modelId
stringlengths 4
112
| lastModified
stringlengths 24
24
| tags
list | pipeline_tag
stringclasses 21
values | files
list | publishedBy
stringlengths 2
37
| downloads_last_month
int32 0
9.44M
| library
stringclasses 15
values | modelCard
large_stringlengths 0
100k
|
|---|---|---|---|---|---|---|---|---|
jky594176/BART2_GRU | 2021-05-31T19:41:21.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 9 | transformers | |
jky594176/recipe_BART1 | 2021-05-30T15:15:52.000Z | [
"pytorch",
"bart",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 15 | transformers | |
jky594176/recipe_BART1_GRU | 2021-05-31T05:50:11.000Z | [
"pytorch",
"bart",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 32 | transformers | |
jky594176/recipe_BART1_NN | 2021-05-30T15:16:55.000Z | [
"pytorch",
"bart",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 13 | transformers | |
jky594176/recipe_BART2 | 2021-05-31T21:04:14.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 12 | transformers | |
jky594176/recipe_BART2GRU_0601_BCElogitsloss | 2021-06-01T05:28:27.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 5 | transformers | |
jky594176/recipe_BART2GRU_0601_BCEloss | 2021-06-01T05:29:42.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 5 | transformers | |
jky594176/recipe_GPT2 | 2021-05-30T16:46:45.000Z | [
"pytorch",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"added_tokens.json",
"config.json",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jky594176 | 43 | transformers | |
jky594176/recipe_bart2_v2 | 2021-05-31T21:02:14.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 6 | transformers | |
jky594176/recipe_bart2_v3 | 2021-06-01T06:37:51.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jky594176 | 6 | transformers | |
jnz/electra-ka-anti-gov | 2021-03-30T14:03:28.000Z | [
"pytorch",
"electra",
"text-classification",
"transformers"
]
| text-classification | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jnz | 11 | transformers | |
jnz/electra-ka-anti-opo | 2021-03-30T14:04:36.000Z | [
"pytorch",
"electra",
"text-classification",
"transformers"
]
| text-classification | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jnz | 6 | transformers | |
jnz/electra-ka-discrediting | 2021-03-30T14:01:41.000Z | [
"pytorch",
"electra",
"text-classification",
"transformers"
]
| text-classification | [
".gitattributes",
"config.json",
"pytorch_model.bin"
]
| jnz | 6 | transformers | |
jnz/electra-ka-fake-news-tagging | 2020-11-15T20:41:39.000Z | [
"pytorch",
"electra",
"text-classification",
"transformers"
]
| text-classification | [
".gitattributes",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.txt"
]
| jnz | 23 | transformers | |
jnz/electra-ka | 2020-12-12T21:53:36.000Z | [
"pytorch",
"electra",
"transformers"
]
| [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"vocab.txt"
]
| jnz | 21 | transformers | ### electra-ka is first of its kind, Transformer based, open source Georgian language model.
The model is trained on 33GB of Georgian text collected from 4854621 pages in commoncrowl archive.
|
|
joeddav/bart-large-mnli-yahoo-answers | 2021-06-14T10:44:33.000Z | [
"pytorch",
"jax",
"bart",
"text-classification",
"en",
"dataset:yahoo-answers",
"arxiv:1909.00161",
"transformers",
"zero-shot-classification",
"pipeline_tag:zero-shot-classification"
]
| zero-shot-classification | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joeddav | 4,325 | transformers | ---
language: en
tags:
- text-classification
- pytorch
datasets:
- yahoo-answers
pipeline_tag: zero-shot-classification
---
# bart-lage-mnli-yahoo-answers
## Model Description
This model takes [facebook/bart-large-mnli](https://huggingface.co/facebook/bart-large-mnli) and fine-tunes it on Yahoo Answers topic classification. It can be used to predict whether a topic label can be assigned to a given sequence, whether or not the label has been seen before.
You can play with an interactive demo of this zero-shot technique with this model, as well as the non-finetuned [facebook/bart-large-mnli](https://huggingface.co/facebook/bart-large-mnli), [here](https://huggingface.co/zero-shot/).
## Intended Usage
This model was fine-tuned on topic classification and will perform best at zero-shot topic classification. Use `hypothesis_template="This text is about {}."` as this is the template used during fine-tuning.
For settings other than topic classification, you can use any model pre-trained on MNLI such as [facebook/bart-large-mnli](https://huggingface.co/facebook/bart-large-mnli) or [roberta-large-mnli](https://huggingface.co/roberta-large-mnli) with the same code as written below.
#### With the zero-shot classification pipeline
The model can be used with the `zero-shot-classification` pipeline like so:
```python
from transformers import pipeline
nlp = pipeline("zero-shot-classification", model="joeddav/bart-large-mnli-yahoo-answers")
sequence_to_classify = "Who are you voting for in 2020?"
candidate_labels = ["Europe", "public health", "politics", "elections"]
hypothesis_template = "This text is about {}."
nlp(sequence_to_classify, candidate_labels, multi_class=True, hypothesis_template=hypothesis_template)
```
#### With manual PyTorch
```python
# pose sequence as a NLI premise and label as a hypothesis
from transformers import BartForSequenceClassification, BartTokenizer
nli_model = BartForSequenceClassification.from_pretrained('joeddav/bart-large-mnli-yahoo-answers')
tokenizer = BartTokenizer.from_pretrained('joeddav/bart-large-mnli-yahoo-answers')
premise = sequence
hypothesis = f'This text is about {label}.'
# run through model pre-trained on MNLI
x = tokenizer.encode(premise, hypothesis, return_tensors='pt',
max_length=tokenizer.max_len,
truncation_strategy='only_first')
logits = nli_model(x.to(device))[0]
# we throw away "neutral" (dim 1) and take the probability of
# "entailment" (2) as the probability of the label being true
entail_contradiction_logits = logits[:,[0,2]]
probs = entail_contradiction_logits.softmax(dim=1)
prob_label_is_true = probs[:,1]
```
## Training
The model is a pre-trained MNLI classifier further fine-tuned on Yahoo Answers topic classification in the manner originally described in [Yin et al. 2019](https://arxiv.org/abs/1909.00161) and [this blog post](https://joeddav.github.io/blog/2020/05/29/ZSL.html). That is, each sequence is fed to the pre-trained NLI model in place of the premise and each candidate label as the hypothesis, formatted like so: `This text is about {class name}.` For each example in the training set, a true and a randomly-selected false label hypothesis are fed to the model which must predict which labels are valid and which are false.
Since this method studies the ability to classify unseen labels after being trained on a different set of labels, the model is only trained on 5 out of the 10 labels in Yahoo Answers. These are "Society & Culture", "Health", "Computers & Internet", "Business & Finance", and "Family & Relationships".
## Evaluation Results
This model was evaluated with the label-weighted F1 of the _seen_ and _unseen_ labels. That is, for each example the model must predict from one of the 10 corpus labels. The F1 is reported for the labels seen during training as well as the labels unseen during training. We found an F1 score of `.68` and `.72` for the unseen and seen labels, respectively. In order to adjust for the in-vs-out of distribution labels, we subtract a fixed amount of 30% from the normalized probabilities of the _seen_ labels, as described in [Yin et al. 2019](https://arxiv.org/abs/1909.00161) and [our blog post](https://joeddav.github.io/blog/2020/05/29/ZSL.html).
|
joeddav/distilbert-base-uncased-agnews-student | 2021-02-18T20:41:19.000Z | [
"pytorch",
"tf",
"distilbert",
"text-classification",
"en",
"dataset:ag_news",
"transformers",
"tensorflow",
"license:mit"
]
| text-classification | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tf_model.h5",
"tokenizer_config.json",
"training_args.bin",
"vocab.txt"
]
| joeddav | 69 | transformers | ---
language: en
tags:
- text-classification
- pytorch
- tensorflow
datasets:
- ag_news
license: mit
widget:
- text: "Armed conflict has been a near-constant policial and economic burden."
- text: "Tom Brady won his seventh Super Bowl last night."
- text: "Dow falls more than 100 points after disappointing jobs data"
- text: "A new moon has been discovered in Jupter's orbit."
---
# distilbert-base-uncased-agnews-student
## Model Description
This model is distilled from the zero-shot classification pipeline on the unlabeled AG's News dataset using [this
script](https://github.com/huggingface/transformers/tree/master/examples/research_projects/zero-shot-distillation).
It is the result of the demo notebook
[here](https://colab.research.google.com/drive/1mjBjd0cR8G57ZpsnFCS3ngGyo5nCa9ya?usp=sharing), where more details
about the model can be found.
- Teacher model: [roberta-large-mnli](https://huggingface.co/roberta-large-mnli)
- Teacher hypothesis template: `"This text is about {}."`
## Intended Usage
The model can be used like any other model trained on AG's News, but will likely not perform as well as a model
trained with full supervision. It is primarily intended as a demo of how an expensive NLI-based zero-shot model
can be distilled to a more efficient student.
|
joeddav/distilbert-base-uncased-go-emotions-student | 2021-02-19T22:15:52.000Z | [
"pytorch",
"tf",
"distilbert",
"text-classification",
"en",
"dataset:go_emotions",
"transformers",
"tensorflow",
"license:mit"
]
| text-classification | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tf_model.h5",
"tokenizer_config.json",
"vocab.txt"
]
| joeddav | 14,360 | transformers | ---
language: en
tags:
- text-classification
- pytorch
- tensorflow
datasets:
- go_emotions
license: mit
widget:
- text: "I feel lucky to be here."
---
# distilbert-base-uncased-go-emotions-student
## Model Description
This model is distilled from the zero-shot classification pipeline on the unlabeled GoEmotions dataset using [this
script](https://github.com/huggingface/transformers/tree/master/examples/research_projects/zero-shot-distillation).
It was trained with mixed precision for 10 epochs and otherwise used the default script arguments.
## Intended Usage
The model can be used like any other model trained on GoEmotions, but will likely not perform as well as a model
trained with full supervision. It is primarily intended as a demo of how an expensive NLI-based zero-shot model
can be distilled to a more efficient student, allowing a classifier to be trained with only unlabeled data. Note
that although the GoEmotions dataset allow multiple labels per instance, the teacher used single-label
classification to create psuedo-labels.
|
joeddav/xlm-roberta-large-xnli | 2020-12-17T16:39:07.000Z | [
"pytorch",
"tf",
"xlm-roberta",
"text-classification",
"multilingual",
"dataset:multi_nli",
"dataset:xnli",
"arxiv:1911.02116",
"transformers",
"tensorflow",
"license:mit",
"zero-shot-classification",
"pipeline_tag:zero-shot-classification"
]
| zero-shot-classification | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"sentencepiece.bpe.model",
"special_tokens_map.json",
"tf_model.h5",
"tokenizer_config.json"
]
| joeddav | 24,422 | transformers | ---
language: multilingual
tags:
- text-classification
- pytorch
- tensorflow
datasets:
- multi_nli
- xnli
license: mit
pipeline_tag: zero-shot-classification
widget:
- text: "За кого вы голосуете в 2020 году?"
candidate_labels: "politique étrangère, Europe, élections, affaires, politique"
multi_class: true
- text: "لمن تصوت في 2020؟"
candidate_labels: "السياسة الخارجية, أوروبا, الانتخابات, الأعمال, السياسة"
multi_class: true
- text: "2020'de kime oy vereceksiniz?"
candidate_labels: "dış politika, Avrupa, seçimler, ticaret, siyaset"
multi_class: true
---
# xlm-roberta-large-xnli
## Model Description
This model takes [xlm-roberta-large](https://huggingface.co/xlm-roberta-large) and fine-tunes it on a combination of NLI data in 15 languages. It is intended to be used for zero-shot text classification, such as with the Hugging Face [ZeroShotClassificationPipeline](https://huggingface.co/transformers/master/main_classes/pipelines.html#transformers.ZeroShotClassificationPipeline).
## Intended Usage
This model is intended to be used for zero-shot text classification, especially in languages other than English. It is fine-tuned on XNLI, which is a multilingual NLI dataset. The model can therefore be used with any of the languages in the XNLI corpus:
- English
- French
- Spanish
- German
- Greek
- Bulgarian
- Russian
- Turkish
- Arabic
- Vietnamese
- Thai
- Chinese
- Hindi
- Swahili
- Urdu
Since the base model was pre-trained trained on 100 different languages, the
model has shown some effectiveness in languages beyond those listed above as
well. See the full list of pre-trained languages in appendix A of the
[XLM Roberata paper](https://arxiv.org/abs/1911.02116)
For English-only classification, it is recommended to use
[bart-large-mnli](https://huggingface.co/facebook/bart-large-mnli) or
[a distilled bart MNLI model](https://huggingface.co/models?filter=pipeline_tag%3Azero-shot-classification&search=valhalla).
#### With the zero-shot classification pipeline
The model can be loaded with the `zero-shot-classification` pipeline like so:
```python
from transformers import pipeline
classifier = pipeline("zero-shot-classification",
model="joeddav/xlm-roberta-large-xnli")
```
You can then classify in any of the above languages. You can even pass the labels in one language and the sequence to
classify in another:
```python
# we will classify the Russian translation of, "Who are you voting for in 2020?"
sequence_to_classify = "За кого вы голосуете в 2020 году?"
# we can specify candidate labels in Russian or any other language above:
candidate_labels = ["Europe", "public health", "politics"]
classifier(sequence_to_classify, candidate_labels)
# {'labels': ['politics', 'Europe', 'public health'],
# 'scores': [0.9048484563827515, 0.05722189322113991, 0.03792969882488251],
# 'sequence': 'За кого вы голосуете в 2020 году?'}
```
The default hypothesis template is the English, `This text is {}`. If you are working strictly within one language, it
may be worthwhile to translate this to the language you are working with:
```python
sequence_to_classify = "¿A quién vas a votar en 2020?"
candidate_labels = ["Europa", "salud pública", "política"]
hypothesis_template = "Este ejemplo es {}."
classifier(sequence_to_classify, candidate_labels, hypothesis_template=hypothesis_template)
# {'labels': ['política', 'Europa', 'salud pública'],
# 'scores': [0.9109585881233215, 0.05954807624220848, 0.029493311420083046],
# 'sequence': '¿A quién vas a votar en 2020?'}
```
#### With manual PyTorch
```python
# pose sequence as a NLI premise and label as a hypothesis
from transformers import AutoModelForSequenceClassification, AutoTokenizer
nli_model = AutoModelForSequenceClassification.from_pretrained('joeddav/xlm-roberta-large-xnli')
tokenizer = AutoTokenizer.from_pretrained('joeddav/xlm-roberta-large-xnli')
premise = sequence
hypothesis = f'This example is {label}.'
# run through model pre-trained on MNLI
x = tokenizer.encode(premise, hypothesis, return_tensors='pt',
truncation_strategy='only_first')
logits = nli_model(x.to(device))[0]
# we throw away "neutral" (dim 1) and take the probability of
# "entailment" (2) as the probability of the label being true
entail_contradiction_logits = logits[:,[0,2]]
probs = entail_contradiction_logits.softmax(dim=1)
prob_label_is_true = probs[:,1]
```
## Training
This model was pre-trained on set of 100 languages, as described in
[the original paper](https://arxiv.org/abs/1911.02116). It was then fine-tuned on the task of NLI on the concatenated
MNLI train set and the XNLI validation and test sets. Finally, it was trained for one additional epoch on only XNLI
data where the translations for the premise and hypothesis are shuffled such that the premise and hypothesis for
each example come from the same original English example but the premise and hypothesis are of different languages.
|
joelito/bert-base-uncased-sem_eval_2010_task_8 | 2021-05-19T20:50:51.000Z | [
"pytorch",
"tf",
"jax",
"bert",
"text-classification",
"transformers"
]
| text-classification | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"pytorch_model.bin",
"special_tokens_map.json",
"tf_model.h5",
"tokenizer_config.json",
"vocab.txt"
]
| joelito | 16 | transformers | # bert-base-uncased-sem_eval_2010_task_8
Task: sem_eval_2010_task_8
Base Model: bert-base-uncased
Trained for 3 epochs
Batch-size: 6
Seed: 42
Test F1-Score: 0.8 |
joelito/distilbert-based-german-cased-ler | 2020-11-30T12:52:05.000Z | [
"pytorch",
"tf",
"distilbert",
"token-classification",
"transformers"
]
| token-classification | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tf_model.h5",
"tokenizer_config.json",
"vocab.txt"
]
| joelito | 20 | transformers | # distilbert-base-german-cased-ler
Task: ler
Base Model: distilbert-base-german-cased
Trained for 3 epochs
Batch-size: 12
Seed: 42
Test F1-Score: 0.936 |
joelito/gbert-base-ler | 2021-05-19T20:51:41.000Z | [
"pytorch",
"tf",
"bert",
"token-classification",
"transformers"
]
| token-classification | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tf_model.h5",
"tokenizer_config.json",
"vocab.txt"
]
| joelito | 10 | transformers | # gbert-base-ler
Task: ler
Base Model: deepset/gbert-base
Trained for 3 epochs
Batch-size: 6
Seed: 42
Test F1-Score: 0.956 |
johngiorgi/declutr-base | 2021-05-20T17:17:11.000Z | [
"pytorch",
"jax",
"roberta",
"masked-lm",
"arxiv:2006.03659",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| johngiorgi | 1,251 | transformers | # DeCLUTR-base
## Model description
The "DeCLUTR-base" model from our paper: [DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations](https://arxiv.org/abs/2006.03659).
## Intended uses & limitations
The model is intended to be used as a universal sentence encoder, similar to [Google's Universal Sentence Encoder](https://tfhub.dev/google/universal-sentence-encoder/4) or [Sentence Transformers](https://github.com/UKPLab/sentence-transformers).
#### How to use
Please see [our repo](https://github.com/JohnGiorgi/DeCLUTR) for full details. A simple example is shown below.
```python
import torch
from scipy.spatial.distance import cosine
from transformers import AutoModel, AutoTokenizer
# Load the model
tokenizer = AutoTokenizer.from_pretrained("johngiorgi/declutr-base")
model = AutoModel.from_pretrained("johngiorgi/declutr-base")
# Prepare some text to embed
text = [
"A smiling costumed woman is holding an umbrella.",
"A happy woman in a fairy costume holds an umbrella.",
]
inputs = tokenizer(text, padding=True, truncation=True, return_tensors="pt")
# Embed the text
with torch.no_grad():
sequence_output = model(**inputs)[0]
# Mean pool the token-level embeddings to get sentence-level embeddings
embeddings = torch.sum(
sequence_output * inputs["attention_mask"].unsqueeze(-1), dim=1
) / torch.clamp(torch.sum(inputs["attention_mask"], dim=1, keepdims=True), min=1e-9)
# Compute a semantic similarity via the cosine distance
semantic_sim = 1 - cosine(embeddings[0], embeddings[1])
```
### BibTeX entry and citation info
```bibtex
@article{Giorgi2020DeCLUTRDC,
title={DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations},
author={John M Giorgi and Osvald Nitski and Gary D. Bader and Bo Wang},
journal={ArXiv},
year={2020},
volume={abs/2006.03659}
}
``` |
johngiorgi/declutr-sci-base | 2021-05-19T20:52:31.000Z | [
"pytorch",
"jax",
"bert",
"masked-lm",
"arxiv:2006.03659",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.txt"
]
| johngiorgi | 2,583 | transformers | # DeCLUTR-sci-base
## Model description
This is the [allenai/scibert_scivocab_uncased](https://huggingface.co/allenai/scibert_scivocab_uncased) model, with extended pretraining on over 2 million scientific papers from [S2ORC](https://github.com/allenai/s2orc/) using the self-supervised training strategy presented in [DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations](https://arxiv.org/abs/2006.03659).
## Intended uses & limitations
The model is intended to be used as a sentence encoder, similar to [Google's Universal Sentence Encoder](https://tfhub.dev/google/universal-sentence-encoder/4) or [Sentence Transformers](https://github.com/UKPLab/sentence-transformers). It is particularly suitable for scientific text.
#### How to use
Please see [our repo](https://github.com/JohnGiorgi/DeCLUTR) for full details. A simple example is shown below.
```python
import torch
from scipy.spatial.distance import cosine
from transformers import AutoModel, AutoTokenizer
# Load the model
tokenizer = AutoTokenizer.from_pretrained("johngiorgi/declutr-sci-base")
model = AutoModel.from_pretrained("johngiorgi/declutr-sci-base")
# Prepare some text to embed
text = [
"Oncogenic KRAS mutations are common in cancer.",
"Notably, c-Raf has recently been found essential for development of K-Ras-driven NSCLCs.",
]
inputs = tokenizer(text, padding=True, truncation=True, return_tensors="pt")
# Embed the text
with torch.no_grad():
sequence_output = model(**inputs)[0]
# Mean pool the token-level embeddings to get sentence-level embeddings
embeddings = torch.sum(
sequence_output * inputs["attention_mask"].unsqueeze(-1), dim=1
) / torch.clamp(torch.sum(inputs["attention_mask"], dim=1, keepdims=True), min=1e-9)
# Compute a semantic similarity via the cosine distance
semantic_sim = 1 - cosine(embeddings[0], embeddings[1])
```
### BibTeX entry and citation info
```bibtex
@article{Giorgi2020DeCLUTRDC,
title={DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations},
author={John M Giorgi and Osvald Nitski and Gary D. Bader and Bo Wang},
journal={ArXiv},
year={2020},
volume={abs/2006.03659}
}
``` |
johngiorgi/declutr-small | 2021-05-20T17:18:12.000Z | [
"pytorch",
"jax",
"roberta",
"masked-lm",
"arxiv:2006.03659",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| johngiorgi | 2,801 | transformers | # DeCLUTR-small
## Model description
The "DeCLUTR-small" model from our paper: [DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations](https://arxiv.org/abs/2006.03659).
## Intended uses & limitations
The model is intended to be used as a universal sentence encoder, similar to [Google's Universal Sentence Encoder](https://tfhub.dev/google/universal-sentence-encoder/4) or [Sentence Transformers](https://github.com/UKPLab/sentence-transformers).
#### How to use
Please see [our repo](https://github.com/JohnGiorgi/DeCLUTR) for full details. A simple example is shown below.
```python
import torch
from scipy.spatial.distance import cosine
from transformers import AutoModel, AutoTokenizer
# Load the model
tokenizer = AutoTokenizer.from_pretrained("johngiorgi/declutr-small")
model = AutoModel.from_pretrained("johngiorgi/declutr-small")
# Prepare some text to embed
text = [
"A smiling costumed woman is holding an umbrella.",
"A happy woman in a fairy costume holds an umbrella.",
]
inputs = tokenizer(text, padding=True, truncation=True, return_tensors="pt")
# Embed the text
with torch.no_grad():
sequence_output = model(**inputs)[0]
# Mean pool the token-level embeddings to get sentence-level embeddings
embeddings = torch.sum(
sequence_output * inputs["attention_mask"].unsqueeze(-1), dim=1
) / torch.clamp(torch.sum(inputs["attention_mask"], dim=1, keepdims=True), min=1e-9)
# Compute a semantic similarity via the cosine distance
semantic_sim = 1 - cosine(embeddings[0], embeddings[1])
```
### BibTeX entry and citation info
```bibtex
@article{Giorgi2020DeCLUTRDC,
title={DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations},
author={John M Giorgi and Osvald Nitski and Gary D. Bader and Bo Wang},
journal={ArXiv},
year={2020},
volume={abs/2006.03659}
}
``` |
jonasmue/cover-letter-distilgpt2 | 2021-05-23T05:58:55.000Z | [
"pytorch",
"jax",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".DS_Store",
".gitattributes",
"config.json",
"eval_results_lm.txt",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"training_args.bin",
"vocab.json",
"checkpoint-500/optimizer.pt"
]
| jonasmue | 45 | transformers | |
jonasmue/cover-letter-gpt2 | 2021-05-23T06:00:04.000Z | [
"pytorch",
"jax",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".DS_Store",
".gitattributes",
"config.json",
"eval_results_lm.txt",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"training_args.bin",
"vocab.json"
]
| jonasmue | 96 | transformers | |
jonasurth/T5Sum | 2021-01-27T08:53:25.000Z | [
"pytorch",
"t5",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"spiece.model",
"tokenizer_config.json"
]
| jonasurth | 6 | transformers | |
jonatasgrosman/bartuque-bart-base-pretrained-mm-2 | 2021-02-25T23:03:55.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"tokenizer.json"
]
| jonatasgrosman | 6 | transformers | Just a test
|
jonatasgrosman/bartuque-bart-base-pretrained-r-2 | 2021-02-04T00:25:56.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"tokenizer.json"
]
| jonatasgrosman | 11 | transformers | Just a test
|
jonatasgrosman/bartuque-bart-base-pretrained-rm-2 | 2021-02-25T23:07:45.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"tokenizer.json"
]
| jonatasgrosman | 7 | transformers | Just a test
|
jonatasgrosman/bartuque-bart-base-random-r-2 | 2021-02-04T00:27:11.000Z | [
"pytorch",
"bart",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"tokenizer.json"
]
| jonatasgrosman | 8 | transformers | Just a test
|
jonatasgrosman/paraphrase | 2021-05-23T06:01:26.000Z | [
"pytorch",
"jax",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"README.md",
"added_tokens.json",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jonatasgrosman | 49 | transformers | testing
|
jonatasgrosman/wav2vec2-large-english | 2021-06-17T16:54:53.000Z | [
"pytorch",
"wav2vec2",
"en",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 659 | transformers | ---
language: en
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: Wav2Vec2 English by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice en
type: common_voice
args: en
metrics:
- name: Test WER
type: wer
value: 21.16
- name: Test CER
type: cer
value: 9.53
---
# Wav2vec2-Large-English
Fine-tuned [facebook/wav2vec2-large](https://huggingface.co/facebook/wav2vec2-large) on English using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "en"
MODEL_ID = "jonatasgrosman/wav2vec2-large-english"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| "SHE'LL BE ALL RIGHT." | SHE'D BE AL RIGHT |
| SIX | SIX |
| "ALL'S WELL THAT ENDS WELL." | ALL IS WELL THAT ENDS WELL |
| DO YOU MEAN IT? | DO YOU MEAN IT |
| THE NEW PATCH IS LESS INVASIVE THAN THE OLD ONE, BUT STILL CAUSES REGRESSIONS. | THE NEW PATCH IS LESS INVASIVE THAN THE OLD ONE BUT STILL CAUSES REGRESSION |
| HOW IS MOZILLA GOING TO HANDLE AMBIGUITIES LIKE QUEUE AND CUE? | HOW IS MOSTYURLA GOING TO BANDO AMBIHOTIS LIKE YU AND Q |
| "I GUESS YOU MUST THINK I'M KINDA BATTY." | QUESTIONS IN CANTON TE PARC |
| NO ONE NEAR THE REMOTE MACHINE YOU COULD RING? | NO ONE NEAR THE REMOTE MACHINE YOU COULD RING |
| SAUCE FOR THE GOOSE IS SAUCE FOR THE GANDER. | SAUCE FOR THE GOICE IS SAUCE FOR THE GONDER |
| GROVES STARTED WRITING SONGS WHEN SHE WAS FOUR YEARS OLD. | GRAFFES STORTED WRITING SONGS WHEN SHE WAS FOUR YEARS OLD |
## Evaluation
The model can be evaluated as follows on the English (en) test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "en"
MODEL_ID = "jonatasgrosman/wav2vec2-large-english"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
"、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
"『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "ʻ", "ˆ"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-20). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used. Initially, I've tested the model only using the Common Voice dataset. Later I've also tested the model using the LibriSpeech and TIMIT datasets, which are better-behaved datasets than the Common Voice, containing only examples in US English extracted from audiobooks.
---
**Common Voice**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | **19.76%** | **8.60%** |
| jonatasgrosman/wav2vec2-large-english | 21.16% | 9.53% |
| facebook/wav2vec2-large-960h-lv60-self | 22.03% | 10.39% |
| facebook/wav2vec2-large-960h-lv60 | 23.97% | 11.14% |
| facebook/wav2vec2-large-960h | 32.79% | 16.03% |
| boris/xlsr-en-punctuation | 34.81% | 15.51% |
| facebook/wav2vec2-base-960h | 39.86% | 19.89% |
| facebook/wav2vec2-base-100h | 51.06% | 25.06% |
| elgeish/wav2vec2-large-lv60-timit-asr | 59.96% | 34.28% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 66.41% | 36.76% |
| elgeish/wav2vec2-base-timit-asr | 68.78% | 36.81% |
---
**LibriSpeech (clean)**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| facebook/wav2vec2-large-960h-lv60-self | **1.86%** | **0.54%** |
| facebook/wav2vec2-large-960h-lv60 | 2.15% | 0.61% |
| facebook/wav2vec2-large-960h | 2.82% | 0.84% |
| facebook/wav2vec2-base-960h | 3.44% | 1.06% |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | 4.16% | 1.28% |
| facebook/wav2vec2-base-100h | 6.26% | 2.00% |
| jonatasgrosman/wav2vec2-large-english | 8.00% | 2.55% |
| elgeish/wav2vec2-large-lv60-timit-asr | 15.53% | 4.93% |
| boris/xlsr-en-punctuation | 19.28% | 6.45% |
| elgeish/wav2vec2-base-timit-asr | 29.19% | 8.38% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 31.82% | 12.41% |
---
**LibriSpeech (other)**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| facebook/wav2vec2-large-960h-lv60-self | **3.89%** | **1.40%** |
| facebook/wav2vec2-large-960h-lv60 | 4.45% | 1.56% |
| facebook/wav2vec2-large-960h | 6.49% | 2.52% |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | 8.82% | 3.42% |
| facebook/wav2vec2-base-960h | 8.90% | 3.55% |
| jonatasgrosman/wav2vec2-large-english | 13.62% | 5.24% |
| facebook/wav2vec2-base-100h | 13.97% | 5.51% |
| boris/xlsr-en-punctuation | 26.40% | 10.11% |
| elgeish/wav2vec2-large-lv60-timit-asr | 28.39% | 12.08% |
| elgeish/wav2vec2-base-timit-asr | 42.04% | 15.57% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 45.19% | 20.32% |
---
**TIMIT**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| facebook/wav2vec2-large-960h-lv60-self | **5.17%** | **1.33%** |
| facebook/wav2vec2-large-960h-lv60 | 6.24% | 1.54% |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | 6.81% | 2.02% |
| facebook/wav2vec2-large-960h | 9.63% | 2.19% |
| facebook/wav2vec2-base-960h | 11.48% | 2.76% |
| elgeish/wav2vec2-large-lv60-timit-asr | 13.83% | 4.36% |
| jonatasgrosman/wav2vec2-large-english | 13.91% | 4.01% |
| facebook/wav2vec2-base-100h | 16.75% | 4.79% |
| elgeish/wav2vec2-base-timit-asr | 25.40% | 8.16% |
| boris/xlsr-en-punctuation | 25.93% | 9.99% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 51.08% | 19.84% |
|
jonatasgrosman/wav2vec2-large-fr-voxpopuli-french | 2021-06-17T16:14:02.000Z | [
"pytorch",
"wav2vec2",
"fr",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 49 | transformers | ---
language: fr
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: Voxpopuli Wav2Vec2 French by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice fr
type: common_voice
args: fr
metrics:
- name: Test WER
type: wer
value: 19.80
- name: Test CER
type: cer
value: 6.89
---
# Wav2vec2-Large-FR-Voxpopuli-French
Fine-tuned [facebook/wav2vec2-large-fr-voxpopuli](https://huggingface.co/facebook/wav2vec2-large-fr-voxpopuli) on French using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fr"
MODEL_ID = "jonatasgrosman/wav2vec2-large-fr-voxpopuli-french"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| "CE DERNIER A ÉVOLUÉ TOUT AU LONG DE L'HISTOIRE ROMAINE." | CE DERNIER ÉVOLÉ TOUT AU LONG DE L'HISTOIRE ROMAINE |
| CE SITE CONTIENT QUATRE TOMBEAUX DE LA DYNASTIE ACHÉMÉNIDE ET SEPT DES SASSANIDES. | CE SITE CONTIENT QUATRE TOMBEAUX DE LA DYNESTIE ACHÉMÉNIDE ET SEPT DES SASENNIDES |
| "J'AI DIT QUE LES ACTEURS DE BOIS AVAIENT, SELON MOI, BEAUCOUP D'AVANTAGES SUR LES AUTRES." | JAI DIT QUE LES ACTEURS DE BOIS AVAIENT SELON MOI BEAUCOUP DAVANTAGE SUR LES AUTRES |
| LES PAYS-BAS ONT REMPORTÉ TOUTES LES ÉDITIONS. | LE PAYS-BAS ON REMPORTÉ TOUTES LES ÉDITIONS |
| IL Y A MAINTENANT UNE GARE ROUTIÈRE. | IL A MAINTENANT OUSATE DE TIRN |
| HUIT | HUIT |
| DANS L’ATTENTE DU LENDEMAIN, ILS NE POUVAIENT SE DÉFENDRE D’UNE VIVE ÉMOTION | DANS L'ATTENTE DU LENDEMAIN IL NE POUVAIT SE DÉFENDRE DUNE VIVE ÉMOTION |
| LA PREMIÈRE SAISON EST COMPOSÉE DE DOUZE ÉPISODES. | LA PREMIÈRE SAISON EST COMPOSÉE DE DOUZE ÉPISODES |
| ELLE SE TROUVE ÉGALEMENT DANS LES ÎLES BRITANNIQUES. | ELLE SE TROUVE ÉGALEMENT DANS LES ÎLES BRITANNIQUES |
| ZÉRO | ZÉRO ZZ |
## Evaluation
The model can be evaluated as follows on the French (fr) test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fr"
MODEL_ID = "jonatasgrosman/wav2vec2-large-fr-voxpopuli-french"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
"、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
"『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "ʻ", "ˆ"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-16). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-french | **16.86%** | **5.65%** |
| Ilyes/wav2vec2-large-xlsr-53-french | 19.67% | 6.70% |
| jonatasgrosman/wav2vec2-large-fr-voxpopuli-french | 19.80% | 6.89% |
| Nhut/wav2vec2-large-xlsr-french | 24.09% | 8.42% |
| facebook/wav2vec2-large-xlsr-53-french | 25.45% | 10.35% |
| MehdiHosseiniMoghadam/wav2vec2-large-xlsr-53-French | 28.22% | 9.70% |
| Ilyes/wav2vec2-large-xlsr-53-french_punctuation | 29.80% | 11.79% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-fr | 61.06% | 33.31% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-arabic | 2021-05-14T21:42:55.000Z | [
"pytorch",
"wav2vec2",
"ar",
"dataset:common_voice",
"dataset:arabic_speech_corpus",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 65 | transformers | ---
language: ar
datasets:
- common_voice
- arabic_speech_corpus
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Arabic by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice ar
type: common_voice
args: ar
metrics:
- name: Test WER
type: wer
value: 39.59
- name: Test CER
type: cer
value: 18.18
---
# Wav2Vec2-Large-XLSR-53-Arabic
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Arabic using the [Common Voice](https://huggingface.co/datasets/common_voice) and [Arabic Speech Corpus](https://huggingface.co/datasets/arabic_speech_corpus).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ar"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-arabic"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| ألديك قلم ؟ | ألديك قلم |
| ليست هناك مسافة على هذه الأرض أبعد من يوم أمس. | ليست نالك مسافة على هذه الأرض أبعد من يوم الأمس م |
| إنك تكبر المشكلة. | إنك تكبر المشكلة |
| يرغب أن يلتقي بك. | يرغب أن يلتقي بك |
| إنهم لا يعرفون لماذا حتى. | إنهم لا يعرفون لماذا حتى |
| سيسعدني مساعدتك أي وقت تحب. | سيسئدنيمساعدتك أي وقد تحب |
| أَحَبُّ نظريّة علمية إليّ هي أن حلقات زحل مكونة بالكامل من الأمتعة المفقودة. | أحب نظرية علمية إلي هي أن حل قتزح المكوينا بالكامل من الأمت عن المفقودة |
| سأشتري له قلماً. | سأشتري له قلما |
| أين المشكلة ؟ | أين المشكل |
| وَلِلَّهِ يَسْجُدُ مَا فِي السَّمَاوَاتِ وَمَا فِي الْأَرْضِ مِنْ دَابَّةٍ وَالْمَلَائِكَةُ وَهُمْ لَا يَسْتَكْبِرُونَ | ولله يسجد ما في السماوات وما في الأرض من دابة والملائكة وهم لا يستكبرون |
## Evaluation
The model can be evaluated as follows on the Arabic test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ar"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-arabic"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
"、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
"『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "'", "ʻ", "ˆ"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-14). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-arabic | **39.59%** | **18.18%** |
| bakrianoo/sinai-voice-ar-stt | 45.30% | 21.84% |
| othrif/wav2vec2-large-xlsr-arabic | 45.93% | 20.51% |
| kmfoda/wav2vec2-large-xlsr-arabic | 54.14% | 26.07% |
| mohammed/wav2vec2-large-xlsr-arabic | 56.11% | 26.79% |
| anas/wav2vec2-large-xlsr-arabic | 62.02% | 27.09% |
| elgeish/wav2vec2-large-xlsr-53-arabic | 100.00% | 100.56% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn | 2021-05-13T22:33:22.000Z | [
"pytorch",
"wav2vec2",
"zh-CN",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 279 | transformers | ---
language: zh-CN
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Chinese (zh-CN) by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice zh-CN
type: common_voice
args: zh-CN
metrics:
- name: Test WER
type: wer
value: 82.37
- name: Test CER
type: cer
value: 19.03
---
# Wav2Vec2-Large-XLSR-53-Chinese-zh-CN
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Chinese using the [Common Voice](https://huggingface.co/datasets/common_voice), [CSS10](https://github.com/Kyubyong/css10) and [ST-CMDS](http://www.openslr.org/38/).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "zh-CN"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| 宋朝末年年间定居粉岭围。 | 宋朝末年年间定居分定为 |
| 渐渐行动不便 | 建境行动不片 |
| 二十一年去世。 | 二十一年去世 |
| 他们自称恰哈拉。 | 他们自称家哈<unk> |
| 局部干涩的例子包括有口干、眼睛干燥、及阴道干燥。 | 菊物干寺的例子包括有口肝眼睛干照以及阴到干<unk> |
| 嘉靖三十八年，登进士第三甲第二名。 | 嘉靖三十八年登进士第三甲第二名 |
| 这一名称一直沿用至今。 | 这一名称一直沿用是心 |
| 同时乔凡尼还得到包税合同和许多明矾矿的经营权。 | 同时桥凡妮还得到包税合同和许多民繁矿的经营权 |
| 为了惩罚西扎城和塞尔柱的结盟，盟军在抵达后将外城烧毁。 | 为了曾罚西扎城和塞尔素的节盟盟军在抵达后将外曾烧毁 |
| 河内盛产黄色无鱼鳞的鳍射鱼。 | 合类生场环色无鱼林的骑射鱼 |
## Evaluation
The model can be evaluated as follows on the Chinese (zh-CN) test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "zh-CN"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
"、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
"『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "'", "ʻ", "ˆ"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-13). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn | **82.37%** | **19.03%** |
| ydshieh/wav2vec2-large-xlsr-53-chinese-zh-cn-gpt | 84.01% | 20.95% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-dutch | 2021-06-17T23:03:24.000Z | [
"pytorch",
"wav2vec2",
"nl",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 314 | transformers | ---
language: nl
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Dutch by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice nl
type: common_voice
args: nl
metrics:
- name: Test WER
type: wer
value: 13.60
- name: Test CER
type: cer
value: 4.45
---
# Wav2Vec2-Large-XLSR-53-Dutch
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Dutch using the [Common Voice](https://huggingface.co/datasets/common_voice) and [CSS10](https://github.com/Kyubyong/css10).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "nl"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| DE ABORIGINALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË. | DE ABORIGONALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË |
| MIJN TOETSENBORD ZIT VOL STOF | MIJN TOETSEN BORT ZIT VOL STOF. |
| ZE HAD DE BANK BESCHADIGD MET HAAR SKATEBOARD. | ZE HAD DE BANK BESCHADIGD MET HAAR SCHEETBOORD |
| WAAR LAAT JIJ JE ONDERHOUD DOEN? | WAAR LAAT JIJ JE ONDERHOUD DOEN |
| NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN QWERTY TOETSENBORD. | NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN KWERTIETOETSENBORD |
## Evaluation
The model can be evaluated as follows on the Dutch test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "nl"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-21). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-dutch | **13.60%** | **4.45%** |
| wietsedv/wav2vec2-large-xlsr-53-dutch | 16.78% | 5.60% |
| facebook/wav2vec2-large-xlsr-53-dutch | 20.97% | 7.24% |
| nithinholla/wav2vec2-large-xlsr-53-dutch | 21.39% | 7.29% |
| MehdiHosseiniMoghadam/wav2vec2-large-xlsr-53-Dutch | 25.89% | 9.12% |
| simonsr/wav2vec2-large-xlsr-dutch | 38.34% | 13.29% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-english | 2021-06-17T16:44:56.000Z | [
"pytorch",
"wav2vec2",
"en",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 1,019 | transformers | ---
language: en
datasets:
- common_voice
- librispeech_asr
- timit_asr
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 English by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice en
type: common_voice
args: en
metrics:
- name: Test WER
type: wer
value: 19.76
- name: Test CER
type: cer
value: 8.60
---
# Wav2Vec2-Large-XLSR-53-English
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on English using the [Common Voice](https://huggingface.co/datasets/common_voice), [LibriSpeech](https://huggingface.co/datasets/librispeech_asr) and [TIMIT](https://huggingface.co/datasets/timit_asr),.
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "en"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-english"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| "SHE'LL BE ALL RIGHT." | SHE'D BE ALRIGHT |
| SIX | SIX |
| "ALL'S WELL THAT ENDS WELL." | ALL IS WELL THAT ENDS WELL |
| DO YOU MEAN IT? | DO YOU MEAN IT |
| THE NEW PATCH IS LESS INVASIVE THAN THE OLD ONE, BUT STILL CAUSES REGRESSIONS. | THE NEW PATCH IS LESS INVASIVE THAN THE OLD ONE BUT STILL CAUSES REGRESSION |
| HOW IS MOZILLA GOING TO HANDLE AMBIGUITIES LIKE QUEUE AND CUE? | HOW IS MUSILA GOING TO HANDLE ANB HOOTIES LIKE QU AND QU |
| "I GUESS YOU MUST THINK I'M KINDA BATTY." | RISIONAS INCI IN TE BACTY |
| NO ONE NEAR THE REMOTE MACHINE YOU COULD RING? | NO ONE NEAR THE REMOTE MACHINE YOU COULD RING |
| SAUCE FOR THE GOOSE IS SAUCE FOR THE GANDER. | SAUCE FOR THE GUISE IS SAUCE FOR THE GONDER |
| GROVES STARTED WRITING SONGS WHEN SHE WAS FOUR YEARS OLD. | GRAFS STARTED WRITING SOUNDS WHEN SHE WAS FOUR YEARS OLD |
## Evaluation
The model can be evaluated as follows on the English test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "en"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-english"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
"、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
"『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "ʻ", "ˆ"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
# uncomment the following lines to eval using other datasets
# test_dataset = load_dataset("librispeech_asr", "clean", split="test")
# test_dataset = load_dataset("librispeech_asr", "other", split="test")
# test_dataset = load_dataset("timit_asr", split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["file"] if "file" in batch else batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["text"] if "text" in batch else batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-20). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used. Initially, I've tested the model only using the Common Voice dataset. Later I've also tested the model using the LibriSpeech and TIMIT datasets, which are better-behaved datasets than the Common Voice, containing only examples in US English extracted from audiobooks.
---
**Common Voice**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | **19.76%** | **8.60%** |
| jonatasgrosman/wav2vec2-large-english | 21.16% | 9.53% |
| facebook/wav2vec2-large-960h-lv60-self | 22.03% | 10.39% |
| facebook/wav2vec2-large-960h-lv60 | 23.97% | 11.14% |
| facebook/wav2vec2-large-960h | 32.79% | 16.03% |
| boris/xlsr-en-punctuation | 34.81% | 15.51% |
| facebook/wav2vec2-base-960h | 39.86% | 19.89% |
| facebook/wav2vec2-base-100h | 51.06% | 25.06% |
| elgeish/wav2vec2-large-lv60-timit-asr | 59.96% | 34.28% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 66.41% | 36.76% |
| elgeish/wav2vec2-base-timit-asr | 68.78% | 36.81% |
---
**LibriSpeech (clean)**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| facebook/wav2vec2-large-960h-lv60-self | **1.86%** | **0.54%** |
| facebook/wav2vec2-large-960h-lv60 | 2.15% | 0.61% |
| facebook/wav2vec2-large-960h | 2.82% | 0.84% |
| facebook/wav2vec2-base-960h | 3.44% | 1.06% |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | 4.16% | 1.28% |
| facebook/wav2vec2-base-100h | 6.26% | 2.00% |
| jonatasgrosman/wav2vec2-large-english | 8.00% | 2.55% |
| elgeish/wav2vec2-large-lv60-timit-asr | 15.53% | 4.93% |
| boris/xlsr-en-punctuation | 19.28% | 6.45% |
| elgeish/wav2vec2-base-timit-asr | 29.19% | 8.38% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 31.82% | 12.41% |
---
**LibriSpeech (other)**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| facebook/wav2vec2-large-960h-lv60-self | **3.89%** | **1.40%** |
| facebook/wav2vec2-large-960h-lv60 | 4.45% | 1.56% |
| facebook/wav2vec2-large-960h | 6.49% | 2.52% |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | 8.82% | 3.42% |
| facebook/wav2vec2-base-960h | 8.90% | 3.55% |
| jonatasgrosman/wav2vec2-large-english | 13.62% | 5.24% |
| facebook/wav2vec2-base-100h | 13.97% | 5.51% |
| boris/xlsr-en-punctuation | 26.40% | 10.11% |
| elgeish/wav2vec2-large-lv60-timit-asr | 28.39% | 12.08% |
| elgeish/wav2vec2-base-timit-asr | 42.04% | 15.57% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 45.19% | 20.32% |
---
**TIMIT**
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| facebook/wav2vec2-large-960h-lv60-self | **5.17%** | **1.33%** |
| facebook/wav2vec2-large-960h-lv60 | 6.24% | 1.54% |
| jonatasgrosman/wav2vec2-large-xlsr-53-english | 6.81% | 2.02% |
| facebook/wav2vec2-large-960h | 9.63% | 2.19% |
| facebook/wav2vec2-base-960h | 11.48% | 2.76% |
| elgeish/wav2vec2-large-lv60-timit-asr | 13.83% | 4.36% |
| jonatasgrosman/wav2vec2-large-english | 13.91% | 4.01% |
| facebook/wav2vec2-base-100h | 16.75% | 4.79% |
| elgeish/wav2vec2-base-timit-asr | 25.40% | 8.16% |
| boris/xlsr-en-punctuation | 25.93% | 9.99% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-en | 51.08% | 19.84% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-finnish | 2021-04-24T16:59:56.000Z | [
"pytorch",
"wav2vec2",
"fi",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jonatasgrosman | 25 | transformers | ---
language: fi
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Finnish by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice fi
type: common_voice
args: fi
metrics:
- name: Test WER
type: wer
value: 41.60
- name: Test CER
type: cer
value: 8.23
---
# Wav2Vec2-Large-XLSR-53-Finnish
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Finnish using the [Common Voice](https://huggingface.co/datasets/common_voice) and [CSS10](https://github.com/Kyubyong/css10).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fi"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-finnish"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| MYSTEERIMIES OLI OPPINUT MORAALINSA TARUISTA, ELOKUVISTA JA PELEISTÄ. | MYSTEERIMIES OLI OPPINUT MORALINSA TARUISTA ELOKUVISTA JA PELEISTÄ |
| ÄÄNESTIN MIETINNÖN PUOLESTA! | ÄÄNESTIN MIETINNÖN PUOLESTA |
| VAIN TUNTIA AIKAISEMMIN OLIMME MIEHENI KANSSA TUNTENEET SUURINTA ILOA. | PAIN TUNTIA AIKAISEMMIN OLIN MIEHENI KANSSA TUNTENEET SUURINTA ILAA |
| ENSIMMÄISELLE MIEHELLE SAI KOLME LASTA. | ENSIMMÄISELLE MIEHELLE SAI KOLME LASTA |
| ÄÄNESTIN MIETINNÖN PUOLESTA, SILLÄ POHJIMMILTAAN SIINÄ VASTUSTETAAN TÄTÄ SUUNTAUSTA. | ÄÄNESTIN MIETINNÖN PUOLESTA SILLÄ POHJIMMILTAAN SIINÄ VASTOTTETAAN TÄTÄ SUUNTAUSTA |
## Evaluation
The model can be evaluated as follows on the Finnish test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fi"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-finnish"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-21). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| aapot/wav2vec2-large-xlsr-53-finnish | **32.51%** | **5.34%** |
| Tommi/wav2vec2-large-xlsr-53-finnish | 35.22% | 5.81% |
| vasilis/wav2vec2-large-xlsr-53-finnish | 38.24% | 6.49% |
| jonatasgrosman/wav2vec2-large-xlsr-53-finnish | 41.60% | 8.23% |
| birgermoell/wav2vec2-large-xlsr-finnish | 53.51% | 9.18% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-french | 2021-06-17T15:11:28.000Z | [
"pytorch",
"wav2vec2",
"fr",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 724 | transformers | ---
language: fr
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 French by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice fr
type: common_voice
args: fr
metrics:
- name: Test WER
type: wer
value: 16.86
- name: Test CER
type: cer
value: 5.65
---
# Wav2Vec2-Large-XLSR-53-French
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on French using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fr"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-french"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| "CE DERNIER A ÉVOLUÉ TOUT AU LONG DE L'HISTOIRE ROMAINE." | CE DERNIER EST VOLÉ TOUT AU LONG DE L'HISTOIRE ROMAINE |
| CE SITE CONTIENT QUATRE TOMBEAUX DE LA DYNASTIE ACHÉMÉNIDE ET SEPT DES SASSANIDES. | CE SITE CONTIENT QUATRE TOMBEAUX DE LA DYNASTIE ASHÉMÉNIDE ET SEPT DES SASANNIDES |
| "J'AI DIT QUE LES ACTEURS DE BOIS AVAIENT, SELON MOI, BEAUCOUP D'AVANTAGES SUR LES AUTRES." | JAI DIT QUE LES ACTEURS DE BOIS AVAIENT SELON MOI BEAUCOUP DAVANTAGES SUR LES AUTRES |
| LES PAYS-BAS ONT REMPORTÉ TOUTES LES ÉDITIONS. | LE PAYS-BAS AN REMPORTAIT TOUTES LES ÉDITIONS |
| IL Y A MAINTENANT UNE GARE ROUTIÈRE. | IL A MA ANDIN GARD DETIRON |
| HUIT | HUIT |
| DANS L’ATTENTE DU LENDEMAIN, ILS NE POUVAIENT SE DÉFENDRE D’UNE VIVE ÉMOTION | DANS L'ATTENTE DU LENDEMAIN IL NE POUVAIT SE DÉFENDRE D'UNE VIVE ÉMOTION |
| LA PREMIÈRE SAISON EST COMPOSÉE DE DOUZE ÉPISODES. | LA PREMIÈRE SAISON EST COMPOSÉE DE DOUX ÉPISODES |
| ELLE SE TROUVE ÉGALEMENT DANS LES ÎLES BRITANNIQUES. | ELLE SE TROUVE ÉGALEMENT DANS LES ÎLES BRITANNIQUES |
| ZÉRO | ZÉRO |
## Evaluation
The model can be evaluated as follows on the French test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fr"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-french"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
"、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
"『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "ʻ", "ˆ"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-16). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-french | **16.86%** | **5.65%** |
| Ilyes/wav2vec2-large-xlsr-53-french | 19.67% | 6.70% |
| jonatasgrosman/wav2vec2-large-fr-voxpopuli-french | 19.80% | 6.89% |
| Nhut/wav2vec2-large-xlsr-french | 24.09% | 8.42% |
| facebook/wav2vec2-large-xlsr-53-french | 25.45% | 10.35% |
| MehdiHosseiniMoghadam/wav2vec2-large-xlsr-53-French | 28.22% | 9.70% |
| Ilyes/wav2vec2-large-xlsr-53-french_punctuation | 29.80% | 11.79% |
| facebook/wav2vec2-base-10k-voxpopuli-ft-fr | 61.06% | 33.31% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-german | 2021-06-17T23:22:27.000Z | [
"pytorch",
"wav2vec2",
"de",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 394 | transformers | ---
language: de
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 German by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice de
type: common_voice
args: de
metrics:
- name: Test WER
type: wer
value: 11.85
- name: Test CER
type: cer
value: 3.17
---
# Wav2Vec2-Large-XLSR-53-German
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on German using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "de"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-german"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| ZIEHT EUCH BITTE DRAUSSEN DIE SCHUHE AUS. | ZIEHT EUCH BITTE DRAUSSEN DIE SCHUHE AUS |
| ES KOMMT ZUM SHOWDOWN IN GSTAAD. | ES GRONTEHILSCHONDEBAR ENBESTACDEN |
| IHRE FOTOSTRECKEN ERSCHIENEN IN MODEMAGAZINEN WIE DER VOGUE, HARPER’S BAZAAR UND MARIE CLAIRE. | IHRE FROTESTRECKEN ERSCHIENEN IN MODEMAGAZINEN WIE DER VOLKE-APERS BASAR VAREQER |
| FELIPE HAT EINE AUCH FÜR MONARCHEN UNGEWÖHNLICH LANGE TITELLISTE. | FIELIPPE HATE EINE AUCH FÜR MONACHEN UNGEWÖHNLICH LANGE TITELLISTE |
| ER WURDE ZU EHREN DES REICHSKANZLERS OTTO VON BISMARCK ERRICHTET. | ER WURDE ZU EHREN DES REICHSKANZLERS OTTO VON BISMARK ERRICHTET |
## Evaluation
The model can be evaluated as follows on the German test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "de"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-german"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-german | **11.85%** | **3.17%** |
| maxidl/wav2vec2-large-xlsr-german | 13.10% | 3.64% |
| marcel/wav2vec2-large-xlsr-53-german | 15.97% | 4.37% |
| flozi00/wav2vec-xlsr-german | 16.13% | 4.33% |
| facebook/wav2vec2-large-xlsr-53-german | 17.15% | 5.79% |
| MehdiHosseiniMoghadam/wav2vec2-large-xlsr-53-German | 19.31% | 5.41% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-greek | 2021-04-24T17:01:54.000Z | [
"pytorch",
"wav2vec2",
"el",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 12 | transformers | ---
language: el
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Greek by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice el
type: common_voice
args: el
metrics:
- name: Test WER
type: wer
value: 14.39
- name: Test CER
type: cer
value: 4.18
---
# Wav2Vec2-Large-XLSR-53-Greek
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Greek using the [Common Voice](https://huggingface.co/datasets/common_voice) and [CSS10](https://github.com/Kyubyong/css10).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "el"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-greek"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| ΤΟ ΒΑΣΙΛΌΠΟΥΛΟ, ΠΟΥ ΜΟΙΆΖΕΙ ΛΕΟΝΤΑΡΆΚΙ ΚΑΙ ΑΕΤΟΥΔΆΚΙ | ΤΟ ΒΑΣΙΛΌΠΟΥΛΟ ΠΟΥ ΜΙΑ ΣΥΛΕΥΕΝΤΑΓΑΡΚΉ ΚΑΙ ΑΙΤΟΥΔΆΚΙ Ο |
| ΣΥΝΆΜΑ ΞΕΠΡΌΒΑΛΑΝ ΑΠΌ ΜΈΣΑ ΑΠΌ ΤΑ ΔΈΝΤΡΑ, ΔΕΞΙΆ, ΑΡΜΑΤΩΜΈΝΟΙ ΚΑΒΑΛΑΡΈΟΙ. | ΣΥΝΆΜΑ ΞΕΠΡΌΒΑΛΑΝ ΑΠΌ ΜΈΣΑ ΑΠΌ ΤΑ ΔΈΝΤΡΑ ΒΕΞΙΆ ΑΡΜΑΤΩΜΈΝΟΙ ΚΑΒΑΛΑΡΈΟ |
| ΤΑ ΣΥΣΚΕΥΑΣΜΈΝΑ ΒΙΟΛΟΓΙΚΆ ΛΑΧΑΝΙΚΆ ΔΕΝ ΠΕΡΙΈΧΟΥΝ ΣΥΝΤΗΡΗΤΙΚΆ ΚΑΙ ΟΡΜΌΝΕΣ | ΤΑ ΣΥΣΚΕΠΑΣΜΈΝΑ ΒΙΟΛΟΓΙΚΆ ΛΑΧΑΝΙΚΆ ΔΕΝ ΠΕΡΙΈΧΟΥΝ ΣΙΝΤΗΡΗΤΙΚΆ ΚΑΙ ΟΡΜΏΝΕΣ |
| ΑΚΟΛΟΥΘΉΣΕΤΕ ΜΕ! | ΑΚΟΛΟΥΘΉΣΤΕ ΜΕ |
| ΚΑΙ ΠΟΎ ΜΠΟΡΏ ΝΑ ΤΟΝ ΒΡΩ; | ΊΤΙ ΤΟΡΟ ΝΟ ΤΙ ΕΒΡΩ |
## Evaluation
The model can be evaluated as follows on the Greek test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "el"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-greek"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| lighteternal/wav2vec2-large-xlsr-53-greek | **10.13%** | **2.66%** |
| jonatasgrosman/wav2vec2-large-xlsr-53-greek | 14.39% | 4.18 |
| vasilis/wav2vec2-large-xlsr-53-greek | 19.09% | 5.88% |
| PereLluis13/wav2vec2-large-xlsr-53-greek | 20.16% | 5.71% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-hungarian | 2021-04-24T17:02:34.000Z | [
"pytorch",
"wav2vec2",
"hu",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 16 | transformers | ---
language: hu
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Hungarian by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice hu
type: common_voice
args: hu
metrics:
- name: Test WER
type: wer
value: 31.40
- name: Test CER
type: cer
value: 6.20
---
# Wav2Vec2-Large-XLSR-53-Hungarian
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Hungarian using the [Common Voice](https://huggingface.co/datasets/common_voice) and [CSS10](https://github.com/Kyubyong/css10).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "hu"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-hungarian"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| BÜSZKÉK VAGYUNK A MAGYAR EMBEREK NAGYSZERŰ SZELLEMI ALKOTÁSAIRA. | BÜSZKÉK VAGYUNK A MAGYAR EMBEREK NAGYSZERŰ SZELLEMI ALKOTÁSAIRE |
| A NEMZETSÉG TAGJAI KÖZÜL EZT TERMESZTIK A LEGSZÉLESEBB KÖRBEN ÍZLETES TERMÉSÉÉRT. | A NEMZETSÉG TAGJAI KÖZÜL ESZSZERMESZTIK A LEGSZELESEBB KÖRBEN IZLETES TERMÉSSÉÉRT |
| A VÁROSBA VÁGYÓDOTT A LEGJOBBAN, ÉPPEN MERT ODA NEM JUTHATOTT EL SOHA. | A VÁROSBA VÁGYÓDOTT A LEGJOBBAN ÉPPEN MERT ODA NEM JUTHATOTT EL SOHA |
| SÍRJA MÁRA MEGSEMMISÜLT. | SIMGI A MANDO MEG SEMMICSEN |
| MINDEN ZENESZÁMOT DRÁGAKŐNEK NEVEZETT. | MINDEN ZENA SZÁMODRAGAKŐNEK NEVEZETT |
## Evaluation
The model can be evaluated as follows on the Hungarian test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "hu"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-hungarian"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-hungarian | **31.40%** | **6.20%** |
| anton-l/wav2vec2-large-xlsr-53-hungarian | 42.39% | 9.39% |
| gchhablani/wav2vec2-large-xlsr-hu | 46.42% | 10.04% |
| birgermoell/wav2vec2-large-xlsr-hungarian | 46.93% | 10.31% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-italian | 2021-06-17T23:47:35.000Z | [
"pytorch",
"wav2vec2",
"it",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 25 | transformers | ---
language: it
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Italian by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice it
type: common_voice
args: it
metrics:
- name: Test WER
type: wer
value: 11.90
- name: Test CER
type: cer
value: 2.94
---
# Wav2Vec2-Large-XLSR-53-Italian
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Italian using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "it"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-italian"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| POI LEI MORÌ. | POI LEI MORÌ |
| IL LIBRO HA SUSCITATO MOLTE POLEMICHE A CAUSA DEI SUOI CONTENUTI. | IL LIBRO HA SUSCITATO MOLTE POLEMICHE A CAUSA DEI SUOI CONTENUTI |
| "FIN DALL'INIZIO LA SEDE EPISCOPALE È STATA IMMEDIATAMENTE SOGGETTA ALLA SANTA SEDE." | FIN DALL'INIZIO LA SEDE EPISCOPALE È STATA IMMEDIATAMENTE SOGGETTA ALLA SANTA SEDE |
| IL VUOTO ASSOLUTO? | IL VUOTO ASSOLUTO |
| DOPO ALCUNI ANNI, EGLI DECISE DI TORNARE IN INDIA PER RACCOGLIERE ALTRI INSEGNAMENTI. | DOPO ALCUNI ANNI EGLI DECISE DI TORNARE IN INDIA PER RACCOGLIERE ALTRI INSEGNAMENTI |
## Evaluation
The model can be evaluated as follows on the Italian test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "it"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-italian"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-21). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-italian | **11.90%** | **2.94%** |
| joorock12/wav2vec2-large-xlsr-italian | 12.60% | 3.18% |
| gchhablani/wav2vec2-large-xlsr-it | 12.99% | 3.11% |
| facebook/wav2vec2-large-xlsr-53-italian | 22.08% | 6.36% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-japanese | 2021-05-13T23:56:39.000Z | [
"pytorch",
"wav2vec2",
"ja",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 357 | transformers | ---
language: ja
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Japanese by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice ja
type: common_voice
args: ja
metrics:
- name: Test WER
type: wer
value: 81.80
- name: Test CER
type: cer
value: 20.16
---
# Wav2Vec2-Large-XLSR-53-Japanese
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Japanese using the [Common Voice](https://huggingface.co/datasets/common_voice), [CSS10](https://github.com/Kyubyong/css10) and [JSUT](https://sites.google.com/site/shinnosuketakamichi/publication/jsut).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ja"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-japanese"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| 祖母は、おおむね機嫌よく、サイコロをころがしている。 | 人母は重にきね起くさいがしている |
| 財布をなくしたので、交番へ行きます。 | 財布をなく手端ので勾番へ行きます |
| 飲み屋のおやじ、旅館の主人、医者をはじめ、交際のある人にきいてまわったら、みんな、私より収入が多いはずなのに、税金は安い。 | ノ宮屋のお親じ旅館の主に医者をはじめ交際のアル人トに聞いて回ったらみんな私より収入が多いはなうに税金は安い |
| 新しい靴をはいて出かけます。 | だらしい靴をはいて出かけます |
| このためプラズマ中のイオンや電子の持つ平均運動エネルギーを温度で表現することがある | このためプラズマ中のイオンや電子の持つ平均運動エネルギーを温度で表弁することがある |
| 松井さんはサッカーより野球のほうが上手です。 | 松井さんはサッカーより野球のほうが上手です |
| 新しいお皿を使います。 | 新しいお皿を使います |
| 結婚以来三年半ぶりの東京も、旧友とのお酒も、夜行列車も、駅で寝て、朝を待つのも久しぶりだ。 | 結婚ル二来三年半降りの東京も吸とのお酒も野越者も駅で寝て朝を待つの久しぶりた |
| これまで、少年野球、ママさんバレーなど、地域スポーツを支え、市民に密着してきたのは、無数のボランティアだった。 | これまで少年野球<unk>三バレーなど地域スポーツを支え市民に満着してきたのは娘数のボランティアだった |
| 靴を脱いで、スリッパをはきます。 | 靴を脱いでスイパーをはきます |
## Evaluation
The model can be evaluated as follows on the Japanese test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ja"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-japanese"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
"、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
"『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "'", "ʻ", "ˆ"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-05-10). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-japanese | **81.80%** | **20.16%** |
| vumichien/wav2vec2-large-xlsr-japanese | 1108.86% | 23.40% |
| qqhann/w2v_hf_jsut_xlsr53 | 1012.18% | 70.77% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-persian | 2021-04-24T17:04:22.000Z | [
"pytorch",
"wav2vec2",
"fa",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 426 | transformers | ---
language: fa
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Persian by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice fa
type: common_voice
args: fa
metrics:
- name: Test WER
type: wer
value: 30.12
- name: Test CER
type: cer
value: 7.37
---
# Wav2Vec2-Large-XLSR-53-Persian
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Persian using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fa"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-persian"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| از مهمونداری کنار بکشم | از مهمانداری کنار بکشم |
| برو از مهرداد بپرس. | برو از ماقدعاد به پرس |
| خب ، تو چیكار می كنی؟ | خوب تو چیکار می کنی |
| مسقط پایتخت عمان در عربی به معنای محل سقوط است | مسقط پایتخت عمان در عربی به بعنای محل سقوط است |
| آه، نه اصلاُ! | اهنه اصلا |
## Evaluation
The model can be evaluated as follows on the Persian test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "fa"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-persian"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-persian | **30.12%** | **7.37%** |
| m3hrdadfi/wav2vec2-large-xlsr-persian-v2 | 33.85% | 8.79% |
| m3hrdadfi/wav2vec2-large-xlsr-persian | 34.37% | 8.98% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-polish | 2021-04-24T17:05:17.000Z | [
"pytorch",
"wav2vec2",
"pl",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 65 | transformers | ---
language: pl
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Polish by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice pl
type: common_voice
args: pl
metrics:
- name: Test WER
type: wer
value: 11.17
- name: Test CER
type: cer
value: 2.87
---
# Wav2Vec2-Large-XLSR-53-Polish
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Polish using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "pl"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-polish"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| """CZY DRZWI BYŁY ZAMKNIĘTE?""" | CZY DRZWI BYŁY ZAMKNIĘTE |
| GDZIEŻ TU POWÓD DO WYRZUTÓW? | GDZIEŻ TO POWÓD DO WYŻYTÓW |
| """O TEM JEDNAK NIE BYŁO MOWY.""" | O TEM JEDNAK NIE BYŁO MOWY N |
| LUBIĘ GO. | LUBIĘ GO |
| — TO MI NIE POMAGA. | TO MI NIE POMAGA |
## Evaluation
The model can be evaluated as follows on the Polish test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "pl"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-polish"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-polish | **11.17%** | **2.87%** |
| facebook/wav2vec2-large-xlsr-53-polish | 20.17% | 5.38% |
| alexcleu/wav2vec2-large-xlsr-polish | 21.72% | 5.17% |
| mbien/wav2vec2-large-xlsr-polish | 22.93% | 5.13% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-portuguese | 2021-06-17T23:33:23.000Z | [
"pytorch",
"wav2vec2",
"pt",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 444 | transformers | ---
language: pt
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Portuguese by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice pt
type: common_voice
args: pt
metrics:
- name: Test WER
type: wer
value: 11.83
- name: Test CER
type: cer
value: 4.09
---
# Wav2Vec2-Large-XLSR-53-Portuguese
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Portuguese using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "pt"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-portuguese"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| NEM O RADAR NEM OS OUTROS INSTRUMENTOS DETECTARAM O BOMBARDEIRO STEALTH. | NENHUM VADA ME OS SALTOWS INSTRUMENTOS DE TECTERÁM UM BOMBADEIRO STER |
| PEDIR DINHEIRO EMPRESTADO ÀS PESSOAS DA ALDEIA | EDIR DINHEIRO EMPRESTADO ÀS PESSOAS DO ALDEIRA |
| OITO | OITO |
| TRANCÁ-LOS | TRAM COULTS |
| REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA | REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA |
## Evaluation
The model can be evaluated as follows on the Portuguese test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "pt"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-portuguese"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-portuguese | **11.83%** | **4.09%** |
| joorock12/wav2vec2-large-xlsr-portuguese-a | 15.52% | 5.12% |
| joorock12/wav2vec2-large-xlsr-portuguese | 15.95% | 5.31% |
| gchhablani/wav2vec2-large-xlsr-pt | 17.64% | 6.04% |
| Rubens/Wav2Vec2-Large-XLSR-53-a-Portuguese | 19.79% | 6.57% |
| Rubens/Wav2Vec2-Large-XLSR-53-Portuguese | 20.85% | 7.08% |
| facebook/wav2vec2-large-xlsr-53-portuguese | 26.73% | 9.27% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-russian | 2021-04-24T17:06:35.000Z | [
"pytorch",
"wav2vec2",
"ru",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 563 | transformers | ---
language: ru
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Russian by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice ru
type: common_voice
args: ru
metrics:
- name: Test WER
type: wer
value: 16.79
- name: Test CER
type: cer
value: 3.68
---
# Wav2Vec2-Large-XLSR-53-Russian
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Russian using the [Common Voice](https://huggingface.co/datasets/common_voice) and [CSS10](https://github.com/Kyubyong/css10).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ru"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-russian"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| ОН РАБОТАТЬ, А ЕЕ НЕ УДЕРЖАТЬ НИКАК — БЕГАЕТ ЗА КЛЁШЕМ КАЖДОГО БУЛЬВАРНИКА. | ОН РАБОТАТЬ А ЕЕ НЕ УДЕРЖАТНИКАК БЕГАЕТ ЗА КЛЕШОМ КАЖДОГО БУЛЬВАРНИКА |
| ЕСЛИ НЕ БУДЕТ ВОЗРАЖЕНИЙ, Я БУДУ СЧИТАТЬ, ЧТО АССАМБЛЕЯ СОГЛАСНА С ЭТИМ ПРЕДЛОЖЕНИЕМ. | ЕСЛИ НЕ БУДЕТ ВОЗРАЖЕНИЙ Я БУДУ СЧИТАТЬ ЧТО АССАМБЛЕЯ СОГЛАСНА С ЭТИМ ПРЕДЛОЖЕНИЕМ |
| ПАЛЕСТИНЦАМ НЕОБХОДИМО СНАЧАЛА УСТАНОВИТЬ МИР С ИЗРАИЛЕМ, А ЗАТЕМ ДОБИВАТЬСЯ ПРИЗНАНИЯ ГОСУДАРСТВЕННОСТИ. | ПАЛЕСТИНЦАМ НЕОБХОДИМО СНАЧАЛА УСТАНОВИТЬ С НИ МИР С ИЗРАИЛЕМ А ЗАТЕМ ДОБИВАТЬСЯ ПРИЗНАНИЯ ГОСУДАРСТВЕННОВСКИЙ |
| У МЕНЯ БЫЛО ТАКОЕ ЧУВСТВО, ЧТО ЧТО-ТО ТАКОЕ ОЧЕНЬ ВАЖНОЕ Я ПРИБАВЛЯЮ. | У МЕНЯ БЫЛО ТАКОЕ ЧУВСТВО ЧТО ЧТО-ТО ТАКОЕ ОЧЕНЬ ВАЖНОЕ Е ПРЕДБАВЛЯЕТ |
| ТОЛЬКО ВРЯД ЛИ ПОЙМЕТ. | ТОЛЬКО ВРЯД ЛИ ПОЙМЕТ |
## Evaluation
The model can be evaluated as follows on the Russian test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "ru"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-russian"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-russian | **16.79%** | **3.68%** |
| anton-l/wav2vec2-large-xlsr-53-russian | 19.49% | 4.15% |
|
jonatasgrosman/wav2vec2-large-xlsr-53-spanish | 2021-06-17T18:13:26.000Z | [
"pytorch",
"wav2vec2",
"es",
"dataset:common_voice",
"transformers",
"audio",
"automatic-speech-recognition",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"vocab.json"
]
| jonatasgrosman | 2,193 | transformers | ---
language: es
datasets:
- common_voice
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Spanish by Jonatas Grosman
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice es
type: common_voice
args: es
metrics:
- name: Test WER
type: wer
value: 10.07
- name: Test CER
type: cer
value: 3.04
---
# Wav2Vec2-Large-XLSR-53-Spanish
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Spanish using the [Common Voice](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "es"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-spanish"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| HABITA EN AGUAS POCO PROFUNDAS Y ROCOSAS. | HABITAN AGUAS POCO PROFUNDAS Y ROCOSAS |
| OPERA PRINCIPALMENTE VUELOS DE CABOTAJE Y REGIONALES DE CARGA. | OPERA PRINCIPALMENTE VUELO DE CARBOTAJES Y REGIONALES DE CARGAN |
| PARA VISITAR CONTACTAR PRIMERO CON LA DIRECCIÓN. | PARA VISITAR CONTACTAR PRIMERO CON LA DIRECCIÓN |
| TRES | TRES |
| REALIZÓ LOS ESTUDIOS PRIMARIOS EN FRANCIA, PARA CONTINUAR LUEGO EN ESPAÑA. | REALIZÓ LOS ESTUDIOS PRIMARIOS EN FRANCIA PARA CONTINUAR LUEGO EN ESPAÑA |
## Evaluation
The model can be evaluated as follows on the Spanish test data of Common Voice.
```python
import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "es"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-spanish"
DEVICE = "cuda"
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
"؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
"=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
test_dataset = load_dataset("common_voice", LANG_ID, split="test")
wer = load_metric("wer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/wer.py
cer = load_metric("cer.py") # https://github.com/jonatasgrosman/wav2vec2-sprint/blob/main/cer.py
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = re.sub(chars_to_ignore_regex, "", batch["sentence"]).upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to(DEVICE), attention_mask=inputs.attention_mask.to(DEVICE)).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
predictions = [x.upper() for x in result["pred_strings"]]
references = [x.upper() for x in result["sentence"]]
print(f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
print(f"CER: {cer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}")
```
**Test Result**:
In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2021-04-22). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.
| Model | WER | CER |
| ------------- | ------------- | ------------- |
| jonatasgrosman/wav2vec2-large-xlsr-53-spanish | **10.07%** | **3.04%** |
| pcuenq/wav2vec2-large-xlsr-53-es | 10.55% | 3.20% |
| facebook/wav2vec2-large-xlsr-53-spanish | 16.99% | 5.40% |
| mrm8488/wav2vec2-large-xlsr-53-spanish | 19.20% | 5.96% |
|
jonathangh/gpt2_pretrain | 2021-03-02T01:42:01.000Z | []
| [
".gitattributes"
]
| jonathangh | 0 | |||
jonx18/DialoGPT-small-Creed-Odyssey | 2021-05-23T06:02:34.000Z | [
"pytorch",
"jax",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jonx18 | 11 | transformers | # Summary
The app was conceived with the idea of recreating and generate new dialogs for existing games.
In order to generate a dataset for training the steps followed were:
1. Download from [Assassins Creed Fandom Wiki](https://assassinscreed.fandom.com/wiki/Special:Export) from the category "Memories relived using the Animus HR-8.5".
2. Keep only text elements from XML.
3. Keep only the dialog section.
4. Parse wikimarkup with [wikitextparser](https://pypi.org/project/wikitextparser/).
5. Clean description of dialog's context.
Due to the small size of the dataset obtained, a transfer learning approach was considered based on a pretrained ["Dialog GPT" model](https://huggingface.co/microsoft/DialoGPT-small). |
joorock12/model-sid-voxforge-cetuc-0 | 2021-04-09T14:05:15.000Z | [
"pytorch",
"wav2vec2",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 6 | transformers | ||
joorock12/model-sid-voxforge-cetuc-1 | 2021-04-05T14:06:32.000Z | [
"pytorch",
"wav2vec2",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 8 | transformers | ||
joorock12/model-sid-voxforge-cetuc-2 | 2021-04-05T14:00:19.000Z | [
"pytorch",
"wav2vec2",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 7 | transformers | ||
joorock12/model-sid-voxforge-cv-cetuc-0 | 2021-04-09T15:13:26.000Z | [
"pytorch",
"wav2vec2",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 6 | transformers | ||
joorock12/model-sid-voxforge-cv-cetuc-1 | 2021-04-14T01:22:49.000Z | [
"pytorch",
"wav2vec2",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 7 | transformers | ||
joorock12/model-sid-voxforge-cv-cetuc-2 | 2021-04-13T23:19:03.000Z | [
"pytorch",
"wav2vec2",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 8 | transformers | ||
joorock12/wav2vec2-cetuc-sid-voxforge-mls-0 | 2021-06-12T11:21:32.000Z | [
"pytorch",
"wav2vec2",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 4 | transformers | ||
joorock12/wav2vec2-large-100k-voxpopuli-pt | 2021-05-18T18:29:08.000Z | [
"pytorch",
"wav2vec2",
"pt",
"dataset:common_voice",
"transformers",
"audio",
"speech",
"apache-2.0",
"portuguese-speech-corpus",
"automatic-speech-recognition",
"PyTorch",
"voxpopuli",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 133 | transformers | ---
language: pt
datasets:
- common_voice
metrics:
- wer
tags:
- audio
- speech
- wav2vec2
- pt
- apache-2.0
- portuguese-speech-corpus
- automatic-speech-recognition
- speech
- PyTorch
- voxpopuli
license: apache-2.0
model-index:
- name: JoaoAlvarenga Wav2Vec2 Large 100k VoxPopuli Portuguese
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice pt
type: common_voice
args: pt
metrics:
- name: Test WER
type: wer
value: 19.735723%
---
# Wav2Vec2-Large-100k-VoxPopuli-Portuguese
Fine-tuned [facebook/wav2vec2-large-100k-voxpopuli](https://huggingface.co/facebook/wav2vec2-large-100k-voxpopuli) on Portuguese using the [Common Voice](https://huggingface.co/datasets/common_voice) dataset.
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "pt", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])
```
## Evaluation
The model can be evaluated as follows on the Portuguese test data of Common Voice.
You need to install Enelvo, an open-source spell correction trained with Twitter user posts
`pip install enelvo`
```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
from enelvo import normaliser
import re
test_dataset = load_dataset("common_voice", "pt", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
model.to("cuda")
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\‘\”\�]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
norm = normaliser.Normaliser()
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = [norm.normalise(i) for i in processor.batch_decode(pred_ids)]
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
**Test Result (wer)**: 19.735723%
## Training
The Common Voice `train`, `validation` datasets were used for training.
|
joorock12/wav2vec2-large-xlsr-53-spanish | 2021-03-23T13:09:05.000Z | [
"es",
"dataset:common_voice",
"audio",
"speech",
"wav2vec2",
"apache-2.0",
"spanish-speech-corpus",
"automatic-speech-recognition",
"xlsr-fine-tuning-week",
"PyTorch",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md"
]
| joorock12 | 0 | ---
language: es
datasets:
- common_voice
metrics:
- wer
tags:
- audio
- speech
- wav2vec2
- es
- apache-2.0
- spanish-speech-corpus
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
- PyTorch
license: apache-2.0
model-index:
- name: JoaoAlvarenga XLSR Wav2Vec2 Large 53 Spanish
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice ES
type: common_voice
args: es
metrics:
- name: Test WER
type: wer
value: Training
---
# Wav2Vec2-Large-XLSR-53-Spanish
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Spanish using the [Common Voice](https://huggingface.co/datasets/common_voice) dataset.
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "es", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-53-spanish")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-53-spanish")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
\treturn batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
\tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])
```
## Evaluation
The model can be evaluated as follows on the Portuguese test data of Common Voice.
```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
test_dataset = load_dataset("common_voice", "es", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-53-spanish")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-53-spanish")
model.to("cuda")
chars_to_ignore_regex = '[\\,\\?\\.\\!\\-\\;\\:\\"\\“]' # TODO: adapt this list to include all special characters you removed from the data
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
\tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
\treturn batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
\tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
\twith torch.no_grad():
\t\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
\tbatch["pred_strings"] = processor.batch_decode(pred_ids)
\treturn batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
**Test Result (wer) **: Training
## Training
The Common Voice `train`, `validation` datasets were used for training.
The script used for training can be found at: https://github.com/joaoalvarenga/wav2vec2-large-xlsr-53-spanish/blob/main/fine-tuning.py |
|
joorock12/wav2vec2-large-xlsr-italian | 2021-03-27T13:52:46.000Z | [
"pytorch",
"wav2vec2",
"it",
"dataset:common_voice",
"transformers",
"audio",
"speech",
"apache-2.0",
"portuguese-speech-corpus",
"automatic-speech-recognition",
"xlsr-fine-tuning-week",
"PyTorch",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 50 | transformers | ---
language: it
datasets:
- common_voice
metrics:
- wer
tags:
- audio
- speech
- wav2vec2
- it
- apache-2.0
- portuguese-speech-corpus
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
- PyTorch
license: apache-2.0
model-index:
- name: JoaoAlvarenga XLSR Wav2Vec2 Large 53 Italian
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice it
type: common_voice
args: it
metrics:
- name: Test WER
type: wer
value: 13.914924%
---
# Wav2Vec2-Large-XLSR-53-Italian
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Italian using the [Common Voice](https://huggingface.co/datasets/common_voice) dataset.
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "it", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-italian")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-italian")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])
```
## Evaluation
The model can be evaluated as follows on the Italian test data of Common Voice.
```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
test_dataset = load_dataset("common_voice", "it", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-italian")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-italian")
model.to("cuda")
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\'\�]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
**Test Result (wer)**: 13.914924%
## Training
The Common Voice `train`, `validation` datasets were used for training.
The script used for training can be found at: https://github.com/joaoalvarenga/wav2vec2-large-xlsr-53-italian/blob/main/fine_tuning.py
|
joorock12/wav2vec2-large-xlsr-portuguese-a | 2021-03-27T01:58:48.000Z | [
"pytorch",
"wav2vec2",
"pt",
"dataset:common_voice",
"transformers",
"audio",
"speech",
"apache-2.0",
"portuguese-speech-corpus",
"automatic-speech-recognition",
"xlsr-fine-tuning-week",
"PyTorch",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 167 | transformers | ---
language: pt
datasets:
- common_voice
metrics:
- wer
tags:
- audio
- speech
- wav2vec2
- pt
- apache-2.0
- portuguese-speech-corpus
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
- PyTorch
license: apache-2.0
model-index:
- name: JoaoAlvarenga XLSR Wav2Vec2 Large 53 Portuguese A
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice pt
type: common_voice
args: pt
metrics:
- name: Test WER
type: wer
value: 15.037146%
---
# Wav2Vec2-Large-XLSR-53-Portuguese
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Portuguese using the [Common Voice](https://huggingface.co/datasets/common_voice) dataset.
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "pt", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese-a")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese-a")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])
```
## Evaluation
The model can be evaluated as follows on the Portuguese test data of Common Voice.
```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
test_dataset = load_dataset("common_voice", "pt", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese-a")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese-a")
model.to("cuda")
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\'\�]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
**Test Result (wer)**: 15.037146%
## Training
The Common Voice `train`, `validation` datasets were used for training.
The script used for training can be found at: https://github.com/joaoalvarenga/wav2vec2-large-xlsr-53-portuguese/blob/main/fine-tuning.py
|
joorock12/wav2vec2-large-xlsr-portuguese | 2021-03-29T15:19:27.000Z | [
"pytorch",
"wav2vec2",
"pt",
"dataset:common_voice",
"transformers",
"audio",
"speech",
"apache-2.0",
"portuguese-speech-corpus",
"automatic-speech-recognition",
"xlsr-fine-tuning-week",
"PyTorch",
"license:apache-2.0"
]
| automatic-speech-recognition | [
".gitattributes",
"README.md",
"added_tokens.json",
"config.json",
"preprocessor_config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| joorock12 | 42 | transformers | ---
language: pt
datasets:
- common_voice
metrics:
- wer
tags:
- audio
- speech
- wav2vec2
- pt
- apache-2.0
- portuguese-speech-corpus
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
- PyTorch
license: apache-2.0
model-index:
- name: JoaoAlvarenga XLSR Wav2Vec2 Large 53 Portuguese
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice pt
type: common_voice
args: pt
metrics:
- name: Test WER
type: wer
value: 13.766801%
---
# Wav2Vec2-Large-XLSR-53-Portuguese
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Portuguese using the [Common Voice](https://huggingface.co/datasets/common_voice) dataset.
## Usage
The model can be used directly (without a language model) as follows:
```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "pt", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])
```
## Evaluation
The model can be evaluated as follows on the Portuguese test data of Common Voice.
You need to install Enelvo, an open-source spell correction trained with Twitter user posts
`pip install enelvo`
```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
from enelvo import normaliser
import re
test_dataset = load_dataset("common_voice", "pt", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese-a")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-xlsr-portuguese-a")
model.to("cuda")
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\'\�]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
norm = normaliser.Normaliser()
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = [norm.normalise(i) for i in processor.batch_decode(pred_ids)]
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
**Test Result (wer)**: 13.766801%
## Training
The Common Voice `train`, `validation` datasets were used for training.
The script used for training can be found at: https://github.com/joaoalvarenga/wav2vec2-large-xlsr-53-portuguese/blob/main/fine-tuning.py
|
jordan-m-young/buzz-article-gpt-2 | 2021-05-23T06:03:38.000Z | [
"pytorch",
"jax",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"training_args.bin",
"vocab.json"
]
| jordan-m-young | 41 | transformers | |
jordan-m-young/buzz-title-gpt-2 | 2021-01-10T20:08:33.000Z | []
| [
".gitattributes"
]
| jordan-m-young | 0 | |||
jordimas/julibert | 2021-05-20T17:19:38.000Z | [
"pytorch",
"jax",
"roberta",
"masked-lm",
"ca",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"tokenizer_config.json",
"vocab.json"
]
| jordimas | 797 | transformers | ---
language: ca
---
## Introduction
Download the model here:
* Catalan Roberta model: [julibert-2020-11-10.zip](https://www.softcatala.org/pub/softcatala/julibert/julibert-2020-11-10.zip)
## What's this?
Source code: https://github.com/Softcatala/julibert
* Corpus: Oscar Catalan Corpus (3,8G)
* Model type: Roberta
* Vocabulary size: 50265
* Steps: 500000
|
joseangelatm/PanamianModel | 2021-05-20T17:22:40.000Z | [
"pytorch",
"jax",
"roberta",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"config.json",
"eval_results.txt",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"tokenizer_config.json",
"vocab.json"
]
| joseangelatm | 17 | transformers | |
joseangelatm/spanishpanama | 2021-05-20T17:25:08.000Z | [
"pytorch",
"jax",
"tensorboard",
"roberta",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
".gitignore",
"config.json",
"eval_results.txt",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"run_language_modeling.py",
"special_tokens_map.json",
"tokenizer_config.json",
"training_args.bin",
"vocab.json",
"data/dataset.txt",
"data/dev.txt",
"data/roberta_cached_lm_510_dev.txt",
"data/roberta_cached_lm_510_train.txt",
"data/train.txt",
"runs/Dec16_14-41-57_a11ab1dcd551/events.out.tfevents.1608129717.a11ab1dcd551.554.0"
]
| joseangelatm | 18 | transformers | |
joshuacalloway/csc575finalproject | 2021-03-16T00:46:04.000Z | [
"pytorch",
"distilbert",
"text-classification",
"transformers"
]
| text-classification | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.txt"
]
| joshuacalloway | 11 | transformers | |
josmunpen/mt5-small-spanish-summarization | 2021-06-12T11:30:16.000Z | [
"pytorch",
"mt5",
"seq2seq",
"es",
"dataset:larazonpublico",
"dataset:es",
"transformers",
"summarization",
"spanish",
"license:apache-2.0",
"text2text-generation"
]
| summarization | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"spiece.model",
"tokenizer.json",
"tokenizer_config.json",
"training_args.bin"
]
| josmunpen | 201 | transformers |
---
language:
- es
thumbnail:
tags:
- summarization
- mt5
- spanish
license: apache-2.0
datasets:
- larazonpublico
- es
metrics:
- rouge
widget:
- text: "La Guardia Civil ha desarticulado un grupo organizado dedicado a copiar en los examenes teoricos para la obtencion del permiso de conducir. Para ello, empleaban receptores y camaras de alta tecnologia y operaban desde la misma sede del Centro de examenes de la Direccion General de Trafico (DGT) en Mostoles. Es lo que han llamado la Operacion pinga.
El grupo desarticulado ofrecia el servicio de transporte y tecnologia para copiar y poder aprobar. Por dicho servicio cobraban 1.000 euros. Los investigadores sorprendieron in fraganti a una mujer intentando copiar en el examen. Portaba una chaqueta con dispositivos electronicos ocultos, concretamente un telefono movil al que estaba conectada una camara que habia sido insertada en la parte frontal de la chaqueta para transmitir online el examen y que orientada al ordenador del Centro de Examenes en el que aparecen las preguntas, permitia visualizar las imagenes en otro ordenador alojado en el interior de un vehiculo estacionado en las inmediaciones del centro. En este vehiculo, se encontraban el resto del grupo desarticulado con varios ordenadores portatiles y tablets abiertos y conectados a paginas de test de la DGT para consultar las respuestas. Estos, comunicaban con la mujer que estaba en el aula haciendo el examen a traves de un diminuto receptor bluetooth que portaba en el interior de su oido.
Luis de Lama, portavoz de la Guardia Civil de Trafico destaca que los ciudadanos, eran de origen chino, y copiaban en el examen utilizando la tecnologia facilitada por una organizacion. Destaca que, ademas de parte del fraude que supone copiar en un examen muchos de estos ciudadanos desconocian el idioma, no hablan ni entienden el español lo que supone un grave riesgo para la seguridad vial por desconocer las señales y letreros que avisan en carretera de muchas incidencias.
"
---
# mt5-small-spanish-summarization
## Model description
This is a mt5-small model finetuned for generating headlines from the body of the news in Spanish.
## Training data
The model was trained with 58425 news extracted from the La Razón (31477) and Público (26948) newspapers. These news belong to the following categories: "España", "Cultura", "Economía", "Igualdad" and "Política".
## Training procedure
It was trained with Google Colab's GPU Tesla P100-PCIE-16GB for 2 epochs.
### Hyperparameters
{evaluation_strategy = "epoch",
learning_rate = 2e-4,
per_device_train_batch_size = 6,
per_device_eval_batch_size = 6,
weight_decay = 0.01,
save_total_limi t= 3,
num_train_epochs = 2,
predict_with_generate = True,
fp16 = False}
## Eval results
| metric | score |
| --- | ----- |
| rouge1 | 44.03 |
| rouge2 | 28.2900 |
| rougeL | 40.54 |
| rougeLsum | 40.5587 |
### BibTeX entry and citation info
```bibtex
@inproceedings{ mt5lrpjosmunpen,
year={2020},
author = {José Manuel Muñiz Peña},
}
``` |
jpcorb20/pegasus-large-reddit_tifu-samsum-256 | 2021-03-20T15:14:53.000Z | [
"pytorch",
"pegasus",
"seq2seq",
"en",
"dataset:samsum",
"transformers",
"google/pegasus-reddit_tifu",
"summarization",
"samsum",
"text2text-generation"
]
| summarization | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"spiece.model",
"tokenizer_config.json"
]
| jpcorb20 | 44 | transformers | ---
language:
- en
thumbnail:
tags:
- pytorch
- google/pegasus-reddit_tifu
- summarization
- samsum
license:
datasets:
- samsum
metrics:
- rouge
---
# Samsum Pegasus (Reddit/TIFU) for conversational summaries
## Model description
Pegasus (Reddit/TIFU) for conversational summaries trained on the samsum dataset!
## Training data
The data is the [samsum](https://huggingface.co/datasets/samsum) dataset for conversional summaries.
The initial weigths were from the [google/pegasus-reddit_tifu](https://huggingface.co/google/pegasus-reddit_tifu). The hypothesis being that it would help the convergence on the samsum dataset to have weights trained on a larger summarization dataset first like the Reddit TIFU using casual language.
## Training procedure
Used the _example/seq2seq/run_summarization.py_ script from the transformers source _4.5.0dev0_.
n_epochs: 3,\
batch_size: 8, \
max_source_length: 256,\
max_target_length: 128
## Eval results
eval_gen_len: 35.9939,\
eval_loss: 1.4284523725509644,\
eval_rouge1: 46.5613,\
eval_rouge2: 23.6137,\
eval_rougeL: 37.2397,\
eval_rougeLsum: 42.7126,\
eval_samples_per_second: 4.302
## Example
from transformers import PegasusForConditionalGeneration, PegasusTokenizer
model_name = "jpcorb20/pegasus-large-reddit_tifu-samsum-256"
tokenizer = PegasusTokenizer.from_pretrained(model_name)
model = PegasusForConditionalGeneration.from_pretrained(model_name)
src_text = """Carter: Hey Alexis, I just wanted to let you know that I had a really nice time with you tonight.\r\nAlexis: Thanks Carter. Yeah, I really enjoyed myself as well.\r\nCarter: If you are up for it, I would really like to see you again soon.\r\nAlexis: Thanks Carter, I'm flattered. But I have a really busy week coming up.\r\nCarter: Yeah, no worries. I totally understand. But if you ever want to go grab dinner again, just let me know.\r\nAlexis: Yeah of course. Thanks again for tonight. Carter: Sure. Have a great night.\r\n"""
token_params = dict(max_length=256, truncation=True, padding='longest', return_tensors="pt")
batch = tokenizer(src_text, **token_params)
translated = model.generate(**batch)
decode_params = dict(num_beams=5, min_length=16, max_length=128, length_penalty=2)
tgt_text = tokenizer.batch_decode(translated, skip_special_tokens=True, **decode_params)
print(tgt_text) |
jpcorb20/pegasus-large-reddit_tifu-samsum-512 | 2021-03-26T12:59:56.000Z | [
"pytorch",
"pegasus",
"seq2seq",
"en",
"dataset:samsum",
"transformers",
"google/pegasus-reddit_tifu",
"summarization",
"samsum",
"text2text-generation"
]
| summarization | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"spiece.model",
"tokenizer_config.json"
]
| jpcorb20 | 109 | transformers | ---
language:
- en
thumbnail:
tags:
- pytorch
- google/pegasus-reddit_tifu
- summarization
- samsum
license:
datasets:
- samsum
metrics:
- rouge
---
# Samsum Pegasus (Reddit/TIFU) for conversational summaries
## Model description
Pegasus (Reddit/TIFU) for conversational summaries trained on the samsum dataset!
## Training data
The data is the [samsum](https://huggingface.co/datasets/samsum) dataset for conversional summaries.
The initial weigths were from the [google/pegasus-reddit_tifu](https://huggingface.co/google/pegasus-reddit_tifu). The hypothesis being that it would help the convergence on the samsum dataset to have weights trained on a larger summarization dataset first like the Reddit TIFU using casual language.
## Training procedure
Used the example/seq2seq/run_summarization.py script from the transformers source 4.5.0dev0.
n_epochs: 3,\
batch_size: 4, \
max_source_length: 512,\
max_target_length: 128
## Eval results
eval_gen_len: 35.89,\
eval_loss: 1.3807392120361328,\
eval_rouge1: 47.3372,\
eval_rouge2: 24.4728,\
eval_rougeL: 37.9078,\
eval_rougeLsum: 43.5744,\
eval_samples_per_second: 2.814
## Example
from transformers import PegasusForConditionalGeneration, PegasusTokenizer
model_name = "jpcorb20/pegasus-large-reddit_tifu-samsum-256"
tokenizer = PegasusTokenizer.from_pretrained(model_name)
model = PegasusForConditionalGeneration.from_pretrained(model_name)
src_text = """Carter: Hey Alexis, I just wanted to let you know that I had a really nice time with you tonight.\\r\
Alexis: Thanks Carter. Yeah, I really enjoyed myself as well.\\r\
Carter: If you are up for it, I would really like to see you again soon.\\r\
Alexis: Thanks Carter, I'm flattered. But I have a really busy week coming up.\\r\
Carter: Yeah, no worries. I totally understand. But if you ever want to go grab dinner again, just let me know.\\r\
Alexis: Yeah of course. Thanks again for tonight. Carter: Sure. Have a great night.\\r\
"""
token_params = dict(max_length=512, truncation=True, padding='longest', return_tensors="pt")
batch = tokenizer(src_text, **token_params)
translated = model.generate(**batch)
decode_params = dict(num_beams=5, min_length=16, max_length=128, length_penalty=2)
tgt_text = tokenizer.batch_decode(translated, skip_special_tokens=True, **decode_params)
print(tgt_text) |
jpcorb20/toxic-detector-distilroberta | 2021-05-20T17:25:58.000Z | [
"pytorch",
"jax",
"roberta",
"text-classification",
"transformers"
]
| text-classification | [
".gitattributes",
"README.md",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jpcorb20 | 128 | transformers | # Distilroberta for toxic comment detection
See my GitHub repo [toxic-comment-server](https://github.com/jpcorb20/toxic-comment-server)
The model was trained from [DistilRoberta](https://huggingface.co/distilroberta-base) on [Kaggle Toxic Comments](https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge) with the BCEWithLogits loss for Multi-Label prediction. Thus, please use the sigmoid activation on the logits (not made to use the softmax output, e.g. like the HF widget).
## Evaluation
F1 scores:
toxic: 0.72
severe_toxic: 0.38
obscene: 0.72
threat: 0.52
insult: 0.69
identity_hate: 0.60
Macro-F1: 0.61 |
jpelhaw/longformer-base-plagiarism-detection | 2021-02-10T15:26:04.000Z | [
"pytorch",
"longformer",
"text-classification",
"ISO 639-1 code for your language, or `multilingual`",
"dataset:array of dataset identifiers",
"arxiv:2004.05150",
"transformers",
"array",
"of",
"tags",
"license:any valid license identifier"
]
| text-classification | [
".gitattributes",
"README.md",
"config.json",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jpelhaw | 23 | transformers | ---
language: "ISO 639-1 code for your language, or `multilingual`"
thumbnail: "url to a thumbnail used in social sharing"
tags:
- array
- of
- tags
license: "any valid license identifier"
datasets:
- array of dataset identifiers
metrics:
- array of metric identifiers
widget:
- text: "Plagiarism is the representation of another author's writing, thoughts, ideas, or expressions as one's own work."
---
# T5-large for Word Sense Disambiguation
This is the checkpoint for T5-large after being trained on the Machine-Paraphrased Plagiarism Dataset: [](https://doi.org/10.5281/zenodo.3608000)
Additional information about this model:
* [The longformer-base-4096 model page](https://huggingface.co/allenai/longformer-base-4096)
* [Longformer: The Long-Document Transformer](https://arxiv.org/pdf/2004.05150.pdf)
* [Official implementation by AllenAI](https://github.com/allenai/longformer)
The model can be loaded to perform Plagiarism like so:
```py
from transformers import AutoModelForSequenceClassification, AutoTokenizer
AutoModelForSequenceClassification("jpelhaw/longformer-base-plagiarism-detection")
AutoTokenizer.from_pretrained("jpelhaw/longformer-base-plagiarism-detection")
input = 'Plagiarism is the representation of another author's writing, thoughts, ideas, or expressions as one's own work.'
example = tokenizer.tokenize(input, add_special_tokens=True)
answer = model(**example)
# "plagiarised"
``` |
jpelhaw/t5-word-sense-disambiguation | 2021-02-12T12:25:20.000Z | [
"pytorch",
"t5",
"seq2seq",
"ISO 639-1 code for your language, or `multilingual`",
"dataset:array of dataset identifiers",
"arxiv:1910.10683",
"transformers",
"array",
"of",
"tags",
"license:any valid license identifier",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"spiece.model",
"tokenizer_config.json"
]
| jpelhaw | 89 | transformers | ---
language: "ISO 639-1 code for your language, or `multilingual`"
thumbnail: "url to a thumbnail used in social sharing"
tags:
- array
- of
- tags
license: "any valid license identifier"
datasets:
- array of dataset identifiers
metrics:
- array of metric identifiers
widget:
- text: "question: which description describes the word \" java \" best in the following context? descriptions: [ \" A drink consisting of an infusion of ground coffee beans \" , \" a platform-independent programming lanugage \" , or \" an island in Indonesia to the south of Borneo \" ] context: I like to drink ' java ' in the morning ."
---
# T5-large for Word Sense Disambiguation
This is the checkpoint for T5-large after being trained on the [SemCor 3.0 dataset](http://lcl.uniroma1.it/wsdeval/).
Additional information about this model:
* [The t5-large model page](https://huggingface.co/t5-large)
* [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/pdf/1910.10683.pdf)
* [Official implementation by Google](https://github.com/google-research/text-to-text-transfer-transformer)
The model can be loaded to perform a few-shot classification like so:
```py
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
AutoModelForSeq2SeqLM.from_pretrained("jpelhaw/t5-word-sense-disambiguation")
AutoTokenizer.from_pretrained("jpelhaw/t5-word-sense-disambiguation")
input = 'question: which description describes the word " java " best in the following context? \
descriptions:[ " A drink consisting of an infusion of ground coffee beans " ,
" a platform-independent programming lanugage "
, or " an island in Indonesia to the south of Borneo " ]
context: I like to drink " java " in the morning .'
example = tokenizer.tokenize(input, add_special_tokens=True)
answer = model.generate(input_ids=example['input_ids'],
attention_mask=example['attention_mask'],
max_length=135)
# "a distinguishing trait"
```
|
jpelhaw/xlnet-base-plagiarism-detection | 2021-02-16T09:24:23.000Z | [
"pytorch",
"xlnet",
"text-classification",
"ISO 639-1 code for your language, or `multilingual`",
"dataset:array of dataset identifiers",
"arxiv:1906.08237",
"transformers",
"array",
"of",
"tags",
"license:any valid license identifier"
]
| text-classification | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"spiece.model",
"tokenizer_config.json"
]
| jpelhaw | 16 | transformers | ---
language: "ISO 639-1 code for your language, or `multilingual`"
thumbnail: "url to a thumbnail used in social sharing"
tags:
- array
- of
- tags
license: "any valid license identifier"
datasets:
- array of dataset identifiers
metrics:
- array of metric identifiers
widget:
- text: "Copyright infringement is viewed as an infringement of scholarly uprightness and a penetrate of editorial morals."
---
# XLNet-LMGC-M for Machine-Paraphrased Plagiarism Detection
This is the checkpoint for LMGC based on XLNet-base after being trained on the Machine-Paraphrased Plagiarism Dataset: [](https://doi.org/10.5281/zenodo.3608000)
Additional information about this model:
* [The xlnet-base model page](https://huggingface.co/xlnet-base-cased)
* [XLNet: Generalized Autoregressive Pretraining for Language Understanding](https://arxiv.org/pdf/1906.08237.pdf)
The model can be loaded to perform Plagiarism like so:
```py
from transformers import AutoModelForSequenceClassification, AutoTokenizer
AutoModelForSequenceClassification("jpelhaw/xlnet-base-plagiarism-detection")
AutoTokenizer.from_pretrained("jpelhaw/xlnet-base-plagiarism-detection")
input = 'Copyright infringement is viewed as an infringement of scholarly uprightness and a penetrate of editorial morals.'
example = tokenizer.tokenize(input, add_special_tokens=True)
answer = model(**example)
# "plagiarism"
``` |
jplu/tf-camembert-base | 2020-12-11T21:47:52.000Z | [
"tf",
"camembert",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"sentencepiece.bpe.model",
"tf_model.h5"
]
| jplu | 1,967 | transformers | # Tensorflow CamemBERT
In this repository you will find different versions of the CamemBERT model for Tensorflow.
## CamemBERT
[CamemBERT](https://camembert-model.fr/) is a state-of-the-art language model for French based on the RoBERTa architecture pretrained on the French subcorpus of the newly available multilingual corpus OSCAR.
## Model Weights
| Model | Downloads
| -------------------------------- | ---------------------------------------------------------------------------------------------------------------
| `jplu/tf-camembert-base` | [`config.json`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-camembert-base/config.json) • [`tf_model.h5`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-camembert-base/tf_model.h5)
## Usage
With Transformers >= 2.4 the Tensorflow models of CamemBERT can be loaded like:
```python
from transformers import TFCamembertModel
model = TFCamembertModel.from_pretrained("jplu/tf-camembert-base")
```
## Huggingface model hub
All models are available on the [Huggingface model hub](https://huggingface.co/jplu).
## Acknowledgments
Thanks to all the Huggingface team for the support and their amazing library!
|
jplu/tf-flaubert-base-cased | 2020-04-24T16:01:44.000Z | [
"tf",
"flaubert",
"lm-head",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"config.json",
"merges.txt",
"tf_model.h5",
"vocab.json"
]
| jplu | 172 | transformers | |
jplu/tf-flaubert-base-uncased | 2020-04-24T16:01:46.000Z | [
"tf",
"flaubert",
"lm-head",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"config.json",
"merges.txt",
"tf_model.h5",
"vocab.json"
]
| jplu | 31 | transformers | |
jplu/tf-flaubert-large-cased | 2020-04-24T16:01:48.000Z | [
"tf",
"flaubert",
"lm-head",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"config.json",
"merges.txt",
"tf_model.h5",
"vocab.json"
]
| jplu | 31 | transformers | |
jplu/tf-flaubert-small-cased | 2020-05-22T19:30:16.000Z | [
"tf",
"flaubert",
"lm-head",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"config.json",
"merges.txt",
"tf_model.h5",
"vocab.json"
]
| jplu | 103 | transformers | |
jplu/tf-xlm-r-ner-40-lang | 2020-12-11T21:47:56.000Z | [
"tf",
"xlm-roberta",
"token-classification",
"multilingual",
"arxiv:1911.02116",
"transformers"
]
| token-classification | [
".gitattributes",
"README.md",
"config.json",
"sentencepiece.bpe.model",
"special_tokens_map.json",
"tf_model.h5",
"tokenizer_config.json"
]
| jplu | 12,300 | transformers | ---
language: multilingual
---
# XLM-R + NER
This model is a fine-tuned [XLM-Roberta-base](https://arxiv.org/abs/1911.02116) over the 40 languages proposed in [XTREME](https://github.com/google-research/xtreme) from [Wikiann](https://aclweb.org/anthology/P17-1178). This is still an on-going work and the results will be updated everytime an improvement is reached.
The covered labels are:
```
LOC
ORG
PER
O
```
## Metrics on evaluation set:
### Average over the 40 languages
Number of documents: 262300
```
precision recall f1-score support
ORG 0.81 0.81 0.81 102452
PER 0.90 0.91 0.91 108978
LOC 0.86 0.89 0.87 121868
micro avg 0.86 0.87 0.87 333298
macro avg 0.86 0.87 0.87 333298
```
### Afrikaans
Number of documents: 1000
```
precision recall f1-score support
ORG 0.89 0.88 0.88 582
PER 0.89 0.97 0.93 369
LOC 0.84 0.90 0.86 518
micro avg 0.87 0.91 0.89 1469
macro avg 0.87 0.91 0.89 1469
```
### Arabic
Number of documents: 10000
```
precision recall f1-score support
ORG 0.83 0.84 0.84 3507
PER 0.90 0.91 0.91 3643
LOC 0.88 0.89 0.88 3604
micro avg 0.87 0.88 0.88 10754
macro avg 0.87 0.88 0.88 10754
```
### Basque
Number of documents: 10000
```
precision recall f1-score support
LOC 0.88 0.93 0.91 5228
ORG 0.86 0.81 0.83 3654
PER 0.91 0.91 0.91 4072
micro avg 0.89 0.89 0.89 12954
macro avg 0.89 0.89 0.89 12954
```
### Bengali
Number of documents: 1000
```
precision recall f1-score support
ORG 0.86 0.89 0.87 325
LOC 0.91 0.91 0.91 406
PER 0.96 0.95 0.95 364
micro avg 0.91 0.92 0.91 1095
macro avg 0.91 0.92 0.91 1095
```
### Bulgarian
Number of documents: 1000
```
precision recall f1-score support
ORG 0.86 0.83 0.84 3661
PER 0.92 0.95 0.94 4006
LOC 0.92 0.95 0.94 6449
micro avg 0.91 0.92 0.91 14116
macro avg 0.91 0.92 0.91 14116
```
### Burmese
Number of documents: 100
```
precision recall f1-score support
LOC 0.60 0.86 0.71 37
ORG 0.68 0.63 0.66 30
PER 0.44 0.44 0.44 36
micro avg 0.57 0.65 0.61 103
macro avg 0.57 0.65 0.60 103
```
### Chinese
Number of documents: 10000
```
precision recall f1-score support
ORG 0.70 0.69 0.70 4022
LOC 0.76 0.81 0.78 3830
PER 0.84 0.84 0.84 3706
micro avg 0.76 0.78 0.77 11558
macro avg 0.76 0.78 0.77 11558
```
### Dutch
Number of documents: 10000
```
precision recall f1-score support
ORG 0.87 0.87 0.87 3930
PER 0.95 0.95 0.95 4377
LOC 0.91 0.92 0.91 4813
micro avg 0.91 0.92 0.91 13120
macro avg 0.91 0.92 0.91 13120
```
### English
Number of documents: 10000
```
precision recall f1-score support
LOC 0.83 0.84 0.84 4781
PER 0.89 0.90 0.89 4559
ORG 0.75 0.75 0.75 4633
micro avg 0.82 0.83 0.83 13973
macro avg 0.82 0.83 0.83 13973
```
### Estonian
Number of documents: 10000
```
precision recall f1-score support
LOC 0.89 0.92 0.91 5654
ORG 0.85 0.85 0.85 3878
PER 0.94 0.94 0.94 4026
micro avg 0.90 0.91 0.90 13558
macro avg 0.90 0.91 0.90 13558
```
### Finnish
Number of documents: 10000
```
precision recall f1-score support
ORG 0.84 0.83 0.84 4104
LOC 0.88 0.90 0.89 5307
PER 0.95 0.94 0.94 4519
micro avg 0.89 0.89 0.89 13930
macro avg 0.89 0.89 0.89 13930
```
### French
Number of documents: 10000
```
precision recall f1-score support
LOC 0.90 0.89 0.89 4808
ORG 0.84 0.87 0.85 3876
PER 0.94 0.93 0.94 4249
micro avg 0.89 0.90 0.90 12933
macro avg 0.89 0.90 0.90 12933
```
### Georgian
Number of documents: 10000
```
precision recall f1-score support
PER 0.90 0.91 0.90 3964
ORG 0.83 0.77 0.80 3757
LOC 0.82 0.88 0.85 4894
micro avg 0.84 0.86 0.85 12615
macro avg 0.84 0.86 0.85 12615
```
### German
Number of documents: 10000
```
precision recall f1-score support
LOC 0.85 0.90 0.87 4939
PER 0.94 0.91 0.92 4452
ORG 0.79 0.78 0.79 4247
micro avg 0.86 0.86 0.86 13638
macro avg 0.86 0.86 0.86 13638
```
### Greek
Number of documents: 10000
```
precision recall f1-score support
ORG 0.86 0.85 0.85 3771
LOC 0.88 0.91 0.90 4436
PER 0.91 0.93 0.92 3894
micro avg 0.88 0.90 0.89 12101
macro avg 0.88 0.90 0.89 12101
```
### Hebrew
Number of documents: 10000
```
precision recall f1-score support
PER 0.87 0.88 0.87 4206
ORG 0.76 0.75 0.76 4190
LOC 0.85 0.85 0.85 4538
micro avg 0.83 0.83 0.83 12934
macro avg 0.82 0.83 0.83 12934
```
### Hindi
Number of documents: 1000
```
precision recall f1-score support
ORG 0.78 0.81 0.79 362
LOC 0.83 0.85 0.84 422
PER 0.90 0.95 0.92 427
micro avg 0.84 0.87 0.85 1211
macro avg 0.84 0.87 0.85 1211
```
### Hungarian
Number of documents: 10000
```
precision recall f1-score support
PER 0.95 0.95 0.95 4347
ORG 0.87 0.88 0.87 3988
LOC 0.90 0.92 0.91 5544
micro avg 0.91 0.92 0.91 13879
macro avg 0.91 0.92 0.91 13879
```
### Indonesian
Number of documents: 10000
```
precision recall f1-score support
ORG 0.88 0.89 0.88 3735
LOC 0.93 0.95 0.94 3694
PER 0.93 0.93 0.93 3947
micro avg 0.91 0.92 0.92 11376
macro avg 0.91 0.92 0.92 11376
```
### Italian
Number of documents: 10000
```
precision recall f1-score support
LOC 0.88 0.88 0.88 4592
ORG 0.86 0.86 0.86 4088
PER 0.96 0.96 0.96 4732
micro avg 0.90 0.90 0.90 13412
macro avg 0.90 0.90 0.90 13412
```
### Japanese
Number of documents: 10000
```
precision recall f1-score support
ORG 0.62 0.61 0.62 4184
PER 0.76 0.81 0.78 3812
LOC 0.68 0.74 0.71 4281
micro avg 0.69 0.72 0.70 12277
macro avg 0.69 0.72 0.70 12277
```
### Javanese
Number of documents: 100
```
precision recall f1-score support
ORG 0.79 0.80 0.80 46
PER 0.81 0.96 0.88 26
LOC 0.75 0.75 0.75 40
micro avg 0.78 0.82 0.80 112
macro avg 0.78 0.82 0.80 112
```
### Kazakh
Number of documents: 1000
```
precision recall f1-score support
ORG 0.76 0.61 0.68 307
LOC 0.78 0.90 0.84 461
PER 0.87 0.91 0.89 367
micro avg 0.81 0.83 0.82 1135
macro avg 0.81 0.83 0.81 1135
```
### Korean
Number of documents: 10000
```
precision recall f1-score support
LOC 0.86 0.89 0.88 5097
ORG 0.79 0.74 0.77 4218
PER 0.83 0.86 0.84 4014
micro avg 0.83 0.83 0.83 13329
macro avg 0.83 0.83 0.83 13329
```
### Malay
Number of documents: 1000
```
precision recall f1-score support
ORG 0.87 0.89 0.88 368
PER 0.92 0.91 0.91 366
LOC 0.94 0.95 0.95 354
micro avg 0.91 0.92 0.91 1088
macro avg 0.91 0.92 0.91 1088
```
### Malayalam
Number of documents: 1000
```
precision recall f1-score support
ORG 0.75 0.74 0.75 347
PER 0.84 0.89 0.86 417
LOC 0.74 0.75 0.75 391
micro avg 0.78 0.80 0.79 1155
macro avg 0.78 0.80 0.79 1155
```
### Marathi
Number of documents: 1000
```
precision recall f1-score support
PER 0.89 0.94 0.92 394
LOC 0.82 0.84 0.83 457
ORG 0.84 0.78 0.81 339
micro avg 0.85 0.86 0.85 1190
macro avg 0.85 0.86 0.85 1190
```
### Persian
Number of documents: 10000
```
precision recall f1-score support
PER 0.93 0.92 0.93 3540
LOC 0.93 0.93 0.93 3584
ORG 0.89 0.92 0.90 3370
micro avg 0.92 0.92 0.92 10494
macro avg 0.92 0.92 0.92 10494
```
### Portuguese
Number of documents: 10000
```
precision recall f1-score support
LOC 0.90 0.91 0.91 4819
PER 0.94 0.92 0.93 4184
ORG 0.84 0.88 0.86 3670
micro avg 0.89 0.91 0.90 12673
macro avg 0.90 0.91 0.90 12673
```
### Russian
Number of documents: 10000
```
precision recall f1-score support
PER 0.93 0.96 0.95 3574
LOC 0.87 0.89 0.88 4619
ORG 0.82 0.80 0.81 3858
micro avg 0.87 0.88 0.88 12051
macro avg 0.87 0.88 0.88 12051
```
### Spanish
Number of documents: 10000
```
precision recall f1-score support
PER 0.95 0.93 0.94 3891
ORG 0.86 0.88 0.87 3709
LOC 0.89 0.91 0.90 4553
micro avg 0.90 0.91 0.90 12153
macro avg 0.90 0.91 0.90 12153
```
### Swahili
Number of documents: 1000
```
precision recall f1-score support
ORG 0.82 0.85 0.83 349
PER 0.95 0.92 0.94 403
LOC 0.86 0.89 0.88 450
micro avg 0.88 0.89 0.88 1202
macro avg 0.88 0.89 0.88 1202
```
### Tagalog
Number of documents: 1000
```
precision recall f1-score support
LOC 0.90 0.91 0.90 338
ORG 0.83 0.91 0.87 339
PER 0.96 0.93 0.95 350
micro avg 0.90 0.92 0.91 1027
macro avg 0.90 0.92 0.91 1027
```
### Tamil
Number of documents: 1000
```
precision recall f1-score support
PER 0.90 0.92 0.91 392
ORG 0.77 0.76 0.76 370
LOC 0.78 0.81 0.79 421
micro avg 0.82 0.83 0.82 1183
macro avg 0.82 0.83 0.82 1183
```
### Telugu
Number of documents: 1000
```
precision recall f1-score support
ORG 0.67 0.55 0.61 347
LOC 0.78 0.87 0.82 453
PER 0.73 0.86 0.79 393
micro avg 0.74 0.77 0.76 1193
macro avg 0.73 0.77 0.75 1193
```
### Thai
Number of documents: 10000
```
precision recall f1-score support
LOC 0.63 0.76 0.69 3928
PER 0.78 0.83 0.80 6537
ORG 0.59 0.59 0.59 4257
micro avg 0.68 0.74 0.71 14722
macro avg 0.68 0.74 0.71 14722
```
### Turkish
Number of documents: 10000
```
precision recall f1-score support
PER 0.94 0.94 0.94 4337
ORG 0.88 0.89 0.88 4094
LOC 0.90 0.92 0.91 4929
micro avg 0.90 0.92 0.91 13360
macro avg 0.91 0.92 0.91 13360
```
### Urdu
Number of documents: 1000
```
precision recall f1-score support
LOC 0.90 0.95 0.93 352
PER 0.96 0.96 0.96 333
ORG 0.91 0.90 0.90 326
micro avg 0.92 0.94 0.93 1011
macro avg 0.92 0.94 0.93 1011
```
### Vietnamese
Number of documents: 10000
```
precision recall f1-score support
ORG 0.86 0.87 0.86 3579
LOC 0.88 0.91 0.90 3811
PER 0.92 0.93 0.93 3717
micro avg 0.89 0.90 0.90 11107
macro avg 0.89 0.90 0.90 11107
```
### Yoruba
Number of documents: 100
```
precision recall f1-score support
LOC 0.54 0.72 0.62 36
ORG 0.58 0.31 0.41 35
PER 0.77 1.00 0.87 36
micro avg 0.64 0.68 0.66 107
macro avg 0.63 0.68 0.63 107
```
## Reproduce the results
Download and prepare the dataset from the [XTREME repo](https://github.com/google-research/xtreme#download-the-data). Next, from the root of the transformers repo run:
```
cd examples/ner
python run_tf_ner.py \
--data_dir . \
--labels ./labels.txt \
--model_name_or_path jplu/tf-xlm-roberta-base \
--output_dir model \
--max-seq-length 128 \
--num_train_epochs 2 \
--per_gpu_train_batch_size 16 \
--per_gpu_eval_batch_size 32 \
--do_train \
--do_eval \
--logging_dir logs \
--mode token-classification \
--evaluate_during_training \
--optimizer_name adamw
```
## Usage with pipelines
```python
from transformers import pipeline
nlp_ner = pipeline(
"ner",
model="jplu/tf-xlm-r-ner-40-lang",
tokenizer=(
'jplu/tf-xlm-r-ner-40-lang',
{"use_fast": True}),
framework="tf"
)
text_fr = "Barack Obama est né à Hawaï."
text_en = "Barack Obama was born in Hawaii."
text_es = "Barack Obama nació en Hawai."
text_zh = "巴拉克·奧巴馬（Barack Obama）出生於夏威夷。"
text_ar = "ولد باراك أوباما في هاواي."
nlp_ner(text_fr)
#Output: [{'word': '▁Barack', 'score': 0.9894659519195557, 'entity': 'PER'}, {'word': '▁Obama', 'score': 0.9888848662376404, 'entity': 'PER'}, {'word': '▁Hawa', 'score': 0.998701810836792, 'entity': 'LOC'}, {'word': 'ï', 'score': 0.9987035989761353, 'entity': 'LOC'}]
nlp_ner(text_en)
#Output: [{'word': '▁Barack', 'score': 0.9929141998291016, 'entity': 'PER'}, {'word': '▁Obama', 'score': 0.9930834174156189, 'entity': 'PER'}, {'word': '▁Hawaii', 'score': 0.9986202120780945, 'entity': 'LOC'}]
nlp_ner(test_es)
#Output: [{'word': '▁Barack', 'score': 0.9944776296615601, 'entity': 'PER'}, {'word': '▁Obama', 'score': 0.9949177503585815, 'entity': 'PER'}, {'word': '▁Hawa', 'score': 0.9987911581993103, 'entity': 'LOC'}, {'word': 'i', 'score': 0.9984861612319946, 'entity': 'LOC'}]
nlp_ner(test_zh)
#Output: [{'word': '夏威夷', 'score': 0.9988449215888977, 'entity': 'LOC'}]
nlp_ner(test_ar)
#Output: [{'word': '▁با', 'score': 0.9903655648231506, 'entity': 'PER'}, {'word': 'راك', 'score': 0.9850614666938782, 'entity': 'PER'}, {'word': '▁أوباما', 'score': 0.9850308299064636, 'entity': 'PER'}, {'word': '▁ها', 'score': 0.9477543234825134, 'entity': 'LOC'}, {'word': 'وا', 'score': 0.9428229928016663, 'entity': 'LOC'}, {'word': 'ي', 'score': 0.9319471716880798, 'entity': 'LOC'}]
```
|
jplu/tf-xlm-roberta-base | 2020-12-11T21:48:00.000Z | [
"tf",
"xlm-roberta",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"sentencepiece.bpe.model",
"tf_model.h5"
]
| jplu | 12,015 | transformers | # Tensorflow XLM-RoBERTa
In this repository you will find different versions of the XLM-RoBERTa model for Tensorflow.
## XLM-RoBERTa
[XLM-RoBERTa](https://ai.facebook.com/blog/-xlm-r-state-of-the-art-cross-lingual-understanding-through-self-supervision/) is a scaled cross lingual sentence encoder. It is trained on 2.5T of data across 100 languages data filtered from Common Crawl. XLM-R achieves state-of-the-arts results on multiple cross lingual benchmarks.
## Model Weights
| Model | Downloads
| -------------------------------- | ---------------------------------------------------------------------------------------------------------------
| `jplu/tf-xlm-roberta-base` | [`config.json`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-base/config.json) • [`tf_model.h5`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-base/tf_model.h5)
| `jplu/tf-xlm-roberta-large` | [`config.json`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-large/config.json) • [`tf_model.h5`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-large/tf_model.h5)
## Usage
With Transformers >= 2.4 the Tensorflow models of XLM-RoBERTa can be loaded like:
```python
from transformers import TFXLMRobertaModel
model = TFXLMRobertaModel.from_pretrained("jplu/tf-xlm-roberta-base")
```
Or
```
model = TFXLMRobertaModel.from_pretrained("jplu/tf-xlm-roberta-large")
```
## Huggingface model hub
All models are available on the [Huggingface model hub](https://huggingface.co/jplu).
## Acknowledgments
Thanks to all the Huggingface team for the support and their amazing library!
|
jplu/tf-xlm-roberta-large | 2020-12-11T21:48:04.000Z | [
"tf",
"xlm-roberta",
"masked-lm",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"sentencepiece.bpe.model",
"tf_model.h5"
]
| jplu | 3,592 | transformers | # Tensorflow XLM-RoBERTa
In this repository you will find different versions of the XLM-RoBERTa model for Tensorflow.
## XLM-RoBERTa
[XLM-RoBERTa](https://ai.facebook.com/blog/-xlm-r-state-of-the-art-cross-lingual-understanding-through-self-supervision/) is a scaled cross lingual sentence encoder. It is trained on 2.5T of data across 100 languages data filtered from Common Crawl. XLM-R achieves state-of-the-arts results on multiple cross lingual benchmarks.
## Model Weights
| Model | Downloads
| -------------------------------- | ---------------------------------------------------------------------------------------------------------------
| `jplu/tf-xlm-roberta-base` | [`config.json`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-base/config.json) • [`tf_model.h5`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-base/tf_model.h5)
| `jplu/tf-xlm-roberta-large` | [`config.json`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-large/config.json) • [`tf_model.h5`](https://s3.amazonaws.com/models.huggingface.co/bert/jplu/tf-xlm-roberta-large/tf_model.h5)
## Usage
With Transformers >= 2.4 the Tensorflow models of XLM-RoBERTa can be loaded like:
```python
from transformers import TFXLMRobertaModel
model = TFXLMRobertaModel.from_pretrained("jplu/tf-xlm-roberta-base")
```
Or
```
model = TFXLMRobertaModel.from_pretrained("jplu/tf-xlm-roberta-large")
```
## Huggingface model hub
All models are available on the [Huggingface model hub](https://huggingface.co/jplu).
## Acknowledgments
Thanks to all the Huggingface team for the support and their amazing library!
|
jplu/tiny-tf-bert-random | 2021-05-19T20:52:55.000Z | [
"tf",
"bert",
"transformers"
]
| [
".gitattributes",
"config.json",
"tf_model.h5"
]
| jplu | 3,854 | transformers | ||
jppaolim/homerGPT2 | 2021-05-23T06:05:48.000Z | [
"pytorch",
"jax",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"README.md",
"added_tokens.json",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jppaolim | 7 | transformers | First model for storytelling
|
jppaolim/homerGPT2L | 2021-05-23T06:09:00.000Z | [
"pytorch",
"jax",
"gpt2",
"lm-head",
"causal-lm",
"transformers",
"text-generation"
]
| text-generation | [
".gitattributes",
"README.md",
"added_tokens.json",
"config.json",
"flax_model.msgpack",
"merges.txt",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.json"
]
| jppaolim | 6 | transformers | Second model for storytelling
|
jrbarnard/t5-generate-answer | 2021-06-18T08:51:34.000Z | [
"pytorch",
"t5",
"seq2seq",
"transformers",
"text2text-generation"
]
| text2text-generation | [
".gitattributes",
"config.json",
"eval_results.txt",
"model_args.json",
"pytorch_model.bin",
"special_tokens_map.json",
"spiece.model",
"tokenizer_config.json",
"training_args.bin",
"training_progress_scores.csv"
]
| jrbarnard | 0 | transformers | |
jsm33/d4-adaptation | 2021-05-30T18:31:03.000Z | [
"pytorch",
"roberta",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"bpe.codes",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.txt"
]
| jsm33 | 11 | transformers | ||
jsm33/d4-primary | 2021-05-30T18:19:39.000Z | [
"pytorch",
"roberta",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"bpe.codes",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.txt"
]
| jsm33 | 15 | transformers | ||
jsm33/irony-classifier | 2021-05-20T17:26:40.000Z | [
"pytorch",
"roberta",
"transformers"
]
| [
".gitattributes",
"added_tokens.json",
"bpe.codes",
"config.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"vocab.txt"
]
| jsm33 | 19 | transformers | ||
jtfields/hf-tutorial | 2021-05-18T19:13:01.000Z | []
| [
".gitattributes"
]
| jtfields | 0 | |||
jtubert/test | 2021-04-20T16:13:33.000Z | []
| [
".gitattributes",
"README.md"
]
| jtubert | 0 | |||
juliamendelsohn/framing_immigration_specific | 2021-05-20T17:27:05.000Z | [
"pytorch",
"roberta",
"transformers"
]
| [
".gitattributes",
"config.json",
"eval_results.txt",
"merges.txt",
"model_args.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"training_args.bin",
"training_progress_scores.csv",
"vocab.json"
]
| juliamendelsohn | 7 | transformers | ||
juliamendelsohn/framing_issue_generic | 2021-05-20T17:27:30.000Z | [
"pytorch",
"roberta",
"transformers"
]
| [
".gitattributes",
"config.json",
"eval_results.txt",
"merges.txt",
"model_args.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"training_args.bin",
"training_progress_scores.csv",
"vocab.json"
]
| juliamendelsohn | 6 | transformers | ||
juliamendelsohn/framing_narrative | 2021-05-20T17:28:06.000Z | [
"pytorch",
"roberta",
"transformers"
]
| [
".gitattributes",
"config.json",
"eval_results.txt",
"merges.txt",
"model_args.json",
"pytorch_model.bin",
"special_tokens_map.json",
"tokenizer_config.json",
"training_args.bin",
"training_progress_scores.csv",
"vocab.json"
]
| juliamendelsohn | 9 | transformers | ||
juliawabant/camembert-long | 2021-02-16T10:44:27.000Z | [
"pytorch",
"camembert",
"masked-lm",
"arxiv:2004.05150",
"transformers",
"fill-mask"
]
| fill-mask | [
".gitattributes",
"README.md",
"config.json",
"pytorch_model.bin",
"sentencepiece.bpe.model",
"special_tokens_map.json",
"tokenizer_config.json"
]
| juliawabant | 21 | transformers | This model is an "extension" of [CamemBERT-base model] (https://camembert-model.fr/) to support up to 4096 tokens.
The architecture was previously extended using [AllenAI's code](https://github.com/allenai/longformer/blob/master/scripts/convert_model_to_long.ipynb) for the procedure descriped in the [Longformer paper](https://arxiv.org/abs/2004.05150). [Their code](https://github.com/allenai/longformer) is Apache-2.0 licensed.
Longformer is an "extension" of the RoBERTa model and CamemBERT use the same architecture as RoBERTa (they don't have same tokenizers though and CamemBERT use Whole Word Masking).
Hence, this "long CamemBERT" can be pretained or directly fine-tuned using AllenAI's code without substantial modifications after being extended.
I created [a notebook] (https://colab.research.google.com/drive/1B7Gj32yqM2NETdnOlqt997RArYJwV24D#scrollTo=5VKUZnBb4vvD) in order to see the model in action with a HuggingFace pipeline (fillmask).
The model can be pretrained using FP16, or using TPUs with TensorFlow or PyTorch-XLA. |
julien-c/DPRNNTasNet-ks16_WHAM_sepclean | 2021-01-12T13:40:25.000Z | [
"pytorch",
"dataset:wham",
"dataset:sep_clean",
"arxiv:2005.04132",
"asteroid",
"audio",
"audio-source-separation",
"license:cc-by-sa-3.0"
]
| audio-source-separation | [
".gitattributes",
"README.md",
"pytorch_model.bin"
]
| julien-c | 0 | asteroid | ---
tags:
- asteroid
- audio
- audio-source-separation
datasets:
- wham
- sep_clean
license: cc-by-sa-3.0
---
## Asteroid model `mpariente/DPRNNTasNet(ks=16)_WHAM!_sepclean`
♻️ Imported from https://zenodo.org/record/3903795#.X8pMBRNKjUI
This model was trained by Manuel Pariente using the wham/DPRNN recipe in [Asteroid](https://github.com/asteroid-team/asteroid). It was trained on the sep_clean task of the WHAM! dataset.
### Demo: How to use in Asteroid
```python
# coming soon
```
### Training config
- data:
- mode: min
- nondefault_nsrc: None
- sample_rate: 8000
- segment: 2.0
- task: sep_clean
- train_dir: data/wav8k/min/tr
- valid_dir: data/wav8k/min/cv
- filterbank:
- kernel_size: 16
- n_filters: 64
- stride: 8
- main_args:
- exp_dir: exp/train_dprnn_ks16/
- help: None
- masknet:
- bidirectional: True
- bn_chan: 128
- chunk_size: 100
- dropout: 0
- hid_size: 128
- hop_size: 50
- in_chan: 64
- mask_act: sigmoid
- n_repeats: 6
- n_src: 2
- out_chan: 64
- optim:
- lr: 0.001
- optimizer: adam
- weight_decay: 1e-05
- positional arguments:
- training:
- batch_size: 6
- early_stop: True
- epochs: 200
- gradient_clipping: 5
- half_lr: True
- num_workers: 6
#### Results
- `si_sdr`: 18.227683982688003
- `si_sdr_imp`: 18.22883576588251
- `sdr`: 18.617789605060587
- `sdr_imp`: 18.466745426438173
- `sir`: 29.22773720052717
- `sir_imp`: 29.07669302190474
- `sar`: 19.116352171914485
- `sar_imp`: -130.06009796503054
- `stoi`: 0.9722025377865715
- `stoi_imp`: 0.23415680987800583
### Citing Asteroid
```BibTex
@inproceedings{Pariente2020Asteroid,
title={Asteroid: the {PyTorch}-based audio source separation toolkit for researchers},
author={Manuel Pariente and Samuele Cornell and Joris Cosentino and Sunit Sivasankaran and
Efthymios Tzinis and Jens Heitkaemper and Michel Olvera and Fabian-Robert Stöter and
Mathieu Hu and Juan M. Martín-Doñas and David Ditter and Ariel Frank and Antoine Deleforge
and Emmanuel Vincent},
year={2020},
booktitle={Proc. Interspeech},
}
```
Or on arXiv:
```bibtex
@misc{pariente2020asteroid,
title={Asteroid: the PyTorch-based audio source separation toolkit for researchers},
author={Manuel Pariente and Samuele Cornell and Joris Cosentino and Sunit Sivasankaran and Efthymios Tzinis and Jens Heitkaemper and Michel Olvera and Fabian-Robert Stöter and Mathieu Hu and Juan M. Martín-Doñas and David Ditter and Ariel Frank and Antoine Deleforge and Emmanuel Vincent},
year={2020},
eprint={2005.04132},
archivePrefix={arXiv},
primaryClass={eess.AS}
}
``` |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.