SentenceTransformer based on avsolatorio/GIST-Embedding-v0

This is a sentence-transformers model finetuned from avsolatorio/GIST-Embedding-v0. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: avsolatorio/GIST-Embedding-v0
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 tokens
Similarity Function: Cosine Similarity

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("wjunwei/ecommerce_text_embedding")
# Run inference
sentences = [
    'magic the gathering modern horizons  bundle gift edition  deluxe bundle with  collector booster  play boosters fullart lands  exclusive accessories give the gift of magicthe perfect present for any magic fan this gift bundle is full of exclusive accessories and the coolest cards from modern horizons  including a collector booster packed with rare cards shiny foils and altart altframe cards moderns never been more marvelousintroducing a heaping helping of exciting cards for modern one of magics most celebrated formats plus the return of competitive favorites theres something for every fan to love in modern horizons  a collector booster full of treasuresboth powerful and flashy the included collector booster is packed with only the coolest cards and is the only place fans may find foiletched cards textured foil cards or serialized cards printed with a unique number boosters for playing  building decksplay boosters are the best packs for playing magic with friends and are fun to open with a possibility of multiple rares and at least  shining foil card in every pack fullart lands  special alternateart cardevery mh bundle gift edition also comes with  traditional foil promo card featuring bundleexclusive alternate art and  of the  included land cards feature stunning fullart  foil  nonfoil exclusive accessorieseach gift bundle also comes with exclusive accessories including a special die to track your life total as you play and a sturdy box to store everything in bundle gift edition contents collector booster  play boosters  traditional foil altart card  land cards  foil  nonfoil  spindown life counter  card storage box and  reference cards',
    'backdrop stand xft photo video studio adjustable backdrop stand for parties wedding photography advertising display note balloons colorful decorations are not included meaningful family gathering cheerful and memorable birthday party romantic and warm wedding professional photography attractive advertising display and other uses just get  ft backdrop stands adjustable stable backdrop stand loading capacity up to lbkg good flexibility with width ftft  height  ft ft heavy duty spring clamp and backdrop elastic string clip to holds curtains canvas muslin projector screen or seamless paper prevent background slippage photography weight sandbag to stabilize the backdrop support system note this bag comes empty its capacity is about lbub the kit includes  pcs crossbar parts   pcs tripod bracket   pcs spring clamp   pcs backdrop elastic string clip   pcs sand bag   pc carry bag',
    'burts bees baby baby girls mittens noscratch mitts  organic cotton set of  ',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Training Details

Training Dataset

Unnamed Dataset

Size: 5,864 training samples
Columns: sentence_0, sentence_1, and label
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1 label
type string string int
details
min: 5 tokens
mean: 144.3 tokens
max: 512 tokens

min: 5 tokens
mean: 143.08 tokens
max: 512 tokens

0: ~87.50%
1: ~12.50%

	sentence_0	sentence_1	label
type	string	string	int
details	min: 5 tokens mean: 144.3 tokens max: 512 tokens	min: 5 tokens mean: 143.08 tokens max: 512 tokens	0: ~87.50% 1: ~12.50%

Samples:

sentence_0	sentence_1	label
philips norelco shaver rechargeable electric shaver with popup trimmer s comfortcut blades get a clean shave thats comfortable on your skin rounded blade caps shield selfsharpening blades to gently cut hair just above skin level and help the shaver glide smoothly over your skin experience a convenient clean shave with heads that flex and float in directions the head adjusts to the curves of your face ensuring smooth contact with your skin without a lot of pressure d flex heads follow your faces contours for a clean shave popup trimmer for mustache and sideburns finish your look with the builtin trimmer ideal for maintaining your mustache and trimming your sideburns onetouch open for easy cleaning experience a convenient clean shave with heads that flex and float in directions the head adjusts to the curves of your face ensuring smooth contact with your skin without a lot of pressure minutes of cordless shaving from an hour charge thats about shaves or plug it in for instant continuous power	philips norelco shaver rechargeable electric shaver with popup trimmer s comfortcut blades get a clean shave thats comfortable on your skin rounded blade caps shield selfsharpening blades to gently cut hair just above skin level and help the shaver glide smoothly over your skin experience a convenient clean shave with heads that flex and float in directions the head adjusts to the curves of your face ensuring smooth contact with your skin without a lot of pressure d flex heads follow your faces contours for a clean shave popup trimmer for mustache and sideburns finish your look with the builtin trimmer ideal for maintaining your mustache and trimming your sideburns onetouch open for easy cleaning experience a convenient clean shave with heads that flex and float in directions the head adjusts to the curves of your face ensuring smooth contact with your skin without a lot of pressure minutes of cordless shaving from an hour charge thats about shaves or plug it in for instant continuous power	`1`
`speedo girls swimsuit one piece thin straps`	`lace tulle flower girl long dress for wedding oneck princess dresses long sleeve pageant party gown`	`0`
`pyrex blue cup rectangular plastic cover pc pack original genuine pyrex made in the usa genuine pyrex replacement lid nonporous surface does not absorb food odors flavors or stains refrigerator microwave and toprack dishwasher safe will not fit anchor hocking products included pyrex pc blue cup rectangle plastic lids lids only containers not included`	`kpywzer vintage leather sling bag bags for men women backpack shoulder messenger crossbody outdoor travel hiking camping tactical chest pack daypack brown`	`0`

Loss: ContrastiveTensionLoss

Training Hyperparameters

Non-Default Hyperparameters

num_train_epochs: 5
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 8
per_device_eval_batch_size: 8
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
learning_rate: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
num_train_epochs: 5
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: False
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 0
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
past_index: -1
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: False
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
dispatch_batches: None
split_batches: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss
0.6821	500	7.4403
1.3643	1000	4.8536
2.0464	1500	3.8646
2.7285	2000	3.8877
3.4106	2500	3.8789
4.0928	3000	3.8052
4.7749	3500	3.8385

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.41.2
PyTorch: 2.3.0+cu121
Accelerate: 0.31.0
Datasets: 2.20.0
Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

ContrastiveTensionLoss

@inproceedings{carlsson2021semantic,
    title={Semantic Re-tuning with Contrastive Tension},
    author={Fredrik Carlsson and Amaru Cuba Gyllensten and Evangelia Gogoulou and Erik Ylip{"a}{"a} Hellqvist and Magnus Sahlgren},
    booktitle={International Conference on Learning Representations},
    year={2021},
    url={https://openreview.net/forum?id=Ov_sMNau-PF}
}

wjunwei
/

ecommerce_text_embedding