Datasets:

---

DanielPFlorian

/

Transformers-Github-Issues

like 0

[]

[]

[ "Indeed, thanks @ikuyamada!" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "We have new lm_fintetuning scripts thanks to @Rocketknight1 PR #392.\r\nMaybe you can try these ones.\r\nThey are in the `./examples/lm_finetuning/` folder.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Maybe you should ask in the AllenNLP repo as well?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "I realised that the way I was loading model was wrong.\r\n\r\nI used this code\r\n\r\n ```\r\n# Save a trained model \r\n model_to_save = model.module if hasattr(model, 'module') else model # Only save the model it-self \r\n output_model_file = os.path.join(args.output_dir, \"pytorch_model.bin\") \r\n torch.save(model_to_save.state_dict(), output_model_file) \r\n \r\n # Load a trained model that you have fine-tuned \r\n model_state_dict = torch.load(output_model_file) \r\n model = BertForQuestionAnswering.from_pretrained(args.bert_model, state_dict=model_state_dict) \r\n model.to(device)\r\n```" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "I used this code and it worked.\r\n\r\n```\r\n output_model_file = /path/to/pytorch_mode.bin/\r\n model_state_dict = torch.load(output_model_file) \r\n model = BertModel.from_pretrained(bert_model, state_dict=model_state_dict)\r\n```", "Hi @KavyaGujjala \r\n\r\nI was fine-tuning the 'Bert base uncased' model as per the [Google Bert Repository](https://github.com/google-research/bert) on my domain specific data. I used the existing WordPiece vocab and ran pre-training for 50000 steps on the in-domain text to learn the compositionality. I realized that the updated embeddings were improved by manual evaluation. But I really want to add my domain words to the vocab, as they carry importance for my downstream tasks.\r\n\r\nCan you please tell how did you handle the domain vocab while fine-tuning? Did you add your domain words to the vocab file? If yes, did you append the words to the vocab file or just replaced the [unusedXX] words from the file? \r\n \r\nDid you see any improvements after fine-tuning your models? How did you evaluate the embeddings quality for your domain, if not evaluated manually? \r\n", "Hi @harmanpreet93 \r\n\r\nI haven't used fine tuning code from actual google bert repo but used the pretraining code. I used finetuning code from pytorch repo but the embeddings were getting changed everytime I loaded the model. So I am just pretraining for domain specific data.\r\n\r\nMy first question, you are doing fine tuning or pretraining?\r\n\r\nI have pretrained using bert base uncased model for 10000 steps. After that I got the sentence representation using [CLS] token from the final hidden layer. Compared the cosine similarity between sentences. I didnt get good results though. Dont know what I am missing. May be as you have mentioned I need to add words to vocab file.\r\n\r\n", "Hi @KavyaGujjala \r\n \r\n> My first question, you are doing fine tuning or pretraining? \r\n\r\nYes, I too used pre-training code from bert repo. After pretraining on domain data for 50k steps, the embeddings got updated. \r\n\r\n> Compared the cosine similarity between sentences. I didnt get good results though. \r\n\r\n**Q:** How did you evaluate the quality of sentence embeddings? By not getting good results, do you mean that the similar sentences weren't scored at the top or the cosine scores weren't good? \r\n \r\nIn my case the similarity scores decreased, but the relevant sentences started showing up in the topk. \r\n \r\n> May be as you have mentioned I need to add words to vocab file. \r\n \r\nI'm referring to [this](https://github.com/google-research/bert/issues/9) issue from bert repo. Its suggests the following approaches: \r\n> But if you want to add more vocab you can either:\r\n(a) Just replace the \"[unusedX]\" tokens with your vocabulary. Since these were not used they are effectively randomly initialized.\r\n(b) Append it to the end of the vocab, and write a script which generates a new checkpoint that is identical to the pre-trained checkpoint, but but with a bigger vocab where the new embeddings are randomly initialized (for initialized we used tf.truncated_normal_initializer(stddev=0.02)). This will likely require mucking around with some tf.concat() and tf.assign() calls. \r\n \r\n\r\n**Q:** Did you try any of these approaches for the bert repo? I tried approach (b), by adding my domain words to the vocab file and updated the `vocab_size` as per [repo](https://github.com/google-research/bert#pre-training-with-bert). I ran into following error: \r\n`ValueError: Shape of variable bert/embeddings/word_embeddings:0 ((33297, 768)) doesn't match with shape of tensor bert/embeddings/word_embeddings ([30522, 768]) from checkpoint reader.`\r\n\r\n**Q:** How were the sentence embeddings calculated? By averaging the word embeddings or any other approach? ", "Hi @harmanpreet93 \r\n\r\n> Yes, I too used pre-training code from bert repo. After pretraining on domain data for 50k steps, the embeddings got updated.\r\n\r\nWhat is the size of dataset you have used? My dataset has like 1 million sentences domain specific.\r\n\r\n> Q: How did you evaluate the quality of sentence embeddings? By not getting good results, do you mean that the similar sentences weren't scored at the top or the cosine scores weren't good?\r\n\r\nyeah cosine scores aren't good enough. Even dissimilar sentences are getting better similarity scores than the similar ones. Does training for more steps solve this issue?\r\n\r\n> Q: Did you try any of these approaches for the bert repo? I tried approach (b), by adding my domain words to the vocab file and updated the vocab_size as per repo. I ran into following error:\r\n> ValueError: Shape of variable bert/embeddings/word_embeddings:0 ((33297, 768)) doesn't match with shape of tensor bert/embeddings/word_embeddings ([30522, 768]) from checkpoint reader.\r\n\r\nI haven't tried adding domain specific words to vocab but are you using the pretrained model bert_config.json file along with the domain specific trained model checkpoint. If so have you changed vocab size? And did you do it the way jacobdevlin has mentioned like writing script to initialize random weights? I didnt really understand that part, I mean what exactly are we supposed to do to initialize weights.\r\nAlso why didnt you try replacing unusedX tokens?\r\n\r\n> Q: How were the sentence embeddings calculated? By averaging the word embeddings or any other approach?\r\n\r\nI used the [CLS] token embeddings from the hidden layer output ( as they have mentioned in the paper that [CLS] token can be used for sequence level classification ). I initially tried max pooling and mean pooling of word embeddings but got really bad results ( every cosine similarity score was above 0.7 ).\r\nWhich approach did you follow?\r\n", "To add to this, in their paper they mention they get the best results by concatenating the last four layers.\r\n\r\n```python\r\n#! In your model setup\r\n# Indices of layers to concatenate\r\nself.bert_layers = [-1, -2, -3, -4]\r\nself.bert = BertModel.from_pretrained('your-checkpoint.pth')\r\n\r\n#! In your forward method\r\nall_bert_layers, _ = self.bert_layer(bert_ids, attention_mask=bert_mask)\r\nbert_concat = torch.cat(tuple([all_bert_layers[i] for i in self.bert_layers]), dim=-1)\r\n\r\n# If you use a mask:\r\n## Pooling by also setting masked items to zero\r\nbert_mask = torch.FloatTensor(bert_mask).unsqueeze(2)\r\n## Multiply output with mask to only retain non-paddding tokens\r\nbert_pooled = torch.mul(bert_concat, bert_mask)\r\n\r\n# First item ['CLS'] is sentence representation.\r\n# Use bert_concat instead of bert_pooled if you didn't use a mask\r\nfinal_bert = bert_pooled[:, 0, :]\r\n```", "> To add to this, in their paper they mention they get the best results by concatenating the last four layers.\r\n> \r\n> ```python\r\n> #! In your model setup\r\n> # Indices of layers to concatenate\r\n> self.bert_layers = [-1, -2, -3, -4]\r\n> self.bert = BertModel.from_pretrained('your-checkpoint.pth')\r\n> \r\n> #! In your forward method\r\n> all_bert_layers, _ = self.bert_layer(bert_ids, attention_mask=bert_mask)\r\n> bert_concat = torch.cat(tuple([all_bert_layers[i] for i in self.bert_layers]), dim=-1)\r\n> \r\n> # If you use a mask:\r\n> ## Pooling by also setting masked items to zero\r\n> bert_mask = torch.FloatTensor(bert_mask).unsqueeze(2)\r\n> ## Multiply output with mask to only retain non-paddding tokens\r\n> bert_pooled = torch.mul(bert_concat, bert_mask)\r\n> \r\n> # First item ['CLS'] is sentence representation.\r\n> # Use bert_concat instead of bert_pooled if you didn't use a mask\r\n> final_bert = bert_pooled[:, 0, :]\r\n> ```\r\n\r\n@BramVanroy Thanks for the info. I am using original BERT [repo](https://github.com/google-research/bert) pretraining code and used extract_features.py code to get last four layers output for all the tokens in a sequence. By concatenating the last four layers does it mean adding all four layers embeddings of each token?", "@KavyaGujjala If you look at my code, you can see that I mean concatenating across the hidden dim axis. So let's say the output of a single bert layer is batch_size * seq_len * hidden_dim, then concatenating the last four ones ends up with batch_size * seq_len * (hidden_dim*4).", "Hi @harmanpreet93 ,\r\n\r\nYou ran 50k epochs for fine tuning?\r\n\r\nThanks\r\nMahesh\r\n", "@search4mahesh Yes, I ran 50k epochs for fine-tuning!", "@harmanpreet93 that must be lot of compute time, example only says about 3 epochs.\r\n ", "I'm assuming that @harmanpreet93 means 50k steps.", "> @harmanpreet93 that must be a lot of compute time, example only says about 3 epochs.\r\n\r\nYou are right @BramVanroy I meant 50k steps. ", "@harmanpreet93 Have you solve the problem of \"ValueError: Shape of variable bert/embeddings/word_embeddings:0 ((33297, 768)) doesn't match with shape of tensor bert/embeddings/word_embeddings ([30522, 768]) from checkpoint reader.\"?", "Hi @Firmiana1220 \r\n\r\nI was following solutions mentioned [here](https://github.com/google-research/bert/issues/9) from bert repo. It suggests the following two approaches:\r\n\r\n> But if you want to add more vocab you can either:\r\n> (a) Just replace the \"[unusedX]\" tokens with your vocabulary. Since these were not used they are effectively randomly initialized.\r\n> (b) Append it to the end of the vocab, and write a script which generates a new checkpoint that is identical to the pre-trained checkpoint, but with a bigger vocab where the new embeddings are randomly initialized (for initialized we used tf.truncated_normal_initializer(stddev=0.02)). This will likely require mucking around with some tf.concat() and tf.assign() calls.\r\n \r\nI was going for option (b). But I couldn't find any progress. Therefore, I decided to pre-train the model just on my domain data, and not leveraging the already pre-trained models. I'm still looking for a better approach. ", "@KavyaGujjala Thanks for the code sample. Is the masking in it redundant? Unless I'm misunderstanding, you mask out all the unused tokens, but then simply grab the 'CLS' token. The masking would be required if you use all the values, but seems redundant if you're just using the sentence representation.", "> @KavyaGujjala Thanks for the code sample. Is the masking in it redundant? Unless I'm misunderstanding, you mask out all the unused tokens, but then simply grab the 'CLS' token. The masking would be required if you use all the values, but seems redundant if you're just using the sentence representation.\r\n\r\nHi @snard6 , I didnt use that masking code as such, for now I got a finetuned model for my domain specific data and using the cls token for representation which gave comparatively better results.\r\n\r\nI used pytorch bert codes for getting a finetuned model and then extract features code to get the cls token as sentence representation.", "> Hi @harmanpreet93 \r\nI tried the option (a), it works. I replaced the word in vocab.txt with the word in my own domain, other words are [unusedX]. If you don't change the size of vocab.txt, it works. But if your domain words are larger than 30522, maybe you should try the option (b).\r\n", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "I fixed the test failure above in `472857c`. Note that this occurred because I was running torch > 1.0. They fixed the contiguous view issue in https://github.com/pytorch/pytorch/issues/3653. In my view, it would probably make sense to update torch to >1.0 and remove `contiguous()` calls where possible throughout this codebase.", "Hi @CatalinVoss,\r\nThis looks good, thanks!\r\n\r\nWe'll try to keep backward compatibility up to pytorch 0.4.1 for the moment since quite a lot of downstream libraries use this package (like AllenNLP, fair, ParlAI...). So it's good to have the `contiguous()` calls, thanks for that.\r\n\r\nDo you want to add a simple script showcasing finetuning GPT-2 on some dataset (open question, you don't need to)?", "OK, thanks! That all makes sense. Yes, I can add a demo finetuning script, but I probably won't get to it immediately, so probably best to do it in another PR, if that's OK.", "Ok good to merge, thanks @CatalinVoss!" ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hi,\r\n\r\nHow did you get the english embeddings using the bert-base-uncase model.\r\n\r\nI have fine tuned the model on domain specific corpus. I am trying to get embeddings now using the pytorch_model.bin I got after fine tuning.\r\n\r\nAny idea on how to do this?", "Are your models in evaluation mode (`model.eval()`)?\r\nUsually reproductibility issues comes from forgetting to disable the DropOut modules using `model.eval()`.", "> Are your models in evaluation mode (`model.eval()`)?\r\n> Usually reproductibility issues comes from forgetting to disable the DropOut modules using `model.eval()`.\r\n\r\nOh,it's really helpful.My rediculous mistake.Thank you for your answer!!!", "@SupUnicorn Following the above comment, could you please let me know , during the evaluation while reproducing the results using a finetuned model, how to disable the DropOut modules ?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Indeed, I have added backward compatibility to python 2 for GPT-2.\r\nDo you want to submit a PR on this?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Also interested in this! ", "Hi,\r\nIt's quite difficult to use BERT to generate text as BERT is not a causal language model per se.\r\nHere is an example: https://github.com/nyu-dl/bert-gen by @W4ngatang and @kyunghyuncho.", "Bert was not trained for text generation since it's not trained in the classical lm setting. However there are some new approaches that doesn't rely on next word predictions in the classical lm way. Have a look at: [Insertion Transformer](https://arxiv.org/abs/1902.03249) and [Insertion-based Decoding](https://arxiv.org/abs/1902.01370).", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Hi the BERT models are regular PyTorch models, you can just use the usual way we freeze layers in PyTorch. For example you can have a look at the [Transfer Learning tutorial of PyTorch](https://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html#convnet-as-fixed-feature-extractor).\r\n\r\nIn our case freezing the pretrained part of a `BertForSequenceClassification` model would look like this\r\n```python\r\nmodel = BertForSequenceClassification.from_pretrained('bert-base-uncased')\r\n\r\nfor param in model.bert.parameters():\r\n param.requires_grad = False\r\n```\r\nThen only the classification layer should be have `requires_grad=True`.", "Wouldn't this solution freeze all layers, including the classifier layer?", "@nuno-carneiro @thomwolf \r\nI think, this will freeze all the layers including the classifier layer. (Correct me, if I'm wrong) \r\n```\r\nmodel = BertForSequenceClassification.from_pretrained('bert-base-uncased')\r\n\r\nfor param in model.bert.parameters():\r\n param.requires_grad = False\r\n```\r\n\r\n**I think the below code will freeze only the BERT layers (Correct me, if I'm wrong)**\r\n\r\n```\r\nmodel = BertForSequenceClassification.from_pretrained('bert-base-uncased')\r\n\r\nfor param in model.bert.bert.parameters():\r\n param.requires_grad = False\r\n```", "You are right @kalyanks0611 ", "AttributeError: 'BertModel' object has no attribute 'bert'\r\n\r\nI got the above error when doing \r\n\r\nfor param in model.bert.bert.parameters():\r\n param.requires_grad = False\r\n\r\nAny clue? Thanks :)", "@yhl48 Hi, please open a new issue with the problem you're facing. Thank you!", "does anyone know how to do this for the TF 2.0 version too?", "> AttributeError: 'BertModel' object has no attribute 'bert'\r\n> \r\n> I got the above error when doing\r\n> \r\n> for param in model.bert.bert.parameters():\r\n> param.requires_grad = False\r\n> \r\n> Any clue? Thanks :)\r\n\r\nI got the same error. This is my workaround following the one in #1431 :\r\n\r\n\tfor name, param in model.named_parameters():\r\n\t\tif 'classifier' not in name: # classifier layer\r\n\t\t\tparam.requires_grad = False", "You're probably instantiating your model as a `BertModel` whereas @kalyanks0611's example uses `BertForSequenceClassification`.", "> does anyone know how to do this for the TF 2.0 version too?\r\n\r\nThis is the real question", "> does anyone know how to do this for the TF 2.0 version too?\r\n\r\nActually in tf2 framework, pre-trained is saved in `model.bert.weights`, which is a list. So you need to do:\r\n```\r\nfor w in model.bert.weights():\r\n w._trainable= False\r\n```\r\n", "> > does anyone know how to do this for the TF 2.0 version too?\r\n> \r\n> This is the real question\r\n\r\nsee [my post](https://github.com/huggingface/transformers/issues/400#issuecomment-571354368)", "> does anyone know how to do this for the TF 2.0 version too?\r\n\r\nmodel.bert.trainable = False\r\nIs this ok?\r\nBut I found that the training speed has not accelerated", "> > does anyone know how to do this for the TF 2.0 version too?\r\n> \r\n> Actually in tf2 framework, pre-trained is saved in `model.bert.weights`, which is a list. So you need to do:\r\n> \r\n> ```\r\n> for w in model.bert.weights():\r\n> w._trainable= False\r\n> ```\r\n\r\nThanks! I needed a very slight amendment for it to work.\r\n\r\n```\r\nfor w in model.get_layer('tf_distil_bert_model').weights:\r\n w._trainable = False\r\n```\r\n\r\nWorks for me. And training time is halved from 9 hours to 4 hours.", "I am trying to let bert model layer 10 and layer 11 are trainable, like below:\r\n```shell\r\nbert_model = TFBertModel.from_pretrained(\"bert-base-uncased\")\r\nfor w in bert_model.bert.weights:\r\n if w.name.find('layer_._10') == -1 and w.name.find('layer_._11') == -1:\r\n print(w.name)\r\n w._trainable = False\r\n```\r\nBut the result is as below:\r\n```shell\r\nTotal params: 109,483,776\r\nTrainable params: 109,483,776\r\nNon-trainable params: 0\r\n```\r\nHow can i change the specific layers to be trainable?\r\n\r\nThanks", "> \r\n> \r\n> I am trying to let bert model layer 10 and layer 11 are trainable, like below:\r\n> \r\n> ```shell\r\n> bert_model = TFBertModel.from_pretrained(\"bert-base-uncased\")\r\n> for w in bert_model.bert.weights:\r\n> if w.name.find('layer_._10') == -1 and w.name.find('layer_._11') == -1:\r\n> print(w.name)\r\n> w._trainable = False\r\n> ```\r\n> \r\n> But the result is as below:\r\n> \r\n> ```shell\r\n> Total params: 109,483,776\r\n> Trainable params: 109,483,776\r\n> Non-trainable params: 0\r\n> ```\r\n> \r\n> How can i change the specific layers to be trainable?\r\n> \r\n> Thanks\r\n\r\nThe codes below would successfully freezes layers in a model:\r\n```\r\nfrom transformers import TFXLNetModel as transformerModel\r\ntransformer_model_name = 'xlnet-base-cased'\r\ntransformer_model = transformerModel.from_pretrained(transformer_model_name)\r\nfor layer in transformer_model.layers:\r\n layer.trainable = False\r\n```", "> @nuno-carneiro @thomwolf\r\n> I think, this will freeze all the layers including the classifier layer. (Correct me, if I'm wrong)\r\n> \r\n> ```\r\n> model = BertForSequenceClassification.from_pretrained('bert-base-uncased')\r\n> \r\n> for param in model.bert.parameters():\r\n> param.requires_grad = False\r\n> ```\r\n> \r\n> **I think the below code will freeze only the BERT layers (Correct me, if I'm wrong)**\r\n> \r\n> ```\r\n> model = BertForSequenceClassification.from_pretrained('bert-base-uncased')\r\n> \r\n> for param in model.bert.bert.parameters():\r\n> param.requires_grad = False\r\n> ```\r\n\r\nHi, I have \r\n\r\n> @nuno-carneiro @thomwolf\r\n> I think, this will freeze all the layers including the classifier layer. (Correct me, if I'm wrong)\r\n> \r\n> ```\r\n> model = BertForSequenceClassification.from_pretrained('bert-base-uncased')\r\n> \r\n> for param in model.bert.parameters():\r\n> param.requires_grad = False\r\n> ```\r\n> \r\n> **I think the below code will freeze only the BERT layers (Correct me, if I'm wrong)**\r\n> \r\n> ```\r\n> model = BertForSequenceClassification.from_pretrained('bert-base-uncased')\r\n> \r\n> for param in model.bert.bert.parameters():\r\n> param.requires_grad = False\r\n> ```\r\n\r\nI think there are 2 elements in BertForSequenceClassification. One is bert and another one is classifier？or? So I think we dont need model.bert.bert.parameter(). We just need model.bert.parameters()? or?", "`model = BertForSequenceClassification.from_pretrained('bert-base-uncased')`\r\n\r\nWhen you run `model` you will get the following architecture (I skipped many intermediate modules)\r\n```\r\nBertForSequenceClassification(\r\n (bert): BertModel(\r\n (embeddings): BertEmbeddings(\r\n ....\r\n (encoder): BertEncoder(\r\n (layer): ModuleList(\r\n (0): BertLayer(\r\n ...\r\n (output): BertSelfOutput(\r\n (dense): Linear(in_features=768, out_features=768, bias=True)\r\n (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\r\n (dropout): Dropout(p=0.1, inplace=False)\r\n )\r\n )\r\n ...\r\n (pooler): BertPooler(\r\n (dense): Linear(in_features=768, out_features=768, bias=True)\r\n (activation): Tanh()\r\n )\r\n )\r\n (dropout): Dropout(p=0.1, inplace=False)\r\n (classifier): Linear(in_features=768, out_features=2, bias=True)\r\n)\r\n```\r\n\r\nSo basically `model` has 3 main submodules `bert`, `dropout`, and `classifier` (you can see this from the indentation as well.). Try running `model.bert`, `model.classifier`. When you call `model.bert `and freeze all the params, it will freeze entire encoder blocks(12 of them).\r\n\r\n\r\nTherefore, the following code\r\n```\r\nfor param in model.bert.bert.parameters():\r\n param.requires_grad = False\r\n```\r\nshould be\r\n\r\n```\r\nfor param in model.bert.parameters():\r\n param.requires_grad = False\r\n```", "> `model = BertForSequenceClassification.from_pretrained('bert-base-uncased')`\r\n> \r\n> When you run `model` you will get the following architecture (I skipped many intermediate modules)\r\n> \r\n> ```\r\n> BertForSequenceClassification(\r\n> (bert): BertModel(\r\n> (embeddings): BertEmbeddings(\r\n> ....\r\n> (encoder): BertEncoder(\r\n> (layer): ModuleList(\r\n> (0): BertLayer(\r\n> ...\r\n> (output): BertSelfOutput(\r\n> (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\r\n> (dropout): Dropout(p=0.1, inplace=False)\r\n> )\r\n> )\r\n> ...\r\n> (pooler): BertPooler(\r\n> (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> (activation): Tanh()\r\n> )\r\n> )\r\n> (dropout): Dropout(p=0.1, inplace=False)\r\n> (classifier): Linear(in_features=768, out_features=2, bias=True)\r\n> )\r\n> ```\r\n> \r\n> So basically `model` has 3 main submodules `bert`, `dropout`, and `classifier` (you can see this from the indentation as well.). Try running `model.bert`, `model.classifier`. When you call `model.bert `and freeze all the params, it will freeze entire encoder blocks(12 of them).\r\n> \r\n> Therefore, the following code\r\n> \r\n> ```\r\n> for param in model.bert.bert.parameters():\r\n> param.requires_grad = False\r\n> ```\r\n> \r\n> should be\r\n> \r\n> ```\r\n> for param in model.bert.parameters():\r\n> param.requires_grad = False\r\n> ```\r\n\r\nI think you are right. And I think there are no model.bert.bert? Or?", "> > `model = BertForSequenceClassification.from_pretrained('bert-base-uncased')`\r\n> > When you run `model` you will get the following architecture (I skipped many intermediate modules)\r\n> > ```\r\n> > BertForSequenceClassification(\r\n> > (bert): BertModel(\r\n> > (embeddings): BertEmbeddings(\r\n> > ....\r\n> > (encoder): BertEncoder(\r\n> > (layer): ModuleList(\r\n> > (0): BertLayer(\r\n> > ...\r\n> > (output): BertSelfOutput(\r\n> > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\r\n> > (dropout): Dropout(p=0.1, inplace=False)\r\n> > )\r\n> > )\r\n> > ...\r\n> > (pooler): BertPooler(\r\n> > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > (activation): Tanh()\r\n> > )\r\n> > )\r\n> > (dropout): Dropout(p=0.1, inplace=False)\r\n> > (classifier): Linear(in_features=768, out_features=2, bias=True)\r\n> > )\r\n> > ```\r\n> > \r\n> > \r\n> > So basically `model` has 3 main submodules `bert`, `dropout`, and `classifier` (you can see this from the indentation as well.). Try running `model.bert`, `model.classifier`. When you call `model.bert `and freeze all the params, it will freeze entire encoder blocks(12 of them).\r\n> > Therefore, the following code\r\n> > ```\r\n> > for param in model.bert.bert.parameters():\r\n> > param.requires_grad = False\r\n> > ```\r\n> > \r\n> > \r\n> > should be\r\n> > ```\r\n> > for param in model.bert.parameters():\r\n> > param.requires_grad = False\r\n> > ```\r\n> \r\n> I think you are right. And I think there are no model.bert.bert? Or?\r\n\r\nI think so, if you want to freeze bert part in bert model, u just say \r\n for param in model.bert.parameters():\r\n param.requires_grad = False\r\nbut how are about optizmer ? \r\n\r\nis look like in this way ? \r\n optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)", "> > > `model = BertForSequenceClassification.from_pretrained('bert-base-uncased')`\r\n> > > When you run `model` you will get the following architecture (I skipped many intermediate modules)\r\n> > > ```\r\n> > > BertForSequenceClassification(\r\n> > > (bert): BertModel(\r\n> > > (embeddings): BertEmbeddings(\r\n> > > ....\r\n> > > (encoder): BertEncoder(\r\n> > > (layer): ModuleList(\r\n> > > (0): BertLayer(\r\n> > > ...\r\n> > > (output): BertSelfOutput(\r\n> > > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > > (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\r\n> > > (dropout): Dropout(p=0.1, inplace=False)\r\n> > > )\r\n> > > )\r\n> > > ...\r\n> > > (pooler): BertPooler(\r\n> > > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > > (activation): Tanh()\r\n> > > )\r\n> > > )\r\n> > > (dropout): Dropout(p=0.1, inplace=False)\r\n> > > (classifier): Linear(in_features=768, out_features=2, bias=True)\r\n> > > )\r\n> > > ```\r\n> > > \r\n> > > \r\n> > > So basically `model` has 3 main submodules `bert`, `dropout`, and `classifier` (you can see this from the indentation as well.). Try running `model.bert`, `model.classifier`. When you call `model.bert `and freeze all the params, it will freeze entire encoder blocks(12 of them).\r\n> > > Therefore, the following code\r\n> > > ```\r\n> > > for param in model.bert.bert.parameters():\r\n> > > param.requires_grad = False\r\n> > > ```\r\n> > > \r\n> > > \r\n> > > should be\r\n> > > ```\r\n> > > for param in model.bert.parameters():\r\n> > > param.requires_grad = False\r\n> > > ```\r\n> > \r\n> > \r\n> > I think you are right. And I think there are no model.bert.bert? Or?\r\n> \r\n> I think so, if you want to freeze bert part in bert model, u just say\r\n> for param in model.bert.parameters():\r\n> param.requires_grad = False\r\n> but how are about optizmer ?\r\n> \r\n> is look like in this way ?\r\n> optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)\r\n\r\nI think you have written right code. But we should write usually 2 parts together. I mean:\r\nfor param in model.bert.parameters():\r\nparam.requires_grad = False\r\noptimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)\r\nCuz we need to know which parameters are frozen. ", "> > > > `model = BertForSequenceClassification.from_pretrained('bert-base-uncased')`\r\n> > > > When you run `model` you will get the following architecture (I skipped many intermediate modules)\r\n> > > > ```\r\n> > > > BertForSequenceClassification(\r\n> > > > (bert): BertModel(\r\n> > > > (embeddings): BertEmbeddings(\r\n> > > > ....\r\n> > > > (encoder): BertEncoder(\r\n> > > > (layer): ModuleList(\r\n> > > > (0): BertLayer(\r\n> > > > ...\r\n> > > > (output): BertSelfOutput(\r\n> > > > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > > > (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\r\n> > > > (dropout): Dropout(p=0.1, inplace=False)\r\n> > > > )\r\n> > > > )\r\n> > > > ...\r\n> > > > (pooler): BertPooler(\r\n> > > > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > > > (activation): Tanh()\r\n> > > > )\r\n> > > > )\r\n> > > > (dropout): Dropout(p=0.1, inplace=False)\r\n> > > > (classifier): Linear(in_features=768, out_features=2, bias=True)\r\n> > > > )\r\n> > > > ```\r\n> > > > \r\n> > > > \r\n> > > > So basically `model` has 3 main submodules `bert`, `dropout`, and `classifier` (you can see this from the indentation as well.). Try running `model.bert`, `model.classifier`. When you call `model.bert `and freeze all the params, it will freeze entire encoder blocks(12 of them).\r\n> > > > Therefore, the following code\r\n> > > > ```\r\n> > > > for param in model.bert.bert.parameters():\r\n> > > > param.requires_grad = False\r\n> > > > ```\r\n> > > > \r\n> > > > \r\n> > > > should be\r\n> > > > ```\r\n> > > > for param in model.bert.parameters():\r\n> > > > param.requires_grad = False\r\n> > > > ```\r\n> > > \r\n> > > \r\n> > > I think you are right. And I think there are no model.bert.bert? Or?\r\n> > \r\n> > \r\n> > I think so, if you want to freeze bert part in bert model, u just say\r\n> > for param in model.bert.parameters():\r\n> > param.requires_grad = False\r\n> > but how are about optizmer ?\r\n> > is look like in this way ?\r\n> > optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)\r\n> \r\n> I think you have written right code. But we should write usually 2 parts together. I mean:\r\n> for param in model.bert.parameters():\r\n> param.requires_grad = False\r\n> optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)\r\n> Cuz we need to know which parameters are frozen.\r\n\r\nsorry, I do not know what do you mean?\r\nso let say, if we just want to freeze the input part ( input embedding). how can I write the code? \r\nbert = BertModel.from_pretrained('bert-base-uncased')\r\n\r\nfor name, param in bert.named_parameters(): \r\n if name.startswith('embeddings'):\r\n param.requires_grad = False\r\nthis above three line, is tell model do not train the embedding right?\r\n\r\nbut how to tell the optimizer to do not change the embedding?\r\n\r\n", "> > > > > `model = BertForSequenceClassification.from_pretrained('bert-base-uncased')`\r\n> > > > > When you run `model` you will get the following architecture (I skipped many intermediate modules)\r\n> > > > > ```\r\n> > > > > BertForSequenceClassification(\r\n> > > > > (bert): BertModel(\r\n> > > > > (embeddings): BertEmbeddings(\r\n> > > > > ....\r\n> > > > > (encoder): BertEncoder(\r\n> > > > > (layer): ModuleList(\r\n> > > > > (0): BertLayer(\r\n> > > > > ...\r\n> > > > > (output): BertSelfOutput(\r\n> > > > > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > > > > (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)\r\n> > > > > (dropout): Dropout(p=0.1, inplace=False)\r\n> > > > > )\r\n> > > > > )\r\n> > > > > ...\r\n> > > > > (pooler): BertPooler(\r\n> > > > > (dense): Linear(in_features=768, out_features=768, bias=True)\r\n> > > > > (activation): Tanh()\r\n> > > > > )\r\n> > > > > )\r\n> > > > > (dropout): Dropout(p=0.1, inplace=False)\r\n> > > > > (classifier): Linear(in_features=768, out_features=2, bias=True)\r\n> > > > > )\r\n> > > > > ```\r\n> > > > > \r\n> > > > > \r\n> > > > > So basically `model` has 3 main submodules `bert`, `dropout`, and `classifier` (you can see this from the indentation as well.). Try running `model.bert`, `model.classifier`. When you call `model.bert `and freeze all the params, it will freeze entire encoder blocks(12 of them).\r\n> > > > > Therefore, the following code\r\n> > > > > ```\r\n> > > > > for param in model.bert.bert.parameters():\r\n> > > > > param.requires_grad = False\r\n> > > > > ```\r\n> > > > > \r\n> > > > > \r\n> > > > > should be\r\n> > > > > ```\r\n> > > > > for param in model.bert.parameters():\r\n> > > > > param.requires_grad = False\r\n> > > > > ```\r\n> > > > \r\n> > > > \r\n> > > > I think you are right. And I think there are no model.bert.bert? Or?\r\n> > > \r\n> > > \r\n> > > I think so, if you want to freeze bert part in bert model, u just say\r\n> > > for param in model.bert.parameters():\r\n> > > param.requires_grad = False\r\n> > > but how are about optizmer ?\r\n> > > is look like in this way ?\r\n> > > optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)\r\n> > \r\n> > \r\n> > I think you have written right code. But we should write usually 2 parts together. I mean:\r\n> > for param in model.bert.parameters():\r\n> > param.requires_grad = False\r\n> > optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)\r\n> > Cuz we need to know which parameters are frozen.\r\n> \r\n> sorry, I do not know what do you mean?\r\n> so let say, if we just want to freeze the input part ( input embedding). how can I write the code?\r\n> bert = BertModel.from_pretrained('bert-base-uncased')\r\n> \r\n> for name, param in bert.named_parameters():\r\n> if name.startswith('embeddings'):\r\n> param.requires_grad = False\r\n> this above three line, is tell model do not train the embedding right?\r\n> \r\n> but how to tell the optimizer to do not change the embedding?\r\n\r\nthese three lines have set False in 'Embedding' right? Then we need just to write what you have written:\r\noptimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=0.00001)", "Are any of the layers in a pre-trained BERT model originally frozen? If not are there any supporting articles or research papers that I can look to? ", "> Are any of the layers in a pre-trained BERT model originally frozen? If not are there any supporting articles or research papers that I can look to?\r\n\r\nI dont think there are any of layers originally frozen. You can read some Papers about Bert firstly, then in Pytoch forum or Tensorflow you could learn how to freeze the layers.", "In tensorflow, other than the classifier layers, to freeze some layers in a underlying base transformer model something like this maybe needed(corrections are welcome):\r\n\r\n```\r\nbertmodel = TFAutoModelForSequenceClassification.from_pretrained(bertModelName, num_labels=output_num_classes, from_pt=True)\r\n\r\nfor i,v in enumerate(bertmodel.layers[0].variables):\r\n if i>=19:\r\n bertmodel.layers[0].variables[i] = tf.stop_gradient(v)\r\n print(\"do freeze {} {}\".format(i, v.name))\r\n else:\r\n print(\"not freeze {} {}\".format(i, v.name))\r\n```", "If I want to freeze the entire model, I can use model.trainable = False and vice vs, however, if I want to freeze part of it, I am setting that up in the loop. It's still not updating the trainable flag. I am using TensorFlow. Any thoughts? \r\n\r\n```\r\n# Adjust the trainable layer weights based on retrain_layer_count\r\n # If retrain_layer_count is 0, then base model is frozen.\r\n # If retrain_layer_count is 12, then the entire base model is trainable.\r\n # And that implies that all the pretrained weights are lost and it relearns\r\n # from the input data.\r\n # If retrain_layer_count is between 1 and 11, then the last n layers of\r\n # the pretrained model retrained.\r\n if retrain_layer_count == 0:\r\n # The pretained model is frozen\r\n model.trainable = False \r\n\r\n elif retrain_layer_count == 12: \r\n # The pretrained model is retrained thru all layers. \r\n model.trainable = True \r\n\r\n else: \r\n # Restrict training to the num_train_layers outer transformer layers\r\n retrain_layer_list = []\r\n #model.trainable = False \r\n for retrain_layer_number in range(retrain_layer_count):\r\n\r\n layer_code = '_' + str(11 - retrain_layer_number)\r\n retrain_layer_list.append(layer_code)\r\n \r\n print('Retrain layers: \\n', retrain_layer_list)\r\n print(\"After adjusting...\")\r\n for layer in model.weights:\r\n layer._trainable = False\r\n print(\"***\", layer.name, layer._trainable)\r\n if 'layer_' in layer.name and layer.name.split(\".\")[1].split(\"/\")[0] in retrain_layer_list:\r\n layer._trainable = True\r\n print(\"$$$\", layer.name, layer._trainable)\r\n elif 'layer_' not in layer.name :\r\n layer._trainable = True\r\n print(\"###\", layer.name, layer._trainable)\r\n \r\n #for weight_details in model.weights:\r\n # print(weight_details.name, weight_details._trainable)\r\n print(f\"Number of trainable parameters : {count_params(model.trainable_weights)}\")\r\n print(f\"Number of non-trainable parameters : {count_params(model.non_trainable_variables)}\")\r\n```\r\n \r\n ", "> If I want to freeze the entire model, I can use model.trainable = False and vice vs, however, if I want to freeze part of it, I am setting that up in the loop. It's still not updating the trainable flag. I am using TensorFlow. Any thoughts?\r\n> \r\n> ```\r\n> # Adjust the trainable layer weights based on retrain_layer_count\r\n> # If retrain_layer_count is 0, then base model is frozen.\r\n> # If retrain_layer_count is 12, then the entire base model is trainable.\r\n> # And that implies that all the pretrained weights are lost and it relearns\r\n> # from the input data.\r\n> # If retrain_layer_count is between 1 and 11, then the last n layers of\r\n> # the pretrained model retrained.\r\n> if retrain_layer_count == 0:\r\n> # The pretained model is frozen\r\n> model.trainable = False \r\n> \r\n> elif retrain_layer_count == 12: \r\n> # The pretrained model is retrained thru all layers. \r\n> model.trainable = True \r\n> \r\n> else: \r\n> # Restrict training to the num_train_layers outer transformer layers\r\n> retrain_layer_list = []\r\n> #model.trainable = False \r\n> for retrain_layer_number in range(retrain_layer_count):\r\n> \r\n> layer_code = '_' + str(11 - retrain_layer_number)\r\n> retrain_layer_list.append(layer_code)\r\n> \r\n> print('Retrain layers: \\n', retrain_layer_list)\r\n> print(\"After adjusting...\")\r\n> for layer in model.weights:\r\n> layer._trainable = False\r\n> print(\"***\", layer.name, layer._trainable)\r\n> if 'layer_' in layer.name and layer.name.split(\".\")[1].split(\"/\")[0] in retrain_layer_list:\r\n> layer._trainable = True\r\n> print(\"$$$\", layer.name, layer._trainable)\r\n> elif 'layer_' not in layer.name :\r\n> layer._trainable = True\r\n> print(\"###\", layer.name, layer._trainable)\r\n> \r\n> #for weight_details in model.weights:\r\n> # print(weight_details.name, weight_details._trainable)\r\n> print(f\"Number of trainable parameters : {count_params(model.trainable_weights)}\")\r\n> print(f\"Number of non-trainable parameters : {count_params(model.non_trainable_variables)}\")\r\n> ```\r\n\r\nI ran into this issue too, where if I first freeze all layers, and then selectively make some trainable, it does not work. Do it the other way round: set everything trainable and then freeze the layers you want!" ]

[]

[]

[ "Hi Aly, GPT-2 is pretrained on an English only corpus.", "Hi @thomwolf , thank you for the clarification." ]

[]

[]

[ "Hi @dirkgr, thanks for this PR.\r\n\r\nI think we will keep all the GPU/multi-GPU logic outside of the main library for now. It makes it easier to integrate the module in downstream libraries and integrating such modifications at the current stage would cause too many breaking changes for the users unfortunately.\r\n\r\nSo I'm closing this PR for now.", "Oh, this was only meant for internal consumption.\r\n\r\nThat said, I thought a bit about how to make it general enough to properly integrate it. I thought the setting for multi-GPU could be part of the config, so it wouldn't affect normal operation. How would you want to do it?\r\n\r\nI didn't get that much interest in this patch internally though, so this is not the most important project right now." ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "@thomwolf Please re-open. If I have the time, I can try to work on this.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Bump so I don't forget about this.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "tests/tokenization_openai_test.py::OpenAIGPTTokenizationTest::test_full_tokenizer FAILED", "Ok, thanks!" ]

[ { "id": 1108649070, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDcw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Need%20more%20information", "name": "Need more information", "color": "d876e3", "default": false, "description": "Further information is requested" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hi,\r\nWhat command are you running to load the model?\r\nAre you loading your own model or one or our pre-trained one?\r\n\r\nIt's normal that there is no `seq_relationship` attribute in a `BertOnlyMLMHead` but our pre-trained model should load without error.", "\r\nI was loading my own pre-trained model by \"convert_tf_checkpoint_to_pytorch.py\". Would it be ok if I skip \"seq_relationship\" in the source code? ", "Maybe, I don't have enough information to really be able to help you at this stage.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Thanks!" ]

[ { "id": 1108649070, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDcw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Need%20more%20information", "name": "Need more information", "color": "d876e3", "default": false, "description": "Further information is requested" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hi,\r\nIs it a model trained from the original Google BERT Tensorflow implementation?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "I get the same error as @leejason. I used the pre-trained BERT base uncased model from the original TF implementation and fine-tuned it on my own training set. I am now trying to use the fine-tuned model to do masked LM. ", "I also have similar issue. I pretrained a bert **from scratch** using **nvidia implementation** with **customized config file and vocab**.\r\nThen I use\r\n```\r\nconvert_tf_checkpoint_to_pytorch.convert_tf_checkpoint_to_pytorch(BERT_MODEL_PATH + 'model.ckpt',\r\n BERT_MODEL_PATH + 'bert_config.json',\r\n BERT_MODEL_PATH + 'pytorch_model.bin')\r\n```\r\n```\r\n...\r\nLoading TF weight cls/predictions/transform/dense/bias with shape [512]\r\nLoading TF weight cls/predictions/transform/dense/bias/adam_m with shape [512]\r\nLoading TF weight cls/predictions/transform/dense/bias/adam_v with shape [512]\r\nLoading TF weight cls/predictions/transform/dense/kernel with shape [512, 512]\r\nLoading TF weight cls/predictions/transform/dense/kernel/adam_m with shape [512, 512]\r\nLoading TF weight cls/predictions/transform/dense/kernel/adam_v with shape [512, 512]\r\nLoading TF weight cls/seq_relationship/output_bias with shape [2]\r\nLoading TF weight cls/seq_relationship/output_weights with shape [2, 512]\r\nLoading TF weight global_step with shape []\r\nLoading TF weight good_steps with shape []\r\nLoading TF weight loss_scale with shape []\r\nSkipping bad_steps\r\n---------------------------------------------------------------------------\r\nAttributeError Traceback (most recent call last)\r\n<ipython-input-20-f28c5910adfc> in <module>\r\n----> 1 bert = BertModel.from_pretrained(BERT_MODEL_PATH, from_tf=True).bert()\r\n\r\n~/InEx/input/huggingface/pytorch-pretrained-BERT-master/pytorch_pretrained_bert/modeling.py in from_pretrained(cls, pretrained_model_name_or_path, *inputs, **kwargs)\r\n 611 # Directly load from a TensorFlow checkpoint\r\n 612 weights_path = os.path.join(serialization_dir, TF_WEIGHTS_NAME)\r\n--> 613 return load_tf_weights_in_bert(model, weights_path)\r\n 614 # Load from a PyTorch state_dict\r\n 615 old_keys = []\r\n\r\n~/InEx/input/huggingface/pytorch-pretrained-BERT-master/pytorch_pretrained_bert/modeling.py in load_tf_weights_in_bert(model, tf_checkpoint_path)\r\n 107 array = np.transpose(array)\r\n 108 try:\r\n--> 109 assert pointer.shape == array.shape\r\n 110 except AssertionError as e:\r\n 111 e.args += (pointer.shape, array.shape)\r\n\r\n~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py in __getattr__(self, name)\r\n 537 return modules[name]\r\n 538 raise AttributeError(\"'{}' object has no attribute '{}'\".format(\r\n--> 539 type(self).__name__, name))\r\n 540 \r\n 541 def __setattr__(self, name, value):\r\n\r\nAttributeError: 'BertModel' object has no attribute 'shape'```", "I solve this problem by bypass some variables in the model, such as \"bad_steps\", “global_step\", \"good_steps\", \"loss_scale\". They don't have attribute 'shape‘ and I don't need them when fineturning the model.\r\n\r\nIn modeling.py, line 121, replace it with \r\n if any(n in [\"adam_v\", \"adam_m\", \"global_step\", \"bad_steps\", \"global_step\", \"good_steps\", \"loss_scale\"] for n in name):\r\nand delete line 151-156.", "> I solve this problem by bypass some variables in the model, such as \"bad_steps\", “global_step\", \"good_steps\", \"loss_scale\". They don't have attribute 'shape‘ and I don't need them when fineturning the model.\r\n> \r\n> In modeling.py, line 121, replace it with\r\n> if any(n in [\"adam_v\", \"adam_m\", \"global_step\", \"bad_steps\", \"global_step\", \"good_steps\", \"loss_scale\"] for n in name):\r\n> and delete line 151-156.\r\n\r\nIt works. Thanks very much !", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "I'm getting a similar error when trying to convert the newer BERT models released at\r\n[tensorflow/models/tree/master/official/nlp/](https://github.com/tensorflow/models/tree/master/official/nlp/bert#pre-trained-models).\r\n\r\nThese models are either BERT models trained with Keras or else checkpoints converted from \r\nthe original [google-research/bert](https://github.com/google-research/bert) repository. I also get the same error when I convert the TF1 to TF2 checkpoints myself using the [tf2_encoder_checkpoint_converter.py](https://github.com/tensorflow/models/blob/master/official/nlp/bert/tf2_encoder_checkpoint_converter.py) script: \r\n\r\nWhat I have tried:\r\n\r\nFirst, I have downloaded a model:\r\n```\r\nwget https://storage.googleapis.com/cloud-tpu-checkpoints/bert/keras_bert/cased_L-12_H-768_A-12.tar.gz\r\n# or\r\nwget https://storage.googleapis.com/cloud-tpu-checkpoints/bert/tf_20/cased_L-12_H-768_A-12.tar.gz \r\n```\r\nAfter unpacking:\r\n```\r\nexport BERT_BASE_DIR=cased_L-12_H-768_A-12\r\n\r\ntransformers-cli convert --model_type bert \\\r\n --tf_checkpoint $BERT_BASE_DIR/bert_model.ckpt \\\r\n --config $BERT_BASE_DIR/bert_config.json \\\r\n --pytorch_dump_output $BERT_BASE_DIR/pytorch_model.bin\r\n```\r\n\r\nThe command prints the configuration but throws the following error: \r\n\r\n```\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_layer/_value_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [768, 12, 64]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_layer_norm/beta/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_layer_norm/gamma/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_output_dense/bias/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_output_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [12, 64, 768]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_intermediate_dense/bias/.ATTRIBUTES/VARIABLE_VALUE with shape [3072]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_intermediate_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [768, 3072]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_dense/bias/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [3072, 768]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_layer_norm/beta/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\nINFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_layer_norm/gamma/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\nINFO:transformers.modeling_bert:Loading TF weight save_counter/.ATTRIBUTES/VARIABLE_VALUE with shape []\r\nINFO:transformers.modeling_bert:Skipping _CHECKPOINTABLE_OBJECT_GRAPH\r\nTraceback (most recent call last):\r\n File \"/home/jbarry/anaconda3/envs/transformers/bin/transformers-cli\", line 30, in <module>\r\n service.run()\r\n File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/transformers/commands/convert.py\", line 62, in run\r\n convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output)\r\n File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/transformers/convert_bert_original_tf_checkpoint_to_pytorch.py\", line 36, in convert_tf_checkpoint_to_pytorch\r\n load_tf_weights_in_bert(model, config, tf_checkpoint_path)\r\n File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/transformers/modeling_bert.py\", line 118, in load_tf_weights_in_bert\r\n assert pointer.shape == array.shape\r\n File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/torch/nn/modules/module.py\", line 585, in __getattr__\r\n type(self).__name__, name))\r\nAttributeError: 'BertForPreTraining' object has no attribute 'shape'\r\n```\r\n\r\nThis is happening in a fresh environment with PyTorch 1.3 installed in Anaconda (Linux), as well as pip-installing `tf-nightly` and `transformers` (2.3.0).\r\n\r\nHas anyone else been able to successfully convert the TF 2.0 version models to PyTorch or know where I'm going wrong? Thanks!", "> I'm getting a similar error when trying to convert the newer BERT models released at\r\n> [tensorflow/models/tree/master/official/nlp/](https://github.com/tensorflow/models/tree/master/official/nlp/bert#pre-trained-models).\r\n> \r\n> These models are either BERT models trained with Keras or else checkpoints converted from\r\n> the original [google-research/bert](https://github.com/google-research/bert) repository. I also get the same error when I convert the TF1 to TF2 checkpoints myself using the [tf2_encoder_checkpoint_converter.py](https://github.com/tensorflow/models/blob/master/official/nlp/bert/tf2_encoder_checkpoint_converter.py) script:\r\n> \r\n> What I have tried:\r\n> \r\n> First, I have downloaded a model:\r\n> \r\n> ```\r\n> wget https://storage.googleapis.com/cloud-tpu-checkpoints/bert/keras_bert/cased_L-12_H-768_A-12.tar.gz\r\n> # or\r\n> wget https://storage.googleapis.com/cloud-tpu-checkpoints/bert/tf_20/cased_L-12_H-768_A-12.tar.gz \r\n> ```\r\n> \r\n> After unpacking:\r\n> \r\n> ```\r\n> export BERT_BASE_DIR=cased_L-12_H-768_A-12\r\n> \r\n> transformers-cli convert --model_type bert \\\r\n> --tf_checkpoint $BERT_BASE_DIR/bert_model.ckpt \\\r\n> --config $BERT_BASE_DIR/bert_config.json \\\r\n> --pytorch_dump_output $BERT_BASE_DIR/pytorch_model.bin\r\n> ```\r\n> \r\n> The command prints the configuration but throws the following error:\r\n> \r\n> ```\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_layer/_value_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [768, 12, 64]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_layer_norm/beta/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_layer_norm/gamma/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_output_dense/bias/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_attention_output_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [12, 64, 768]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_intermediate_dense/bias/.ATTRIBUTES/VARIABLE_VALUE with shape [3072]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_intermediate_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [768, 3072]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_dense/bias/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_dense/kernel/.ATTRIBUTES/VARIABLE_VALUE with shape [3072, 768]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_layer_norm/beta/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\n> INFO:transformers.modeling_bert:Loading TF weight model/layer_with_weights-9/_output_layer_norm/gamma/.ATTRIBUTES/VARIABLE_VALUE with shape [768]\r\n> INFO:transformers.modeling_bert:Loading TF weight save_counter/.ATTRIBUTES/VARIABLE_VALUE with shape []\r\n> INFO:transformers.modeling_bert:Skipping _CHECKPOINTABLE_OBJECT_GRAPH\r\n> Traceback (most recent call last):\r\n> File \"/home/jbarry/anaconda3/envs/transformers/bin/transformers-cli\", line 30, in <module>\r\n> service.run()\r\n> File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/transformers/commands/convert.py\", line 62, in run\r\n> convert_tf_checkpoint_to_pytorch(self._tf_checkpoint, self._config, self._pytorch_dump_output)\r\n> File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/transformers/convert_bert_original_tf_checkpoint_to_pytorch.py\", line 36, in convert_tf_checkpoint_to_pytorch\r\n> load_tf_weights_in_bert(model, config, tf_checkpoint_path)\r\n> File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/transformers/modeling_bert.py\", line 118, in load_tf_weights_in_bert\r\n> assert pointer.shape == array.shape\r\n> File \"/home/jbarry/anaconda3/envs/transformers/lib/python3.6/site-packages/torch/nn/modules/module.py\", line 585, in __getattr__\r\n> type(self).__name__, name))\r\n> AttributeError: 'BertForPreTraining' object has no attribute 'shape'\r\n> ```\r\n> \r\n> This is happening in a fresh environment with PyTorch 1.3 installed in Anaconda (Linux), as well as pip-installing `tf-nightly` and `transformers` (2.3.0).\r\n> \r\n> Has anyone else been able to successfully convert the TF 2.0 version models to PyTorch or know where I'm going wrong? Thanks!\r\n\r\nignore those lines causing erros by changing\r\nhttps://github.com/huggingface/transformers/blob/master/src/transformers/modeling_bert.py#L123-L127 to \r\n try:\r\n assert pointer.shape == array.shape\r\n except:\r\n pass\r\n\r\nsame thing for https://github.com/huggingface/transformers/blob/master/src/transformers/modeling_bert.py#L129\r\n", "> \r\n> \r\n> I solve this problem by bypass some variables in the model, such as \"bad_steps\", “global_step\", \"good_steps\", \"loss_scale\". They don't have attribute 'shape‘ and I don't need them when fineturning the model.\r\n> \r\n> In modeling.py, line 121, replace it with\r\n> if any(n in [\"adam_v\", \"adam_m\", \"global_step\", \"bad_steps\", \"global_step\", \"good_steps\", \"loss_scale\"] for n in name):\r\n> and delete line 151-156.\r\n\r\nIt helps! Thx u so much <3", "I'm running into the same problem, I tried the solution proposed by @yzhang123 but it only makes me run in another error \r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\lrizzello\\AppData\\Local\\JetBrains\\PyCharm 2019.3.4\\plugins\\python\\helpers\\pydev\\pydevd.py\", line 1434, in _exec\r\n pydev_imports.execfile(file, globals, locals) # execute the script\r\n File \"C:\\Users\\lrizzello\\AppData\\Local\\JetBrains\\PyCharm 2019.3.4\\plugins\\python\\helpers\\pydev\\_pydev_imps\\_pydev_execfile.py\", line 18, in execfile\r\n exec(compile(contents+\"\\n\", file, 'exec'), glob, loc)\r\n File \"C:/source/repos/DeduplicationSiamese/PythonDeduper/eval/eval_matcher_tf.py\", line 91, in <module>\r\n load_tf_weights_in_bert(model, config, tf_path)\r\n File \"C:\\Users\\lrizzello\\Anaconda3\\envs\\dedupe_transformer\\lib\\site-packages\\transformers\\modeling_bert.py\", line 129, in load_tf_weights_in_bert\r\n pointer.data = torch.from_numpy(array)\r\nTypeError: expected np.ndarray (got bytes)\r\n\r\nProcess finished with exit code -1\r\n\r\n```\r\n\r\nI got those checkpoints by training an existing huggingface model (namely 'google/bert_uncased_L-12_H-256_A-4') via the TFTrainer/TFTrainingArguments method\r\n\r\nI have tried many other hacks to get this working, such as [this one](https://stackoverflow.com/questions/60539758/biobert-for-keras-version-of-huggingface-transformers) or [this one](https://github.com/huggingface/transformers/issues/676#issuecomment-502101078) but unsuccesfully but nothing worked. I keep running into error after error.\r\n\r\nHas anyone managed to get this working any other way?", "I managed to get it working by going through the pointers in debug mode and checking what variable name corresponded to what. This is the function I ended up using.\r\n\r\n```\r\ndef convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, bert_config_file, pytorch_dump_path):\r\n config_path = os.path.abspath(bert_config_file)\r\n tf_path = os.path.abspath(tf_checkpoint_path)\r\n print(\"Converting TensorFlow checkpoint from {} with config at {}\".format(tf_path, config_path))\r\n # Load weights from TF model\r\n init_vars = tf.train.list_variables(tf_path)\r\n excluded = [\"BERTAdam\", \"_power\", \"global_step\", \"_CHECKPOINTABLE_OBJECT_GRAPH\"]\r\n init_vars = list(filter(lambda x: all([True if e not in x[0] else False for e in excluded]), init_vars))\r\n names = []\r\n arrays = []\r\n for name, shape in init_vars:\r\n print(\"Loading TF weight {} with shape {}\".format(name, shape))\r\n array = tf.train.load_variable(tf_path, name)\r\n names.append(name)\r\n arrays.append(array)\r\n\r\n config = BertConfig.from_json_file(bert_config_file)\r\n print(\"Building PyTorch model from configuration: {}\".format(str(config)))\r\n # Initialise PyTorch model\r\n model = BertForSequenceClassification(config)\r\n\r\n for name, array in zip(names, arrays):\r\n name = name.split(\"/\")\r\n # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v\r\n # which are not required for using pretrained model\r\n if any(n in [\"adam_v\", \"adam_m\", \"global_step\", \"bad_steps\", \"global_step\", \"good_steps\", \"loss_scale\",\r\n \"AdamWeightDecayOptimizer\", \"AdamWeightDecayOptimizer_1\", \"save_counter\", \".OPTIMIZER_SLOT\"] for n in name) or \\\r\n name[0] == \"optimizer\":\r\n print(\"Skipping {}\".format(\"/\".join(name)))\r\n continue\r\n if \".OPTIMIZER_SLOT\" in name:\r\n idx = name.index(\".OPTIMIZER_SLOT\")\r\n name = name[:idx]\r\n elif \".ATTRIBUTES\" in name:\r\n idx = name.index(\".ATTRIBUTES\")\r\n name = name[:idx]\r\n print(name)\r\n pointer = model\r\n for m_name in name:\r\n if re.fullmatch(r\"[A-Za-z]+_\\d+\", m_name):\r\n scope_names = re.split(r\"_(\\d+)\", m_name)\r\n else:\r\n scope_names = [m_name]\r\n if scope_names[0] == \"kernel\" or scope_names[0] == \"gamma\":\r\n pointer = getattr(pointer, \"weight\")\r\n elif scope_names[0] == \"output_bias\" or scope_names[0] == \"beta\":\r\n pointer = getattr(pointer, \"bias\")\r\n elif scope_names[0] == \"output_weights\":\r\n pointer = getattr(pointer, \"weight\")\r\n elif scope_names[0] == \"squad\":\r\n pointer = getattr(pointer, \"classifier\")\r\n elif scope_names[0] == \"dense_output\" or scope_names[0] == \"bert_output\":\r\n pointer = getattr(pointer, \"output\")\r\n elif scope_names[0] == \"self_attention\":\r\n pointer = getattr(pointer, \"self\")\r\n else:\r\n try:\r\n pointer = getattr(pointer, scope_names[0])\r\n except AttributeError:\r\n logger.info(\"Skipping {}\".format(\"/\".join(name)))\r\n continue\r\n if len(scope_names) >= 2:\r\n num = int(scope_names[1])\r\n pointer = pointer[num]\r\n if m_name[-11:] == \"_embeddings\":\r\n pointer = getattr(pointer, \"weight\")\r\n elif m_name == \"kernel\" or m_name == \"gamma\" or m_name == \"output_weights\":\r\n array = np.transpose(array)\r\n # print(\"Initialize PyTorch weight {}\".format(name))\r\n pointer.data = torch.from_numpy(array)\r\n\r\n # Save pytorch-model\r\n print(\"Save PyTorch model to {}\".format(pytorch_dump_path))\r\n torch.save(model.state_dict(), pytorch_dump_path)\r\n\r\n\r\nconvert_tf_checkpoint_to_pytorch(tf_path, config_path, pytorch_dump_path)\r\n```\r\n\r\n", "Hi, this is an actual programming error in modeling_bert.py. If you look at line 145 it's pretty obvious that the code should be continuing to the next iteration of the *outer* loop (over name, array) rather than the inner one (over the path components of name) - otherwise why would the error messages say \"skipping {name}\":\r\n\r\nhttps://github.com/huggingface/transformers/blob/master/src/transformers/models/bert/modeling_bert.py#L145\r\n\r\nTo fix this, simply extract the try/except block so that it wraps the entire loop (lines 127-148). I would supply a patch but I have to work with transformers 3.5.1 for the moment since I'm using sentence-transformers which hasn't been updated to the latest version.", "@thomwolf If the above fix will be added to the master branch this will be great\r\nhttps://github.com/smartshark/transformers/pull/1\r\n", "> \r\n\r\nI revised `modeling_bert.by` following @lrizzello 's code and could save tf1 checkpoint I personally trained into pytorch. I first changed tf1 checkpoint to tf2, and then used the below code. Here is the code I revised in `modeling_bert.py`\r\n \r\n```\r\ndef load_tf_weights_in_bert(model, config, tf_checkpoint_path):\r\n try:\r\n import re\r\n import numpy as np\r\n import tensorflow as tf\r\n except ImportError:\r\n logger.error(\r\n \"Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see \"\r\n \"https://www.tensorflow.org/install/ for installation instructions.\"\r\n )\r\n raise\r\n tf_path = os.path.abspath(tf_checkpoint_path)\r\n logger.info(f\"Converting TensorFlow checkpoint from {tf_path}\")\r\n # Load weights from TF model\r\n init_vars = tf.train.list_variables(tf_path)\r\n names = []\r\n arrays = []\r\n for name, shape in init_vars:\r\n logger.info(f\"Loading TF weight {name} with shape {shape}\")\r\n array = tf.train.load_variable(tf_path, name)\r\n names.append(name)\r\n arrays.append(array)\r\n for name, array in zip(names, arrays):\r\n name = name.split(\"/\")\r\n # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v\r\n # which are not required for using pretrained model\r\n if any(\r\n [\"adam_v\", \"adam_m\", \"global_step\", \"bad_steps\", \"global_step\", \"good_steps\", \"loss_scale\",\r\n \"AdamWeightDecayOptimizer\", \"AdamWeightDecayOptimizer_1\", \"save_counter\", \".OPTIMIZER_SLOT\"] for n in name) or \\\r\n name[0] == \"optimizer\":\r\n # n in [\"adam_v\", \"adam_m\", \"AdamWeightDecayOptimizer\", \"AdamWeightDecayOptimizer_1\", \"global_step\"]\r\n # for n in name\r\n # ):\r\n logger.info(f\"Skipping {'/'.join(name)}\")\r\n continue\r\n if \".OPTIMIZER_SLOT\" in name:\r\n idx = name.index(\".OPTIMIZER_SLOT\")\r\n name = name[:idx]\r\n elif \".ATTRIBUTES\" in name:\r\n idx = name.index(\".ATTRIBUTES\")\r\n name = name[:idx]\r\n print(name)\r\n pointer = model\r\n for m_name in name:\r\n if re.fullmatch(r\"[A-Za-z]+_\\d+\", m_name):\r\n scope_names = re.split(r\"_(\\d+)\", m_name)\r\n else:\r\n scope_names = [m_name]\r\n if scope_names[0] == \"kernel\" or scope_names[0] == \"gamma\":\r\n pointer = getattr(pointer, \"weight\")\r\n elif scope_names[0] == \"output_bias\" or scope_names[0] == \"beta\":\r\n pointer = getattr(pointer, \"bias\")\r\n elif scope_names[0] == \"output_weights\":\r\n pointer = getattr(pointer, \"weight\")\r\n elif scope_names[0] == \"squad\":\r\n pointer = getattr(pointer, \"classifier\")\r\n elif scope_names[0] == \"dense_output\" or scope_names[0] == \"bert_output\":\r\n pointer = getattr(pointer, \"output\")\r\n elif scope_names[0] == \"self_attention\":\r\n pointer = getattr(pointer, \"self\")\r\n else:\r\n try:\r\n pointer = getattr(pointer, scope_names[0])\r\n except AttributeError:\r\n logger.info(\"Skipping {}\".format(\"/\".join(name)))\r\n continue\r\n if len(scope_names) >= 2:\r\n num = int(scope_names[1])\r\n pointer = pointer[num]\r\n if m_name[-11:] == \"_embeddings\":\r\n pointer = getattr(pointer, \"weight\")\r\n elif m_name == \"kernel\" or m_name == \"gamma\" or m_name == \"output_weights\":\r\n array = np.transpose(array)\r\n # try:\r\n # if pointer.shape != array.shape:\r\n # raise ValueError(f\"Pointer shape {pointer.shape} and array shape {array.shape} mismatched\")\r\n # except AssertionError as e:\r\n # e.args += (pointer.shape, array.shape)\r\n # raise\r\n logger.info(f\"Initialize PyTorch weight {name}\")\r\n pointer.data = torch.from_numpy(array)\r\n return model\r\n```\r\n\r\nFor convert_tf_to_pytorch function, I used below.\r\n\r\n```\r\nimport argparse\r\nimport os\r\nimport torch\r\n\r\nfrom transformers import BertConfig, BertForPreTraining, load_tf_weights_in_bert\r\nfrom transformers.utils import logging\r\n\r\nlogging.set_verbosity_info()\r\ndef convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, bert_config_file, pytorch_dump_path):\r\n # Initialise PyTorch model\r\n config = BertConfig.from_json_file(bert_config_file)\r\n print(f\"Building PyTorch model from configuration: {config}\")\r\n model = BertForPreTraining(config)\r\n # Load weights from tf checkpoint\r\n load_tf_weights_in_bert(model, config, tf_checkpoint_path)\r\n # Save pytorch-model\r\n os.makedirs(pytorch_dump_path)\r\n pytorch_dump_path = os.path.join(pytorch_dump_path, '0')\r\n print(f\"Save PyTorch model to {pytorch_dump_path}\")\r\n torch.save(model.state_dict(), pytorch_dump_path)\r\nif __name__ == \"__main__\":\r\n parser = argparse.ArgumentParser()\r\n # Required parameters\r\n parser.add_argument(\r\n \"--tf_checkpoint_path\", default=None, type=str, required=True, help=\"Path to the TensorFlow checkpoint path.\"\r\n )\r\n parser.add_argument(\r\n \"--bert_config_file\",\r\n default=None,\r\n type=str,\r\n required=True,\r\n help=\"The config json file corresponding to the pre-trained BERT model. \\n\"\r\n \"This specifies the model architecture.\",\r\n )\r\n parser.add_argument(\r\n \"--pytorch_dump_path\", default=None, type=str, required=True, help=\"Path to the output PyTorch model.\"\r\n )\r\n args = parser.parse_args()\r\n convert_tf_checkpoint_to_pytorch(args.tf_checkpoint_path, args.bert_config_file, args.pytorch_dump_path)\r\n```\r\n\r\nHope this help!", "I use @Jwmc999 's method and successfully converted tensoeflow 2.x ckpt file to Pytorch.bin.\r\nCheck the below jupyterbook if needed.\r\n[Converting-TF-ckpt-to-Pytorch-model](https://github.com/suchunxie/Converting-TF-ckpt-to-Pytorch-model) " ]

[]

[]

[ "Also, build_py2 has failed because I deliberately did not include Py2 compatibility - it's only a few months from end-of-life now. If you really want me to, I can go back and include it, but we should be trying to let it go by now!", "This is really great @Rocketknight1!\r\nThanks for taking the time to make a very clear and informative README.\r\nI've fixed the formatting issues in the README, slightly re-worded it and added a link to it in the main README.\r\nI'm merging it now.\r\nGreat job!" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Cola is probably one of the most unstable tasks for BERT. For us it mostly boiled down to running many seeds. If all you care about is a good pre-trained model checkpoint, we have a 65 / 61 run at https://github.com/zphang/bert_on_stilts ", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "@cooelf hi, can you reproduce cola score of BERT now? I work on this recently but still can't reach the reported score on the test set, even if restart with different seed multiple times, and I further noticed that improvement on dev set is inconsistent with that on the test set at all.", "Thanks for all the feedbacks! \r\n@scissorsy The results of dev and test are indeed incosistent. I have changed the seeds, lr, warmup rates and max_epochs, and finally the test result reached about 60% for some runs. Using multi-tasking or transfer from bigger dataset such as MNLI seems to be more stable.\r\n\r\n", "> Thanks for all the feedbacks!\r\n> @scissorsy The results of dev and test are indeed incosistent. I have changed the seeds, lr, warmup rates and max_epochs, and finally the test result reached about 60% for some runs. Using multi-tasking or transfer from bigger dataset such as MNLI seems to be more stable.\r\n\r\nThanks!", "Hi @cooelf, what parameter number did you change in order to fit a better result, thanks!" ]

[]

[]

[ "Me too.It is probably Network connection problem.", "The network connection check has been relaxed in the now merged #500.\r\nIt will be included in the next PyPI release (probably next week).\r\nIn the meantime you can install from `master`.", "@thomwolf Thank you.", "The new release is on pypi!" ]

[]

[]

[ "Hi @lukovnikov, yes I think the various schedules you have added to your fork are very nice!\r\nDo you want to add them in this PR as well?\r\nOtherwise, I'll merge it.", "Merging it for now. Thanks @lukovnikov ", "Hi, sorry, lost track of this, will make a new PR soon." ]

[]

[]

[ "Hi @ananyahjha93,\r\nThanks for this PR.\r\nDo you have some results from fine-tuning BERT on the other tasks?\r\nAlso, I think we should add some details on the available tasks in the readme as well.", "@thomwolf I have added results on GLUE dev set in the README and details on how to run any GLUE task. But, I have also added a warning stating that all GLUE tasks have not been tested with half-precision training. With the new tasks and metrics being added, there should be one round of tests with half-precision training as well. Unfortunately, I do not have direct access to a V100 or any of the RTX cards in order to run that. \r\n\r\nAlso, for some reason the previous pull request was passing each build_py2 test but now it is failing the 'OpenAIGPTTokenizationTest'. I only added changes to the README in my latest commit.", "This looks great, thanks @ananyahjha93!" ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "It is better to imitate original Bert repo to add separate argument `args.vocab_file`, and during prediction, argument `bert_model` is the directory containing the fine-tuned model. ", "Make sense indeed. Would you like to submit a PR on that?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Hi Catalin, thanks for that!\r\nYes backward compatibility is an important things to keep in mind here.\r\nI think the changes look nice.\r\nWe should also:\r\n- document them in the readme, and\r\n- add associated tests in the relevant test files.", "Just before this gets merged - I've noticed that the GPT(1) tokenizer ignores whether or not a string contains special tokens, and hence doesn't encode them properly. \r\nWe've got around this by splitting `text` on whitespace inside the `tokenize` method and iterating, checking if a word is a special token or not, like so:\r\n```\r\ndef tokenize(self, text):\r\n split_tokens = []\r\n for t in text.split(' '):\r\n if t in self.special_tokens:\r\n split_tokens.extend([t])\r\n else:\r\n #current tokenization stuff\r\n```\r\nWould be nice if this could be included :)", "@andrewPoulton we can include this, but only for the whitespace tokenizer fallback of GPT-1. The original tokenizer (SpaCy) would split tokens like `<\\w>` in pieces which is the reason it was not included originally.\r\n\r\nOverall I must say I'm not a huge fan of the not-split-specific-tokens feature (the `never_split` option). We've added it due to popular request but it is very dependent on the underlying behavior of the tokenizer and the character content of special tokens (does it contains spaces, dashes...) and from the issues it looks like a common source of bugs and un-intended behavior (see #410 for a latest example).", "Hi @CatalinVoss, from the state of the PR I understand you are still working on it.\r\nMaybe ping me when you think it's ready for merging in master and I'll have a look again?", "Hey @thomwolf yeah, just didn't get around to it yet, but then I needed the BERT decoding piece yesterday so I merged it in. It's imperfect. If we renamed everything to be consistent with words, tokens, token IDs, etc. we would have to change the method names, per my comment above. Do you want to do that? Otherwise perhaps better to do that in a separate PR and target some sort of v0.7 branch?", "Ok we still have to add the docstrings, test and details in the readme for these methods.\r\nHaven't find time this week. I will see if I can find time next week. ", "@CatalinVoss I also was wondering about this while writing some extra functionality for the GPT2 tokenizer to encode new strings for finetuning tasks, but ended up writing a `special_token_to_id` functionality for getting IDs of special tokens -- the pattern of usage of the GPT1 tokenizer for finetuning tasks seems to be to add the special token IDs after encoding the rest of the string to process.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Hey @thomwolf, what was the decision on this? I can revisit if you want. We still wanted this in our fork… thanks much!", "I never really had time to add tests and documentation to the PR but it's a good idea.\r\nLet's add this feature in the new release.", "Looks like this was taken care of with your refactor in `tokenization_utils.py`. Very nice!!" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "We can't now. The code is still incomplete. Is it possible recently? I really want to help but not familiar with tensorflow.", "A related issue is #376.\r\n\r\nHowever, pytorch-pretraned-BERT was mostly designed to provide easy and fast access to pretrained models.\r\n\r\nIf you want to train a BERT model from scratch you will need a more robust code base for training and data-processing than the simple examples that are provided in this repo.\r\n\r\nI would probably advise to move to a more integrated codebase like the nice [XLM repo](https://github.com/facebookresearch/XLM) of @glample and @aconneau.", "I've been able to use the codebase for this, and didn't see much issues, however I might be overlooking something. If you construct and initialize a new model instead of loading from pretrained, you can use the `simple_lm_finetuning` script to train on new data.\r\n\r\nThomas, did you have any specific other issues in mind? ", "NVidia recently [released](https://medium.com/future-vision/bert-meets-gpus-403d3fbed848?fbclid=IwAR0bFskUVVKDRyYF-9cQGgRXeq7dTvteGHi10HaTG5zI7_eE8oW-BfrxYQw) TF and PyTorch code to pretrain Bert from scratch. I wrapped it in a script to launch on multiple machines on AWS [here](https://github.com/cybertronai/Megatron-LM/blob/master/launch_pretrain_bert.py). Currently I'm still figuring out why the 64-GPU AWS throughput is 2x worse than what they are getting locally", "Thanks @yaroslavvb!", "Thanks! @yaroslavvb", "@yaroslavvb [this article](https://medium.com/the-mission/why-building-your-own-deep-learning-computer-is-10x-cheaper-than-aws-b1c91b55ce8c) explains why cloud computing can have inconsistent throughput. I think it's a great read, and I've been working on setting up my own rig.\r\n\r\nI see in [the script](https://github.com/cybertronai/Megatron-LM/blob/master/launch_pretrain_bert.py#L49) that you're using 8 GPUs. have long is the pretraining taking with that? I'm not sure whether to go with gcloud TPUs or AWS. the Bert readme said that a single TPU will take up to 2 weeks to finish pretaining..", "@yaroslavvb hi, did you train bert successfully? I trained it with https://github.com/NVIDIA/Megatron-LM/scripts/pretrain_bert_tfrecords_distributed.sh on 2 machines with 16 GPUS, but when it was sotpped after ' > number of parameters: 336226108' and i got nothing else after that, the GPU-Util is 0%.", "@MarvinLong yes, I was able to launch it on multiple machines and observe the model training, and it's about 600ms per step. I did not try training it to completion as the scaling efficiency on p3dn instances on AWS is only about 50% because of NCCL bug currently. I'm wondering if your machines can't communicate to each other on the right ports. @jrc2139 I have not observed inconsistent throughput, I've used this [codebase](https://github.com/cybertronai/imagenet18) to train imagenet in 19 minutes on 64 GPUs on AWS p3 instances.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "> I've been able to use the codebase for this, and didn't see much issues, however I might be overlooking something. If you construct and initialize a new model instead of loading from pretrained, you can use the `simple_lm_finetuning` script to train on new data.\r\n> \r\n> Thomas, did you have any specific other issues in mind?\r\n\r\nI'm trying to train on my own custom data and I'm a bit confused about how to \"construct and initialize a new model\"—i.e., when not working with pretrained models. Any help appreciated.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "@yaroslavvb Hi, I can launch Megatron-LM to pretrain bert, but my MLM loss stay around 6.8. How about you? Can you pretrain BERT successfully?", "> @yaroslavvb Hi, I can launch Megatron-LM to pretrain bert, but my MLM loss stay around 6.8. How about you? Can you pretrain BERT successfully?\r\n\r\nI was able to pre-train using this repo [https://github.com/google-research/bert]. However, even with one million step, the MLM accuracy was 64.69% and it's loss was 2.4. I am eager to know if someone else has pre-trained and got MLM accuracy higher than this.", "> > @yaroslavvb Hi, I can launch Megatron-LM to pretrain bert, but my MLM loss stay around 6.8. How about you? Can you pretrain BERT successfully?\r\n> \r\n> I was able to pre-train using this repo [https://github.com/google-research/bert]. However, even with one million step, the MLM accuracy was 64.69% and it's loss was 2.4. I am eager to know if someone else has pre-trained and got MLM accuracy higher than this.\r\n\r\nAccording to the pretrian log from gloun-nlp[https://github.com/dmlc/web-data/blob/master/gluonnlp/logs/bert/bert_base_pretrain.log](url), your MLM accuracy seems right though with a higher loss. I think you can try to check it with fintuning. \r\n", "@ibrahimishag I want to know if you pretrain your BERT with Bookscorpus. I cannot find a copy of that. For my pretraining, my bert loss is decreasing so so slowly after removing clip-grad-norm. There must be something wrong with me.", "@JF-D I pre-trained on other domain-specific corpus.", "Can someone please specify why Thomas mention/refers XLM repo from facebook? Is there any fault from huggingface? I thought I would just use hugging face repo without using \"pretrained paramater\" they generously provided for us. \r\n\r\nJust struggling with Facebook repo\"span bert\" and seems it is hard to even run this due to distributed launch issue. Hope it is ok to use hugging face's one to reproduce paper result", "Is it possible to train from scratch using the run_language_modeling.py code? does hugging face support training from scratch. I looked at this example https://huggingface.co/blog/how-to-train but this thread is hitting that training from scratch is not currently supported.", "Any update on training from scratch BERT-like models with huggingface? ", "Yes this has been supported for close to a year now ;)", "@julien-c Thanks. I really appreacite the prompt response.\r\n\r\nIs there any tutorial/example specifically for BERT (/ALBERT) pretraining ?", "Pretraining from scratch is a very rigid demand for users.", "> @julien-c Thanks. I really appreacite the prompt response.\r\n> \r\n> Is there any tutorial/example specifically for BERT (/ALBERT) pretraining ?\r\n\r\nwait example", "This is all there is to pretraining:\r\n```\r\nimport os\r\n\r\nos.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\r\n\r\nfrom pathlib import Path\r\nfrom transformers import BertTokenizer\r\nfrom tokenizers.processors import BertProcessing\r\nfrom transformers import RobertaConfig\r\nfrom transformers import RobertaForMaskedLM\r\nfrom transformers import LineByLineTextDataset\r\nfrom transformers import DataCollatorForLanguageModeling\r\nfrom transformers import Trainer, TrainingArguments\r\nimport torch\r\n\r\ntokenizer = BertTokenizer('./data/vocab.txt')\r\n\r\ntokens = tokenizer.encode(\"b140 m33 c230\")\r\nprint('token ids: {}'.format(tokens))\r\n\r\nconfig = RobertaConfig(\r\n vocab_size=1458,\r\n max_position_embeddings=130,\r\n hidden_size=384,\r\n intermediate_size=1536,\r\n num_attention_heads=4,\r\n num_hidden_layers=4,\r\n type_vocab_size=1,\r\n)\r\n\r\n# FROM SCRATCH\r\nmodel = RobertaForMaskedLM(config=config)\r\n\r\n# CONTINUE TRAINING -- i.e., just load your saved model using \"from_pretrained\"\r\n# model = RobertaForMaskedLM.from_pretrained('./trained_model')\r\n\r\nprint(model.num_parameters())\r\n\r\n# We should save this dataset since it's a bit slow to build each time\r\ndataset = LineByLineTextDataset(\r\n tokenizer=tokenizer,\r\n file_path=\"./data/my_data.txt\",\r\n block_size=128,\r\n)\r\n\r\ndata_collator = DataCollatorForLanguageModeling(\r\n tokenizer=tokenizer, mlm=True, mlm_probability=0.15\r\n)\r\n\r\ntraining_args = TrainingArguments(\r\n output_dir=\"./out/my_run\",\r\n overwrite_output_dir=True,\r\n num_train_epochs=100,\r\n per_device_train_batch_size=128,\r\n save_steps=100,\r\n save_total_limit=2,\r\n)\r\n\r\ntrainer = Trainer(\r\n model=model,\r\n args=training_args,\r\n data_collator=data_collator,\r\n train_dataset=dataset,\r\n prediction_loss_only=True,\r\n)\r\n\r\ntrainer.train()\r\n\r\ntrainer.save_model(\"./trained_model\")\r\n```\r\n\r\nNote that this is a small model, with a specialized, fixed vocabulary, so I'm using the old BERT tokenizer I had working from a previous project. For \"real\" languages you'd use one of the RobertaTokenizer options.\r\n\r\nI'm just getting back to this project after being away for a while, and I'm noticing I'm getting a warning about switching to the Datasets Library. I'll do that at some point, but it's working for now so I won't mess with it.\r\nAlso, I'm curious if anyone can tell me how to set the maximum length of inputs, so that longer inputs truncate?\r\n\r\nUPDATE: Duh, sorry, looks like `tokenizer.encode()` takes `max_length` and `truncation` parameters. Simple.", "One question; I'm noticing that creating the dataset...\r\n```\r\ndataset = LineByLineTextDataset(\r\n tokenizer=tokenizer,\r\n file_path=\"./data/my_data.txt\",\r\n block_size=128,\r\n)\r\n```\r\n...is taking a long time. Is it possible to save that as a file, to avoid the wait when I (re)run training?", "Hi, Is there any specifications of how to generate dataset for \"pretraining from scratch\" with raw texts ?", "> One question; I'm noticing that creating the dataset...\r\n> \r\n> ```\r\n> dataset = LineByLineTextDataset(\r\n> tokenizer=tokenizer,\r\n> file_path=\"./data/my_data.txt\",\r\n> block_size=128,\r\n> )\r\n> ```\r\n> \r\n> ...is taking a long time. Is it possible to save that as a file, to avoid the wait when I (re)run training?\r\n\r\nthe same question", "Detailed Tutorial\r\nhttps://mlcom.github.io/Create-Language-Model/" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Wondering Noone is replying\n\nOn Sat, 25 May, 2019, 4:09 PM stale[bot], <[email protected]> wrote:\n\n> Closed #384\n> <https://github.com/huggingface/pytorch-pretrained-BERT/issues/384>.\n>\n> —\n> You are receiving this because you authored the thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/pytorch-pretrained-BERT/issues/384?email_source=notifications&email_token=AHRBKICW2VUPHXBBGSQD5O3PXEJN5A5CNFSM4G666UB2YY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGORUM5MZY#event-2367280743>,\n> or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/AHRBKIG2BUNIOBDUV6QGHQDPXEJN5ANCNFSM4G666UBQ>\n> .\n>\n" ]

[]

[]

[]

[]

[]

[ "Hi @andrewPoulton, yes indeed we could update that for GPT-2, would be happy to get a PR.\r\nCan you check the generations are identical for a few seeds (it should be)?", "Yeah, sure - what generations do you mean?", "Fixed with #495" ]

[]

[]

[ "Thanks @tseretelitornike!" ]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[]

[ "Good catch! Do you want to submit a PR? Otherwise, I'll fix it in the next release.", "Hi @thomwolf \r\nThe issue still persists, there were two extra indentations and you removed only one to move the line out of inner if-else but, one more indentation should be removed to bring [L620](https://github.com/huggingface/pytorch-pretrained-BERT/blob/master/examples/run_squad.py#L620) out of outer if-else. Doing this produces non-empty `nbest_predictions.json` file." ]

[]

[]

[ "Sorry, I just realized this was mentioned in the [original PR](https://github.com/huggingface/pytorch-pretrained-BERT/pull/124).", "Indeed. Happy to welcome a PR if you want to improve this example!", "Working on it now! One question, though: It seems likely that I'll have to make significant changes. The reason for this is there is a significant random element in the concatenation of sentences, and so the exact number of training examples is difficult to know in advance. The concept of an index into the list of training examples also stops making sense because of this. This makes a simple implementation based on a Dataset object difficult.\r\n\r\nI can see two possible approaches to resolve this:\r\n1) Before training begins, the script does a pass over the dataset and pregenerates all training cases as InputExample objects. These could also be regenerated each epoch to increase diversity. The training cases could be packed into a Dataset object and sampled with RandomSampler, so we could still have meaningful progress bars. This is similar to Google's original implementation, where the data is pregenerated and stored in example files.\r\n\r\n2) The random sampling could occur on the fly at train time. This would avoid the time and storage needed for pregenerating training cases, but would become harder to measure the 'length' of an epoch in advance. \r\n\r\nI think either of these could be different enough that it might be better to implement it as a separate script (though it would still use a lot of the helper functions from the existing script). What do you think?", "Hi @Rocketknight1, yes both solutions make sense and it seems better to have independent scripts indeed.\r\nMaybe you can start by drafting an independent script focusing on the pregenerated case and then see if the current `run_lm_finetuning` script can be updated for the on-the-fly case?", "I think starting with pregenerated makes sense. However, maybe instead of replacing the old script, maybe we can keep that one as is? The original authors mentioned that concatenating sentences into training examples didn't make sense for their use-case. Possibly their dataset was something like chatbot conversations instead of long documents?\r\n\r\nEither way, they (and probably others) have use for the simple \"one sentence per example\" training system so I don't want to delete it entirely!", "Sounds good to me!", "@thomwolf I created PR #392 which includes the new functionality. " ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "maybe you can take a look at the \"cache_dir\" argument. For the run_classifier.py file, it is located at line 498", "I tried with --cache_dir , giving the fine-tunings output directory as cache_dir. \r\nI added these 2 files to the directory: bert_config.json and vocab.txt from the original bert_basic_uncased\r\n(finetune out folder has finetuned model pytorch_model.bin file, which I am not sure if it used at all)\r\n\r\nIt gave exactly same accuracy (by 16 digits) as direct train/eval run_classifier.py directly with bert_basic_uncased. It would seem that with --cache_dir, it saves the original bert_base_uncased to that given --cache_dir. I am not sure if there is another difference. \r\n\r\n(I am runnin a 3-label classifier, for which I used the SST-2 from GLUE as basis, saved data in same format and added 3rd label to code in run_classifier.py\r\n\r\npython run_classifier.py \\\r\n --task_name SST-2 \\\r\n --do_train \\\r\n --do_eval \\\r\n --do_lower_case \\\r\n --data_dir ~/git/x/data/input-sst2/ \\\r\n --bert_model bert-base-uncased \\\r\n --cache_dir out_finetune_140/ \\\r\n --max_seq_length 140 \\\r\n --train_batch_size 16 \\\r\n --learning_rate 2e-5 \\\r\n --num_train_epochs 3.0 \\\r\n --output_dir out_testcache/\r\n\r\n\r\n\r\n", "I can get classifier running on finetuned model by replacing bert-base-uncased with output folder of fine-tuning: --bert_model out_finetune_140/ \\ (and adding bert_config.json , vocab.txt to that folder)\r\n\r\nBut as a result, eval_accuracy went down from 0.918 to 0.916. \r\n(I wonder is it correct to use vocab.txt and bert_config.json from original bert_base_uncased, or would fine-tuned model need updated ones?)\r\n\r\nstep1:\r\npython run_lm_finetuning.py \\\r\n --bert_model bert-base-uncased \\\r\n --do_lower_case \\\r\n --do_train \\\r\n --train_file ~/git/xdata/lm-file.txt \\\r\n --output_dir out_finetune_140/ \\\r\n --num_train_epochs 2.0 \\\r\n --learning_rate 3e-5 \\\r\n --train_batch_size 16 \\\r\n --max_seq_length 140 \\\r\n\r\nstep2:\r\npython run_classifier.py \\\r\n --task_name SST-2 \\\r\n --do_train \\\r\n --do_eval \\\r\n --do_lower_case \\\r\n --data_dir ~/git/x/data/input-sst2/ \\\r\n --bert_model out_finetune_140/ \\\r\n --max_seq_length 140 \\\r\n --train_batch_size 16 \\\r\n --learning_rate 2e-5 \\\r\n --num_train_epochs 3.0 \\\r\n --output_dir out_finetune+class_140/\r\n", "Hi, I think I find a workaround for this issue. You can first load the original model, and then insert this line into your python file (for example, after line 607 and 610 in run_classifier.py): \r\nmodel.load_state_dict(torch.load(\"output_dir/pytorch_model.bin\"))\r\nthen the model will be your customized fine-tuned model. And there is no need to change anything else (for example, the config file or vocab file)", "I would suggest that you add a separate logic to load your fine-tuned model and perform prediction. Your code will be very similar to the eval but you won't need (actually won't have access to) labels during prediction and hence no need of the accuracy code in eval etc. Simply collect your predictions in a list and write to a file called \"pred_results.txt\".\r\nI added some new flags (\"do_pred\" and \"model_path\"), modified the eval logic little bit to ignore labels, and wrote outputs to a file. Things are working for me. ", "Hi LeenaShekhar\r\n\r\nWould you mind showing the code you wrote to perform predictions of a trained model? ", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Just doing `from pytorch_pretrained_bert import BertTokenizer, BertModel` in your project doesn't work?\r\n\r\nThat's what they do in [AllenNLP](https://github.com/allenai/allennlp/blob/3f0953d19de3676ea82e642659fc96d90690e34d/allennlp/modules/token_embedders/bert_token_embedder.py#L14) or [flair](https://github.com/zalandoresearch/flair/blob/797c958d0e8c256531f2cea37508e7becb2026cb/flair/embeddings.py#L14)", "it does work, however, I don't need the higher level classes for my model, so I am sub classing the OpenAIGPTPreTrainedModel and Block and they are not included into the __init__.py. Including them into the __init__.py also solves the issue with the importException:\r\n\r\nfrom .modeling_openai import (OpenAIGPTConfig, OpenAIGPTModel,\r\nOpenAIGPTLMHeadModel, OpenAIGPTDoubleHeadsModel, OpenAIGPTPreTrainedModel, Block, load_tf_weights_in_openai_gpt)\r\n", "I see. I think we can fix this the same way I did in the `bert` case by adding `from __future__ import absolute_import`. If you don't mind I'll do a quick PR on that." ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" } ]

[]

[ "Similar problem. My token ids is something like \"[cls]qqqq[sep]0000cccccccc[sep]00000\". Have you solve it? or is there anyone met the similar problem?", "> Similar problem. My token ids is something like \"[cls]qqqq[sep]0000cccccccc[sep]00000\". Have you solve it? or is there anyone met the similar problem?\r\n\r\nI felt like this is caused by the way how we pretrain BERT. BERT is pretrained on contiguous texts. The way we pad zeros in between broke such continuity so essentially we need to re-train the whole model. \r\n\r\nSince I was trying to separate query and context, the way I tackled this is just masking all query tokens to create the context and same for the query. ", "> > Similar problem. My token ids is something like \"[cls]qqqq[sep]0000cccccccc[sep]00000\". Have you solve it? or is there anyone met the similar problem?\r\n> \r\n> I felt like this is caused by the way how we pretrain BERT. BERT is pretrained on contiguous texts. The way we pad zeros in between broke such continuity so essentially we need to re-train the whole model.\r\n> \r\n> Since I was trying to separate query and context, the way I tackled this is just masking all query tokens to create the context and same for the query.\r\n\r\nIt could be the same way as inputing the query and context separately if we mask the query or context part. Anyway, I will try it. Thanks." ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Why should it be limited to half of 512?", "> Why should it be limited to half of 512?\r\n\r\ncause when do train, we have sentence embedding 0 and 1, but in a single sentence classification task ,we just embedding 0, if this get bad influence", "You can just set the whole sequence to sentence 0. Create a DataProcessor class for your task and set the whole input sequence to `text_a`, example:\r\n\r\n```\r\nclass MyProcessor(DataProcessor):\r\n# some other methods here\r\n\r\n def _create_examples(self, lines, set_type):\r\n \"\"\"Creates examples for the training and dev sets.\"\"\"\r\n examples = []\r\n for (i, line) in enumerate(lines):\r\n guid = \"%s-%s\" % (set_type, i)\r\n text_a = line[1]\r\n label = line[0]\r\n examples.append(\r\n InputExample(guid=guid, text_a=text_a, text_b=None, label=label))\r\n return examples\r\n``` \r\nNotice `text_b=None`.", "How should I do if I have not only a sentence, but a whole text? \r\nI don't clearly understand how to extend `BertForSequenceClassification` with my own dataset for training/evaluating.\r\nI have a dataset consisting of text/label pairs, where text can have multiple sentences. ", "Just send in the whole text as one \"sentence\", the limit on a sequence length that can be sent at once to BERT is 512 tokens", "Ok, thanks.\r\nOne more question related to classification. BERT tokenizes my sentences pretty strange:\r\n> 04/27/2019 16:08:32 - INFO - __main__ - tokens: [CLS] @ bra ##yy ##yy ##ant Так акт ##иви ##ровала ##сь новая карта , ст ##ара ##я и была не ##ак ##тив ##на . [SEP]\r\n\r\nWhy are more than a half of the words a separated with `#`? I mean, these words are on russian and many of them are splited to several parts with `#`, though it is one word. Should this be fixed during training?", "That's the WordPiece tokenization, its a way to match subwords when an out of vocabulary word is encountered. It's explained in the bert paper with references. It's as it should be. ", "Ok, thank you so much.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "wouldn't this cause some kind of indentation error? (I don't have time to test the change sorry)" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Check Jacob Devlin slides starting from slide 26 [here](https://nlp.stanford.edu/seminar/details/jdevlin.pdf?fbclid=IwAR2TBFCJOeZ9cGhxB-z5cJJ17vHN4W25oWsjI8NqJoTEmlYIYEKG7oh4tlY)", "@thomwolf thanks, the slides were helpful. Do you know if there is a recording of the talk publicly available somewhere?", "I don't know!", "It was a very crowded room and these talks are generally not recorded, sorry…", "Is the synthetic self training module in this code?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Same question, will the synthetic self training module be in this code?", "Presently not in the library and there are no short-term plans to add synthetic self-training to the library." ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "If you want only the context you can find the index from the segment vector by finding the last first 1 in the vector and splitting the query_context on that index. Then the context will be everything after the 1 index and everything before will be the question.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1108649070, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDcw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Need%20more%20information", "name": "Need more information", "color": "d876e3", "default": false, "description": "Further information is requested" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "I had the same issue. I'm getting seg fault on an aws deep learning ami with a tesla v100 gpu instance. I get the error with or with out fp16", "I also have the same problem. Did you guys figure out any solution? \r\nI am able to load the data, however at the first epoch 0, I see the error segmentation fault. ", "Can you try running the model:\r\n- with a very small batch size to check if it's an OOM error\r\n- with `CUDA_LAUNCH_BLOCKING=1` to see the exact line causing the error\r\n?", "setting the env variable CUDA_LAUNCH_BLOCKING=1 doesn't give me any additional error messages. I still see the same error. \r\n \r\nTried it with batch size 1 and I get the same issue", "I got the code to run on my school server. I updated the gcc using this `conda install -c psi4 gcc-5`. ", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "I am also experiencing this issue. Anyone else figure it out?" ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" } ]

[]

[ "@Liangtaiwan and @abeljim were the contributors of the `run_squad.py` example.\r\nMaybe they can help.", "Hi @elephantomkk,\r\n\r\nThe non-answerable solving is using [CLS] token as the ground truth.\r\nAs a result, the start login = and end logit = -1.\r\n\r\nYou can find it out in this repo code or official Bert code.\r\nAlso, the method is mentioned in Jacob Devlin's slide.\r\nhttps://nlp.stanford.edu/seminar/details/jdevlin.pdf?fbclid=IwAR2TBFCJOeZ9cGhxB-z5cJJ17vHN4W25oWsjI8NqJoTEmlYIYEKG7oh4tlY", "Thanks for the reply! I tested that and found the performance on the non-answerable is not so good compared with the answerables:(" ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Indeed, this example could be improved. I would happy to welcome a PR on that.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1108649070, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDcw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Need%20more%20information", "name": "Need more information", "color": "d876e3", "default": false, "description": "Further information is requested" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Can you give a simple self-contained script to reproduce your issue?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "I agree with you, this part of the API could be improved.\r\n\r\nThe BERT model is now used in several third-party libraries like AllenNLP and FLAIR so we have to be careful not to make any breaking change on this model.\r\n\r\nWe could add a flag to get full output maybe.", "Mind if I submit a PR adding a flag that allows getting the full output in the case of `if labels is not None`?", "I'm ok to welcome a PR on that but there is one issue with flags in `forward()` call and multi-GPU you may or may not be aware of and we need to think about:\r\nAll the inputs to the `forward()` call are split across GPUs so having a non-batched input like a flag break DataParallel for multi-gpu.\r\nSo maybe we need to add a general flag in the models which can be set. Maybe you can try to draft a PR and we check it behave well on the examples then.\r\nMaybe you can also have a deep look at the way inputs are split in DataParallel and check whether a solution with flag in the arguments of the `forward()` call work.", "Hmm, okay, I was not aware of that. I think adding it as a property on the model itself is arguably an unnecessary over-complication for a simple matter. Returning a tuple value is much cleaner and the user can always discard the information that they don't care. Although I do understand the concern of backward compatibility. So I won't do this PR and will let you guys make the decision between the two options. Personally I lean more towards the API-breaking option.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Thanks!" ]

[]

[]

[ "Thanks @junjieqian!" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "I am also interested in this, it looks like we would have to append the prediction probability to the `all_predictions` JSON output.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Thanks Elon" ]

[]

[]

[ "Thanks @cdjhz " ]

[]

[]

[ "Seems good to me, thanks for that.\r\nLet me just check why we don't have Circle-CI tests on the PR anymore and I'll merge it." ]

[]

[]

[ "GPT is a causal model so each tokens only attend to the left context and masking is not really needed.\r\nJust mask the output according to your lengths (and be such that each input sample start at the very first left token)." ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[]

[ { "id": 1108649070, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDcw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Need%20more%20information", "name": "Need more information", "color": "d876e3", "default": false, "description": "Further information is requested" } ]

[]

[ "Strange error.\r\n\r\nCan you try:\r\n```python\r\nimport pytorch_pretrained_bert as ppb\r\nassert 'bert-large-cased' in ppb.modeling.PRETRAINED_MODEL_ARCHIVE_MAP\r\n```\r\nDo you have an open internet connection on the server that run the script?", "@thomwolf Is there a way to point to a model on disk? This question seems related enough to daisychain with this issue. :-)", "I noticed that this error happens when you exceed the disk space in the temporary directory while downloading BERT.", "I ran into the same problem. When I used the Chinese pre-training model, it was sometimes good and sometimes bad.", "@thomwolf I've been having the same error, and I received an AssertionError when I try \r\n\r\nassert 'bert-based-uncased' in bert.modeling.PRETRAINED_MODEL_ARCHIVE_MAP\r\n\r\nI've tried using both conda install and Pip install to get the package but in both cases I am not able to load any models", "Hi @DuncanCam-Stein,\r\nWhich version of python do you have?\r\nCan you try to install from source?", "@thomwolf @countback \r\nI finally fixed the problem by downloading the tf checkpoints directly from [here](https://github.com/google-research/bert), and then using the '[convert_tf_checkpoint_to_pytorch.py](https://github.com/huggingface/pytorch-pretrained-BERT/blob/master/pytorch_pretrained_bert/convert_tf_checkpoint_to_pytorch.py)' function to create a `pytorch_model.bin` file.\r\nI then used the path to pytorch_model.bin and bert_config.json file in BertModel.from_pretrained('path/to/bin/and/json') instead of 'bert-base-uncased'.\r\n👍 \r\nHelpful info was found [here](https://devhub.io/repos/huggingface-pytorch-pretrained-BERT).", "The network connection check has been relaxed in the now merged #500.\r\nSerialization of the model have also been simplified a lot with #489.\r\n\r\nThese improvements will be included in the next PyPI release (probably next week).\r\n\r\nIn the meantime you can install from `master` and already use the serialization best-practices described in the README [here](https://github.com/huggingface/pytorch-pretrained-BERT#serialization-best-practices)", "As @martiansideofthemoon said, I met this error because I didn't have enough space on disk.\r\n\r\nCheck if you can download the file with :\r\n\r\n`wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased.tar.gz`", "@martiansideofthemoon What does that mean if we can download it via wget but not when we use from_pretrained? is it a disk space problem?\r\n", "@Hannabrahman \r\nIf you can download it via wget, it means you have enough disk space, so the issue is from somewhere else.", "@Colanim Thanks. I figured out it was memory issue on the cache directory. ", "@Hannabrahman \r\n\r\n> @Colanim Thanks. I figured out it was memory issue on the cache directory.\r\n\r\nhow did you solve this issue?", "@raj5287 \r\nFree some disk space on the cache directory or specify another cache directory ", "@Colanim i have enough disk space since i have downloaded the file using \r\n`wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased.tar.gz` but i am not sure how to specify another cache directory or use the downloaded file (i am new to pytorch and ubuntu :| )", "> @thomwolf @countback\r\n> I finally fixed the problem by downloading the tf checkpoints directly from [here](https://github.com/google-research/bert), and then using the '[convert_tf_checkpoint_to_pytorch.py](https://github.com/huggingface/pytorch-pretrained-BERT/blob/master/pytorch_pretrained_bert/convert_tf_checkpoint_to_pytorch.py)' function to create a `pytorch_model.bin` file.\r\n> I then used the path to pytorch_model.bin and bert_config.json file in BertModel.from_pretrained('path/to/bin/and/json') instead of 'bert-base-uncased'.\r\n> +1\r\n> Helpful info was found [here](https://devhub.io/repos/huggingface-pytorch-pretrained-BERT).\r\n\r\n@DuncanCam-Stein i have downloaded and placed _pytorch_model.bin_ and _bert_config.json_ in _bert_tagger_ folder but when i am doing `tokenizer = BertModel.from_pretrained('home/user/Download/bert_pos_tagger/bert_tagger/')` i am still getting the error : `Model name 'home/user/Downloads/bert_pos_tagger/bert_tagger/' was not found in model name list (bert-base-uncased, bert-large-uncased, bert-base-cased, bert-large-cased, bert-base-multilingual-uncased, bert-base-multilingual-cased, bert-base-chinese). We assumed 'home/user/Downloads/bert_pos_tagger/bert_tagger/' was a path or url but couldn't find any file associated to this path or url.`", "try to delete cahe file and rerun the command", "I noticed that the error appears when I execute my script in debug mode (in Visual Studio Code). I fixed it by executing the script over the terminal `python myscriptname.py` once. Afterwards Debug mode works fine. \r\n\r\nBtw. I got the same problem with the tokenizer and this also fixed it.", "> > > > model = BertModel.from_pretrained('bert-large-cased')\r\n> > > > Model name 'bert-large-cased' was not found in model name list (bert-base-uncased, bert-large-uncased, bert-base-cased, bert-large-cased, bert-base-multilingual-uncased, bert-base-multilingual-cased, bert-base-chinese). We assumed 'https://s3.amazonaws.com/models.huggingface.co/bert/bert-large-cased.tar.gz' was a path or url but couldn't find any file associated to this path or url.\r\n\r\nhello，I meet the problem when run the torch bert code 👍 \r\n\r\nOSError: Can't load weights for 'bert-base-uncased'. Make sure that:\r\n\r\n- 'bert-base-uncased' is a correct model identifier listed on 'https://huggingface.co/models'\r\n\r\n- or 'bert-base-uncased' is the correct path to a directory containing a file named one of pytorch_model.bin, tf_model.h5, model.ckpt.\r\nif I can download the bert-base-uncased weight, where I should put the file in ? hope your reply~", "@DTW1004 check your network connection. This happens when I'm behind a proxy and SSL/proxy isn't configured appropriately.", "bro，I've been having the same error. and then I try to debug the specific code\r\nBertTokenizer.from_pretrained(MODEL_NAME)\r\nstep in to the origin code, I find that I could step every line of the transformer in the debug mode. when step out origin code, the tokenizer tool could be used. what‘s more , the code could be run normaly in the next time I run the code. ", "I met the issue and I found the reason is that my server connecting was offline.", "Running into the same issue on AWS Lambda. Neither relative and absolute paths will allow the model to load from pre-trained. ", "Here's what I am doing:\r\n\r\n```shell\r\n!wget -q https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-multilingual-cased.tar.gz\r\n!tar xf bert-base-multilingual-cased.tar.gz\r\n```\r\n\r\nNow, if I do:\r\n\r\n```python\r\nencoder = TFBertModel.from_pretrained(\"bert-base-multilingual-cased\")\r\n```\r\n\r\nI still get:\r\n\r\n```shell\r\nOSError: Can't load config for 'bert-base-multilingual-cased'. Make sure that:\r\n\r\n- 'bert-base-multilingual-cased' is a correct model identifier listed on 'https://huggingface.co/models'\r\n\r\n- or 'bert-base-multilingual-cased' is the correct path to a directory containing a config.json file\r\n```", "Here's what I am doing:\r\n\r\nfrom transformers import pipeline\r\n\r\ndef corret_sentence(sentence,unmasker):\r\n res = unmasker(sentence)\r\n return res \r\n\r\nif __name__=='__main__':\r\n sentence = \"关小彤\"\r\n new_sentence = \"\"\r\n unmasker = pipeline('fill-mask', model='uer/chinese_roberta_L-2_H-512') \r\n for idx,ch in enumerate(sentence): \r\n new_sentence = sentence[:idx] + \"[MASK]\" + sentence[idx+1:]\r\n print(corret_sentence(new_sentence,unmasker))\r\n\r\nI get:\r\n\r\nValueError: Could not load model uer/chinese_roberta_L-2_H-512 with any of the following classes: (<class 'transformers.models.auto.modeling_auto.AutoModelForMaskedLM'>, <class 'transformers.models.bert.modeling_bert.BertForMaskedLM'>).", "> Here's what I am doing:\r\n> \r\n> from transformers import pipeline\r\n> \r\n> def corret_sentence(sentence,unmasker): res = unmasker(sentence) return res\r\n> \r\n> if **name**=='**main**': sentence = \"关小彤\" new_sentence = \"\" unmasker = pipeline('fill-mask', model='uer/chinese_roberta_L-2_H-512') for idx,ch in enumerate(sentence): new_sentence = sentence[:idx] + \"[MASK]\" + sentence[idx+1:] print(corret_sentence(new_sentence,unmasker))\r\n> \r\n> I get:\r\n> \r\n> ValueError: Could not load model uer/chinese_roberta_L-2_H-512 with any of the following classes: (<class 'transformers.models.auto.modeling_auto.AutoModelForMaskedLM'>, <class 'transformers.models.bert.modeling_bert.BertForMaskedLM'>).\r\n\r\nhow could solve this?\r\n\r\n\r\ndid you solve this problem?\r\n\r\ni am also having sample", "> Free some disk space\r\n\r\nhow can I free some disk space.\r\nwhich shell command should i use？", "> @thomwolf @countback I finally fixed the problem by downloading the tf checkpoints directly from [here](https://github.com/google-research/bert), and then using the '[convert_tf_checkpoint_to_pytorch.py](https://github.com/huggingface/pytorch-pretrained-BERT/blob/master/pytorch_pretrained_bert/convert_tf_checkpoint_to_pytorch.py)' function to create a `pytorch_model.bin` file. I then used the path to pytorch_model.bin and bert_config.json file in BertModel.from_pretrained('path/to/bin/and/json') instead of 'bert-base-uncased'. 👍 Helpful info was found [here](https://devhub.io/repos/huggingface-pytorch-pretrained-BERT).\r\n\r\nCan you please specify which model exactly you downloaded and how you ran the function? Thanks" ]

[ { "id": 1260952223, "node_id": "MDU6TGFiZWwxMjYwOTUyMjIz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Discussion", "name": "Discussion", "color": "22870e", "default": false, "description": "Discussion on a topic (keep it focused or open a new issue though)" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "I think my answer here can help you: https://github.com/huggingface/pytorch-pretrained-BERT/issues/332", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Hey @shuvadibp, did you figure out a way of doing it? I would like to talk to you about the same.." ]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1108649070, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDcw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Need%20more%20information", "name": "Need more information", "color": "d876e3", "default": false, "description": "Further information is requested" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[]

[ { "id": 1108649070, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDcw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Need%20more%20information", "name": "Need more information", "color": "d876e3", "default": false, "description": "Further information is requested" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "Probably an issue with `t_total` and the number of training optimization steps similarly to #329.\r\nCould you check the number of total training step sent to the optimizer? Which example script are you using?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Wrong upstream I guess. Closing." ]

[]

[]

[ "Thanks @jplehmann!" ]

[]

[]

[ "Argh, just realized I was on `0.3.0` which is what pip installed due to some dependencies. Upgrading to `0.6.1` and now I'm getting expected scores:\r\n```\r\neval_accuracy = 0.8529411764705882\r\neval_loss = 0.39120761538837473\r\nglobal_step = 345\r\nloss = 0.17308216924252717\r\n\r\neval_accuracy = 0.8431372549019608\r\neval_loss = 0.49456917187746835\r\nglobal_step = 345\r\nloss = 0.12193756103515625\r\n\r\neval_accuracy = 0.875\r\neval_loss = 0.4023934503396352\r\nglobal_step = 345\r\nloss = 0.14832657523777174\r\n\r\neval_accuracy = 0.8553921568627451\r\neval_loss = 0.44353585865567713\r\nglobal_step = 345\r\nloss = 0.17814078952955162\r\n```" ]

[]

[]

[ "how do you fix this issue?", "> how do you fix this issue?\r\n\r\ntry to update your torch version，i found it didn't work in torch 4.0.0, try \"torch >=4.0.1\"" ]

[]

[]

[ "Could be, do you want to submit a PR to update this?", "Closed by #358, thanks @cdjhz!" ]