Datasets:

---

DanielPFlorian

/

Transformers-Github-Issues

like 0

[]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "You'd want to look at the `run_language_modeling.py` script which implements causal language modeling. (do not pass the `--mlm` flag)", "I'm thinking some edit to run_language_modeling.py script maybe would make it work. I don't think just to not pass the --mlm flag you solve the problem @julien-c. Have you found any solution @johncwok? I'm searching the same thing.", "@johncwok GPT2 always uses a causal mask. It's quite hidden in the code. This line https://github.com/huggingface/transformers/blob/0a4b1068e1d6c46525082b91a4ba00a09c9270ac/src/transformers/modeling_gpt2.py#L145 creates the causal mask that is then applied to the weights. The naming can definitely be improved here! So no matter what mask you insert it will only be applied in combination with the causal mask.\r\nAlso take a look at this line that creates the mask:\r\nhttps://github.com/huggingface/transformers/blob/0a4b1068e1d6c46525082b91a4ba00a09c9270ac/src/transformers/modeling_gpt2.py#L107\r\n", "After https://github.com/huggingface/transformers/pull/2715/files is merged, I will do some renaming in the code - seems like a lot of people look for the causal mask in GPT2, CTRL and GPT" ]

[]

[]

[]

[]

[]

[ "You should have used the template, because now we don't have enough information to help you: how are you running the script (torch launch utility? Which command?), which script are you using (your own (give details) or one of the example scripts)?\r\n\r\nBy default, PyTorch will only use one GPU unless you specify it to go DDP.", "Use tags please. Read through this guide. https://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks", "@BramVanroy it was my bad, I am in bit hurry so that I was not able to provide my code path. I have just pushed my code to github. Please check below for path \r\n[https://github.com/tiru1930/bert_intent_classification](url)\r\n\r\nIn this PATH I use src/Train.py to train the model.\r\n\r\n", "As you see, you are just wasting your own time and mine when you are _in a hurry_. In the future, take your time to write a good starting posts so that we _want_ to help you and _can_ help you quickly.\r\n\r\nIn your code, you are calling DataParallel on your mode, which will automatically run your model over multiple GPU's (but under a single process). Remove this line.\r\n\r\nhttps://github.com/tiru1930/bert_intent_classification/blob/master/src/train.py#L80" ]

[]

[]

[ "Indeed, this is not supported right now. We'd welcome a PR though :)", "Before somebody starts on a PR, we need to consider what exactly this should do.\r\n\r\nFor `top_k = 1`, most users probably expect a single forward pass and picking the top prediction for each token. For greater `top_k`, however, picking the k-best prediction at each mask position has increasingly high risk of yielding an inconsistent sequence. A beam search over all possible sequences with some overall objective and returning the overall `top_k` best sequences will be more desirable, but also more work to implement.\r\n\r\nA naive objective could simply multiply the probabilities of each candidate replacement obtained from a single forward pass. However, these probabilities are not conditional on the specific choice for the other mask positions. What exactly these probabilities are when there is more than 1 mask token is not clear to me but I think a reasonable assumption is that the network produces some kind of weighted average of all the probability distributions one would get if one fixes the other mask tokens and makes a forward pass with just one mask token.\r\n\r\nTherefore, I think one must make multiple forward passes to get the probability of each decision step in the gap filling process. It is not clear though in what order to make decisions. Even in the simplest case of continuous mask positions we could proceed left-to-right, right-to-left, from both sides simultaneously, start in the middle or in some other way. The order could also be influenced by the probabilities, e.g. condensating the most confidently predicted token first.\r\n\r\nIt may also be desirable to have a [MASK*] that is expanded to multiple tokens as needed. Then, one may want to have a brevity penalty or normalise by length as otherwise the model will prefer short answers as their probability is higher. One may also want to have a callback to filter candidate substitutions, e.g. for a cloze test one may want to check that the sequence does not start with '##' and that it detokenises to a single word of the target language.", "Please see the following issue https://github.com/huggingface/transformers/issues/10158 and PR https://github.com/huggingface/transformers/pull/10222 for an attempt to take a crack at this", "@jowagner Has made some very valid points. In fact, these are the same concerns I have had previously with how multiple mask filling even works when done simultaneously. However, there are some issues with all of the approaches and I am not quite sure yet as to how it could be resolved.\r\n\r\nTake for example you have 3 mask positions and we follow the method that gives preference first to the most confidently predicted token. There is an intrinsic issue as to what the most confident token would even mean here in the first place given that the other 2 masks are still empty and not filled. My point being, the probability of which word needs to be filled in a particular slot is not necessarily indicative of whether that SHOULD be the first one to be filled. \r\n\r\nDo have a look at https://arxiv.org/abs/2002.03079 's work on Blank Language Model. Most of the valuable suggestions that you provided here start spilling into this paper's realm. \r\n\r\nI would be very happy to discuss further about this with you Joachim", "Hi, I've implemented right to left, left to right, and random mask filling in PyTorch for top k ids that the model thinks are the most probable tokens in a sentence in one of my projects. In this implementation, each time we want to generate a mask, the model looks at the previously generated sentences and decides what is the most probable for the next masked position. So if we have 2 masks in a sentence, by setting top_k=5, we'll have 25 sentences (5 tokens for the first position, and for each of these 5 sentences with one mask we have another 5 tokens for the second mask). It'll output something like this:(I used Persian models for this. I hope you can see how the masks are being filled)\r\n\r\nThen in the next step, we implemented a beam search to choose the most probable sequence of all between all these sentences.\r\n\r\nI'd be glad to help HuggingFace on this issue, I can send my code or send a pull request.\r\n\r\n", "The idea in https://github.com/huggingface/transformers/pull/10222/commits/80a113641a49c73f7680289219096ee5cf7ca620#r605659735 may point to how one can combine left and right direction or even average over all possible sequences of crystallisation.", "Hi, This is the function for different orders of prediction. I hope it helps. \r\nAlso, In the beam search section, we constructed a dictionary of bi tri and four grams in a specific corpus related to our work and scored predictions based on those. I won't include this extensive part here but tell me if it can be useful.\r\n\r\n```\r\ndef predict_seqs_dict(sequence, model, tokenizer, top_k=5, order='right-to-left'):\r\n\r\n\r\n ids_main = tokenizer.encode(sequence,\r\n return_tensors=\"pt\",\r\n add_special_tokens=False)\r\n\r\n ids_ = ids_main.detach().clone()\r\n position = torch.where(ids_main == tokenizer.mask_token_id)\r\n\r\n positions_list = position[1].numpy().tolist()\r\n\r\n if order =='left-to-right':\r\n positions_list.reverse()\r\n\r\n elif order=='random':\r\n random.shuffle(positions_list)\r\n\r\n # print(positions_list)\r\n predictions_ids = {}\r\n predictions_detokenized_sents = {}\r\n\r\n for i in range(len(positions_list)):\r\n predictions_ids[i] = []\r\n predictions_detokenized_sents[i] = []\r\n\r\n \r\n # if it was the first prediction, \r\n # just go on and predict the first predictions\r\n \r\n\r\n if i==0:\r\n model_logits = model(ids_main)['logits'][0][positions_list[0]]\r\n top_k_tokens = torch.topk(model_logits, top_k, dim=0).indices.tolist()\r\n \r\n for j in range(len(top_k_tokens)):\r\n #print(j)\r\n ids_t_ = ids_.detach().clone()\r\n ids_t_[0][positions_list[0]] = top_k_tokens[j]\r\n predictions_ids[i].append(ids_t_)\r\n \r\n pred = tokenizer.decode(ids_t_[0])\r\n predictions_detokenized_sents[i].append(pred)\r\n\r\n # append the sentences and ids of this masked token\r\n\r\n\r\n # if we already have some predictions, go on and fill the rest of the masks\r\n # by continuing the previous predictions\r\n if i!=0:\r\n for pred_ids in predictions_ids[i-1]:\r\n \r\n # get the logits\r\n model_logits = model(pred_ids)['logits'][0][positions_list[i]]\r\n # get the top 5 of this prediction and masked token\r\n top_k_tokens = torch.topk(model_logits, top_k, dim=0).indices.tolist()\r\n\r\n for top_id in top_k_tokens:\r\n \r\n ids_t_i = pred_ids.detach().clone()\r\n ids_t_i[0][positions_list[i]] = top_id\r\n\r\n pred = tokenizer.decode(ids_t_i[0])\r\n\r\n # append the sentences and ids of this masked token\r\n\r\n predictions_ids[i].append(ids_t_i)\r\n predictions_detokenized_sents[i].append(pred)\r\n \r\n return predictions_detokenized_sents\r\n \r\n```\r\n\r\n", "This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.\n\nPlease note that issues that do not follow the [contributing guidelines](https://github.com/huggingface/transformers/blob/master/CONTRIBUTING.md) are likely to be ignored.", "While an external scoring model may produce higher quality results, such an approach would move quite far away from letting the BERT model make the predictions. For example, consider a users who is evaluating the quality of a BERT model using a cloze test. They don't want issues of the BERT model to be smoothed / repaired by the external scoring model.\r\n\r\nFor finding the most confidently predicted token, I don't see why the fact that 3 or more masks may include a mask that has only masked neighbours is a problem. What we need is a measure of confidence that can be derived from the class probability distribution of the MLM head (its softmax layer). BERT gives us a class probability distribution for each masked token. The most confident token is then simply the one for which the confidence measure gives the greatest value.\r\n\r\nI didn't yet find time to read https://arxiv.org/abs/2002.03079 ", "@jowagner Just to reconfirm, your proposition was to fill the slots not in an arbitrary left to right or right to left fashion, but to fill the one that has the highest value in the softmax layer and then utilize that while regenerating clozes for the rest of the masks, correct?\r\n\r\n The high confidence for the position could be by virtue of there not being any other better suitable candidates for that position rather than being an indicator that the model is most confident about that prediction (for us to be filling that prediction first before using that as the seed to move on and fill the rest in a similar fashion). Right? " ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834056635, "node_id": "MDU6TGFiZWwxODM0MDU2NjM1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Tokenization", "name": "Core: Tokenization", "color": "FF4446", "default": false, "description": "Internals of the library; Tokenization." } ]

[]

[ "A PR is welcome!", "created PR #3621 ", "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", "cc @mfuntowicz @n1t0 \r\n\r\nThe problem is that the tokenizer does not allow empty strings to be passed (will lead to index out of bounds error when `text[0].isspace()`). An empty string is possible, according to OP, when using the QQP task which has such a format. OP added a PR here that you can have a look at https://github.com/huggingface/transformers/pull/3621", "This issue speaks about fixing https://github.com/huggingface/transformers/blob/81484b447b7d8504ff5e1cfff38ec35918383963/src/transformers/tokenization_roberta.py#L239 which seems totally reasonable to me, but #3621 does a lot more than that, half of which I don't even understand.\r\n\r\n@boy2000-007man could you update your PR to only fix the relevant line, and maybe add a test?", "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" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3607?src=pr&el=h1) Report\n> Merging [#3607](https://codecov.io/gh/huggingface/transformers/pull/3607?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/81484b447b7d8504ff5e1cfff38ec35918383963&el=desc) will **decrease** coverage by `1.04%`.\n> The diff coverage is `43.33%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3607?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3607 +/- ##\n==========================================\n- Coverage 78.06% 77.02% -1.05% \n==========================================\n Files 100 100 \n Lines 17134 17159 +25 \n==========================================\n- Hits 13375 13216 -159 \n- Misses 3759 3943 +184 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3607?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/pipelines.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9waXBlbGluZXMucHk=) | `72.76% <43.33%> (-2.19%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0.00%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.71% <0.00%> (-10.00%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `95.61% <0.00%> (-2.64%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `73.48% <0.00%> (-2.30%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.20% <0.00%> (-1.35%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl91dGlscy5weQ==) | `93.28% <0.00%> (-0.33%)` | :arrow_down: |\n| [src/transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3607/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==) | `92.10% <0.00%> (+0.12%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3607?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3607?src=pr&el=footer). Last update [81484b4...34623f3](https://codecov.io/gh/huggingface/transformers/pull/3607?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Thank you @enzoampil!", "FYI that I will apply the entity grouping functionality explained above in this same PR", "**This pull request now applies the entity group transformation illustrated above by setting the parameter: `group`=True**.\r\n\r\nThis was done by reflecting the transformation inside `NerPipeline`. I've changed the name of the pull request to better reflect the feature being proposed.\r\n\r\ncc @julien-c @mfuntowicz @petulla \r\n\r\nSample code:\r\n```\r\n# Install branch\r\n# Make sure to restart runtime after installing if using Google Colab\r\n!pip install -e git+git://github.com/enzoampil/transformers.git@add_index_to_ner_pipeline#egg=transformers\r\n\r\n# Grouped NER\r\nfrom transformers import pipeline\r\nnlp = pipeline('ner', group=True)\r\nnlp(\"Enzo works at the Australian National University (AUN)\")\r\n\r\n# [{'entity_group': 'I-PER', 'score': 0.9968132972717285, 'word': 'Enzo'},\r\n# {'entity_group': 'I-ORG', 'score': 0.9970400333404541, 'word': 'Australian National University'},\r\n# {'entity_group': 'I-ORG', 'score': 0.9831967651844025, 'word': 'AUN'}]\r\n\r\n# Ungrouped NER\r\nnlp = pipeline('ner', group=False)\r\nnlp(\"Enzo works at the Australian National University (AUN)\")\r\n\r\n# [{'entity': 'I-PER', 'index': 1, 'score': 0.9983270168304443, 'word': 'En'},\r\n# {'entity': 'I-PER', 'index': 2, 'score': 0.9952995777130127, 'word': '##zo'},\r\n# {'entity': 'I-ORG', 'index': 6, 'score': 0.9984350204467773, 'word': 'Australian'},\r\n# {'entity': 'I-ORG','index': 7, 'score': 0.9967807531356812, 'word': 'National'},\r\n# {'entity': 'I-ORG', 'index': 8 'score': 0.9959043264389038, 'word': 'University'},\r\n# {'entity': 'I-ORG', 'index': 10, 'score': 0.9900023937225342, 'word': 'AU'},\r\n# {'entity': 'I-ORG', 'index': 11, 'score': 0.9763911366462708, 'word': '##N'}]\r\n```\r\n\r\nTutorial on how to do Entity Grouping w/ `NerPipeline` [here](https://colab.research.google.com/drive/1CVLP0n3Q5t5qiWpode7jyhUNZpmLg0mS)\r\n\r\nI'm very keen to get feedback for the above, so please let me know if I should change anything, or perform additional steps to bring its quality to an acceptable level.", "I accidentally deleted the fork for this, so I've recreated this pull request [here](https://github.com/huggingface/transformers/pull/3957). Apologies for any inconvenience caused by this.\r\n\r\nI will close this PR so please refer to the one linked above." ]

[]

[]

[]

[ { "id": 1771187924, "node_id": "MDU6TGFiZWwxNzcxMTg3OTI0", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Pipeline", "name": "Core: Pipeline", "color": "FF7066", "default": false, "description": "Internals of the library; Pipeline." } ]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Hi @Colanim, thanks a lot for your speed comparison :-). \r\n\r\nIt might be possible that the pipelines used different default parameters for `T5` and `Bart` under the hood which strongly influence their running times.\r\nBesides `min_length` and `max_length` could you also insert those parameters into both `T5` and `Bart` to overwrite the default parameters: \r\n\r\n```\r\n \"early_stopping\": True\r\n \"length_penalty\": 2.0\r\n \"no_repeat_ngram_size\": 3\r\n \"num_beams\": 4\r\n```\r\n\r\nIf there is still a big difference in time, then I guess we have to take a closer look! \r\n\r\n", "Thanks for your fast answer @patrickvonplaten \r\n\r\nHere is the link to the modified notebook, with the parameters you mentioned :\r\nhttps://colab.research.google.com/drive/1kCm5ew8qDQqguZjbsC6Ujs9KZBaSfafi\r\n\r\n---\r\n\r\nUnfortunately, there is still a **huge** difference...\r\n\r\n```\r\nBART = 66s\r\nT5 = 226s\r\n```", "Ok, good to know! thanks for doing the comparison @Colanim. This might interest you as well @sshleifer :-) \r\n\r\nOh actually I just remember that Bart caches the decoder hidden key/value outputs when doing auto-regressive decoding (similar to GPT2 - check Visuals under \"GPT-2 Masked Self-Attention\" in this [post](http://jalammar.github.io/illustrated-gpt2/)) and I think T5 does not. \r\n\r\nBut T5 could cache the decoder key/value outputs to speed up decoding as well since it uses a causal mask for the decoder. This could definitely be a Feature Request. What do you think\r\n@sshleifer @craffel @thomwolf ?", "Sounds worth it!" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3604?src=pr&el=h1) Report\n> Merging [#3604](https://codecov.io/gh/huggingface/transformers/pull/3604?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/81484b447b7d8504ff5e1cfff38ec35918383963&el=desc) will **not change** coverage by `%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3604?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3604 +/- ##\n=======================================\n Coverage 78.06% 78.06% \n=======================================\n Files 100 100 \n Lines 17134 17134 \n=======================================\n Hits 13375 13375 \n Misses 3759 3759 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3604?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3604/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl91dGlscy5weQ==) | `93.45% <0.00%> (-0.17%)` | :arrow_down: |\n| [src/transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3604/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==) | `92.10% <0.00%> (+0.12%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3604?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3604?src=pr&el=footer). Last update [81484b4...8845212](https://codecov.io/gh/huggingface/transformers/pull/3604?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3603?src=pr&el=h1) Report\n> Merging [#3603](https://codecov.io/gh/huggingface/transformers/pull/3603?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/81484b447b7d8504ff5e1cfff38ec35918383963&el=desc) will **not change** coverage by `%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3603?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3603 +/- ##\n=======================================\n Coverage 78.06% 78.06% \n=======================================\n Files 100 100 \n Lines 17134 17134 \n=======================================\n Hits 13375 13375 \n Misses 3759 3759 \n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3603?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3603?src=pr&el=footer). Last update [81484b4...32340ca](https://codecov.io/gh/huggingface/transformers/pull/3603?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3602?src=pr&el=h1) Report\n> Merging [#3602](https://codecov.io/gh/huggingface/transformers/pull/3602?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/1a16d9d94a81e95463b166adfce4a8e02cdc47eb&el=desc) will **not change** coverage by `%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3602?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3602 +/- ##\n=======================================\n Coverage 78.06% 78.06% \n=======================================\n Files 100 100 \n Lines 17181 17181 \n=======================================\n Hits 13413 13413 \n Misses 3768 3768 \n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3602?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3602?src=pr&el=footer). Last update [1a16d9d...1a16d9d](https://codecov.io/gh/huggingface/transformers/pull/3602?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Hi, is there any reason that cc-25 was removed and only the fine-tuned one is kept? any way I can quickly enable that? thanks", "When the authors released the CC25 checkpoint, it was shaped differently than `mbart-large-en-ro` and I am not clear on whether that is fixed yet.\r\nSee https://github.com/pytorch/fairseq/issues?q=is%3Aissue+is%3Aopen+sort%3Aupdated-desc+mbart" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3601?src=pr&el=h1) Report\n> Merging [#3601](https://codecov.io/gh/huggingface/transformers/pull/3601?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/f68d22850ced09bb194b30068ff94ca3409f0879&el=desc) will **decrease** coverage by `0.00%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3601?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3601 +/- ##\n==========================================\n- Coverage 78.06% 78.05% -0.01% \n==========================================\n Files 100 100 \n Lines 17134 17134 \n==========================================\n- Hits 13375 13374 -1 \n- Misses 3759 3760 +1 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3601?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3601/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl91dGlscy5weQ==) | `93.28% <0.00%> (-0.17%)` | :arrow_down: |\n| [src/transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3601/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==) | `92.10% <0.00%> (ø)` | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3601?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3601?src=pr&el=footer). Last update [f68d228...857e77e](https://codecov.io/gh/huggingface/transformers/pull/3601?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[ { "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" ]

[]

[]

[ "My apologies, my computer glitched and posted twice. Please close this issue and refer to: https://github.com/huggingface/transformers/issues/3600" ]

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

[]

[ "Please use the template in the future. It is there for a reason. As mentioned in the template, don't post a screenshot. Use code blocks, post your code or the example script that you used, and the error trace. Also provide your version of PyTorch and Python.\r\n\r\nhttps://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks", "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 }, { "id": 1834059054, "node_id": "MDU6TGFiZWwxODM0MDU5MDU0", "url": "https://api.github.com/repos/huggingface/transformers/labels/Ex:%20Generation", "name": "Ex: Generation", "color": "06EFF8", "default": false, "description": "Natural Language Generation" } ]

[]

[ "CTRL uses control codes, as is mentioned in our documentation, with examples on the [original repository](https://github.com/salesforce/ctrl#generations). Have you tried using these control codes?", "How do I specify which control code I want to use? Do I have to do that in the command line and if yes, how? This is the Control code I want to use 16360 (i.e. politics). \r\n\r\nThank you ", "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 }, { "id": 1834056635, "node_id": "MDU6TGFiZWwxODM0MDU2NjM1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Tokenization", "name": "Core: Tokenization", "color": "FF4446", "default": false, "description": "Internals of the library; Tokenization." } ]

[]

[ "Hi,\r\n\r\nIs anyone working on this or is it open for someone to take?\r\nI was able to reproduce the issue.\r\nIf not being worked on by anyone, I would like to take it up.\r\n\r\nThanks", "I'm facing similar issues with batch_encode_plus:\r\n```\r\ntokenizer = transformers.BertTokenizer.from_pretrained('bert-base-cased')\r\na = ['short sentence', 'Larger sentence than short sentence']\r\ninput_ids = torch.tensor(tokenizer.batch_encode_plus(a, pad_to_max_length=True)['input_ids'])\r\n```\r\nIt doesn't work for me, it return this error:\r\n`ValueError: expected sequence of length 2 at dim 1 (got 6)`\r\n\r\nThanks", "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", "This should now be fixed on master with the updated tokenizer API" ]

[]

[]

[]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834056635, "node_id": "MDU6TGFiZWwxODM0MDU2NjM1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Tokenization", "name": "Core: Tokenization", "color": "FF4446", "default": false, "description": "Internals of the library; Tokenization." } ]

[]

[ "Can you show how you initialize `tokenizer`? Which vocab are you using?", "> Can you show how you initialize `tokenizer`? Which vocab are you using?\r\n\r\nsorry I forgot that... I updated the code in the issue already.\r\n\r\nI was using `distilbert-base-multilingual-cased`\r\n\r\n`tokenizer = DistilBertTokenizer.from_pretrained(\"distilbert-base-multilingual-cased\")`\r\n\r\n", "This behaviour seems to have been solved in v2.7.0 as running your code yields the correct result in my environment.", "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 } ]

[]

[ "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", "Any idea about how to solve this apostrophe seperating word token with different word id in bert rokenizer ?" ]

[]

[]

[ "I believe only BART and T5 can do summarization for now. See the [documentation regarding the checkpoints that may summarize](https://huggingface.co/transformers/main_classes/pipelines.html#summarizationpipeline).", "Here is [a notebook](https://github.com/Nikoschenk/bert-extractive-summarizer/blob/master/colab/scibert-summaries.ipynb) using the scibert model based on the [great code](https://github.com/dmmiller612/bert-extractive-summarizer) that Derek provided.\r\n\r\n" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1843377584, "node_id": "MDU6TGFiZWwxODQzMzc3NTg0", "url": "https://api.github.com/repos/huggingface/transformers/labels/Version%20mismatch", "name": "Version mismatch", "color": "ddea7c", "default": false, "description": "" } ]

[]

[ "It seems you have a very old version of this repository (when it was still named `pytorch_transformers`). Please update to the latest version or you won't have all the features (like access to the model hub).", "I fixed the version and the error was fixed. But now I get \r\n\r\n AttributeError: 'BertTokenizer' object has no attribute 'encoder'\r\nDo I have another problem with version? I also used \r\n\r\n pip3 install --upgrade pytorch-pretrained-bert\r\nThank you", "Please post the code and full trace that gave you this error.", " tokenizer = AutoTokenizer.from_pretrained(\"nlpaueb/bert-base-greek-uncased-v1\")\r\n model = AutoModelWithLMHead.from_pretrained(\"nlpaueb/bert-base-greek-uncased-v1\")\r\n model.to(args.device)\r\n # Add special tokens if they are not already added\r\n add_special_tokens_(model, tokenizer)\r\n\r\n def add_special_tokens_(model, tokenizer):\r\n \"\"\" Add special tokens to the tokenizer and the model if they have not already been added. \"\"\"\r\n orig_num_tokens = len(tokenizer.encoder)\r\n num_added_tokens = tokenizer.add_special_tokens(ATTR_TO_SPECIAL_TOKEN) # doesn't add if they are already there\r\n if num_added_tokens > 0:\r\n model.resize_token_embeddings(new_num_tokens=orig_num_tokens + num_added_tokens)\r\n\r\n\r\n\r\n INFO:transformers.modeling_utils:loading weights file https://s3.amazonaws.com /models.huggingface.co/bert/nlpaueb/bert-base-greek-uncased-v1/pytorch_model.bin from cache at /home/hatzimin/.cache/torch/transformers/3a685f5fa6f50a35a4efc31e9cdc74cfe8e2956002ee5c2df350e5e6c54deaf2.2aad66b9b70b2aa069cb5a695a371c8289c0fc672a34efff6188126824ef3b60\r\n INFO:transformers.modeling_utils:Weights from pretrained model not used in BertForMaskedLM: ['cls.seq_relationship.weight', 'cls.seq_relationship.bias']\r\n Traceback (most recent call last):\r\n File \"./traingreek.py\", line 268, in <module>\r\n train()\r\n File \"./traingreek.py\", line 161, in train\r\n add_special_tokens_(model, tokenizer)\r\n File \"./traingreek.py\", line 51, in add_special_tokens_\r\n orig_num_tokens = len(tokenizer.encoder)\r\n AttributeError: 'BertTokenizer' object has no attribute 'encoder'\r\n", "That's because the tokenizer does not have an `encoder` attribute. If you're looking to get the size of the tokenizer, you can do it with `len(tokenizer)`.", "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" ]

[]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Hi @Weilin37,\r\n\r\nHow did you count the length of your output? The `min_length` corresponds to the minimum number of tokens in the output which is `<=` number of words.\r\nTo count the number of tokens you have in your output text, you could use `tokenizer.tokenize(OUTPUT_TEXT)`", "> Hi @Weilin37,\r\n> \r\n> How did you count the length of your output? The `min_length` corresponds to the minimum number of tokens in the output which is `<=` number of words.\r\n> To count the number of tokens you have in your output text, you could use `tokenizer.tokenize(OUTPUT_TEXT)`\r\n\r\nHi. Ah ok thanks for clarifying. I had mistakenly thought it was the # of words.", "Hi @patrickvonplaten,\r\n\r\nIs it possible to set the minimum number of words instead of tokens?" ]

[ { "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": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "model pages: https://huggingface.co/models?filter=polish,question-answering\r\n\r\nThank you!" ]

[]

[]

[ "see https://github.com/huggingface/transformers/issues/3576" ]

[]

[]

[ "We would need more information in order to help you debug, namely transformers version, python version, code example, etc.\r\n\r\nHave you seen the issue template? Respecting it will ensure you get help :slightly_smiling_face: " ]

[]

[]

[ "With distillation, the initialization is really important in order to obtain good results. You can read the [paper](https://arxiv.org/pdf/1910.01108.pdf) to have more information, look for section 3 and \"Student Initialization\"!\r\n\r\nThe important part is the following:\r\n\r\n_Student initialization In addition to the previously described optimization and architectural choices,\r\nan important element in our training procedure is to find the right initialization for the sub-network to\r\nconverge. Taking advantage of the common dimensionality between teacher and student networks,\r\nwe initialize the student from the teacher by taking one layer out of two._", "Why not [0, 2, 4, 6, 8, 10]?" ]

[]

[]

[ "additionally, the `xlnet-large-cased` should not combine with `--do_lower_case` as `xlnet` model only has a cased version." ]

[ { "id": 1834054694, "node_id": "MDU6TGFiZWwxODM0MDU0Njk0", "url": "https://api.github.com/repos/huggingface/transformers/labels/TensorFlow", "name": "TensorFlow", "color": "FF6F00", "default": false, "description": "Anything TensorFlow" }, { "id": 1834081910, "node_id": "MDU6TGFiZWwxODM0MDgxOTEw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Usage", "name": "Usage", "color": "e28436", "default": false, "description": "General questions about the library" } ]

[]

[ "I would assume that you need to unpack the dictionary as you pass it to the model:\r\n\r\n```python\r\nmodel(**features)\r\n```\r\n\r\nEDIT: I was wrong, since the TF version should be used differently than the PT version.", "Thank you @BramVanroy, but another error arose:\r\n\r\n```python\r\nmodel(**features)\r\n```\r\n\r\n```python\r\n---------------------------------------------------------------------------\r\nValueError Traceback (most recent call last)\r\n<ipython-input-20-6248eef0b628> in <module>()\r\n----> 1 model(**features)\r\n\r\n/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer.py in __call__(self, *args, **kwargs)\r\n 798 else:\r\n 799 raise ValueError(\r\n--> 800 'The first argument to `Layer.call` must always be passed.')\r\n 801 \r\n 802 call_context = base_layer_utils.call_context()\r\n\r\nValueError: The first argument to `Layer.call` must always be passed.\r\n\r\n---------------------------------------------------------------------------\r\nNOTE: Current TensorFlow version is 2.2.0-rc2. To use TF 1.x instead,\r\nrestart your runtime (Ctrl+M .) and run \"%tensorflow_version 1.x\" before\r\nyou run \"import tensorflow\".\r\n---------------------------------------------------------------------------\r\n```", "I believe the error comes from the fact that you''re lacking a dimension in your features. All inputs to the model should have a shape of `[batch_size, sequence_length]`, whereas from your output:\r\n\r\n```py\r\n{\r\n 'attention_mask': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], dtype=int32)>,\r\n 'input_ids': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([ 101, 13366, 2131, 1035, 6819, 2094, 1035, 2013, 1035, 24471, 2140, 1006, 24471, 2140, 1007, 102], dtype=int32)>\r\n}\r\n```\r\n\r\nYour tensors are of shape `[sequence_length]`. You can unsqueeze those to add a batch dimension (or simply batch them, since you're already making use of the attention mask), and it should work.", "@LysandreJik,\r\nYes, it was really missing a dimension in `features`. \r\nUsing `features = tokenized_dataset.batch(2)` produces the desired input to transformer:\r\n\r\n```python\r\n({'attention_mask': <tf.Tensor: shape=(2, 16), dtype=int32, numpy=\r\n array([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],\r\n [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]], dtype=int32)>,\r\n 'input_ids': <tf.Tensor: shape=(2, 16), dtype=int32, numpy=\r\n array([[ 101, 13366, 2131, 1035, 6819, 2094, 1035, 2013, 1035,\r\n 24471, 2140, 1006, 24471, 2140, 1007, 102],\r\n [ 101, 13366, 8254, 2050, 1035, 20950, 1035, 2000, 1035,\r\n 24471, 2140, 1035, 2862, 1006, 20950, 102]], dtype=int32)>},\r\n {'attention_mask': <tf.Tensor: shape=(2, 16), dtype=int32, numpy=\r\n array([[1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],\r\n [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]], dtype=int32)>,\r\n 'input_ids': <tf.Tensor: shape=(2, 16), dtype=int32, numpy=\r\n array([[ 101, 27059, 2678, 8909, 2013, 24471, 2140, 1012, 102,\r\n 0, 0, 0, 0, 0, 0, 0],\r\n [ 101, 2358, 2099, 1011, 1028, 2862, 10463, 20950, 2000,\r\n 24471, 2140, 2862, 1012, 2013, 12170, 102]], dtype=int32)>})\r\n```\r\nThank you." ]

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

[]

[ "Have the same problem and do you have any url to download the bart-large-cnn or and other pretrained model", "No chinese support, yet. \r\nDownload:\r\n```bash\r\nwget https://s3.amazonaws.com/models.huggingface.co/bert/facebook/bart-large-cnn/pytorch_model.bin\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 had a similar problem. There is a bug in conversion script, this pull request fixes the issue: https://github.com/huggingface/transformers/pull/3642", "@sdadas Thanks a lot. I've pulled the latest change and it works now.", "Great to hear, closing!" ]

[]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Good catch @dmytyar :-) " ]

[]

[]

[ "**Try using the T5 summarizer instead like below:**\r\n```python\r\nfrom transformers import T5Tokenizer, T5ForConditionalGeneration\r\n\r\ntokenizer = T5Tokenizer.from_pretrained('t5-small')\r\nmodel = T5ForConditionalGeneration.from_pretrained('t5-small')\r\ninputs = tokenizer.batch_encode_plus([\"summarize: \" + example_text], max_length=1024, return_tensors=\"pt\", pad_to_max_length=True) # Batch size 1\r\noutputs = model.generate(inputs['input_ids'], num_beams=4, max_length=50, early_stopping=True)\r\n\r\nprint([tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=False) for g in outputs])\r\n```\r\n\r\n**The above excerpt gave me a summary of:**\r\n\r\n*'the survival rate among patients with metastatic renal-cell carcinoma has plateaued . agents such as sunitinib that target the vascular endothelial growth factor pathway are standard first-line therapy for advanced disease'*\r\n\r\n**If you still want to use Bart:**\r\n\r\nMy assumption is that this is not a bug. I may be wrong, but it seems the Bart summarizer just has a bias towards pointing to the first couple sentences of the original text. It's still abstractive, as can be seen by subtle differences in the summary you're getting. If you specify `min_length` as a higher value, like 100, you start to see that there are pointers to sentences that are not just in the first couple sentences.\r\n\r\n**Trying a `min_length` of a 100 using `bart-large-cnn` gave me the below summary:**\r\n\r\n*'Renal-cell carcinoma is characterized by susceptibility to both immunotherapeutic and antiangiogenic treatment approaches and resistance to cytotoxic chemotherapy. Agents such as sunitinib that target the vascular endothelial growth factor (VEGF) pathway are standard first-line therapy for advanced disease. **We conducted the KEYNOTE-426 trial to determine whether pembrolizumab plus axit inib would result in better outcomes than sunit in patients with previously untreated advanced renal- cell carcinoma.**'`*\r\n\r\nYou can see that the last sentence is not a part of the initial text excerpt", "As @aychang95 suggested you have to play around with the `generate` method arguments to see what works best for your example. Especially take a look at `num_beams`, `max_length`, `min_length`, `early_stopping` and `length_penalty`. \r\n\r\nI just noticed that I forget to add a good default setting to the Bart summarization pipeline. Just uploaded it - see here: https://s3.amazonaws.com/models.huggingface.co/bert/facebook/bart-large-cnn/config.json\r\n\r\nThe summarization pipeline should work better now :-) ", "> As @aychang95 suggested you have to play around with the `generate` method arguments to see what works best for your example. Especially take a look at `num_beams`, `max_length`, `min_length`, `early_stopping` and `length_penalty`.\r\n> \r\n> I just noticed that I forget to add a good default setting to the Bart summarization pipeline. Just uploaded it - see here: https://s3.amazonaws.com/models.huggingface.co/bert/facebook/bart-large-cnn/config.json\r\n> \r\n> The summarization pipeline should work better now :-)\r\n\r\nThank you! How do I go about updating the model? My code is below but I receive an error:\r\n\r\n```\r\nfrom transformers import pipeline, AutoTokenizer, AutoModel\r\ntokenizer = AutoTokenizer.from_pretrained(\"facebook/bart-large-cnn\")\r\nmodel = AutoModel.from_pretrained(\"facebook/bart-large-cnn\")\r\nsummarizer = pipeline(\"summarization\", model = model, tokenizer = tokenizer)\r\n```\r\n\r\n\r\n> OSError: Model name 'facebook/bart-large-cnn' was not found in tokenizers model name list (bart-large, bart-large-mnli, bart-large-cnn, bart-large-xsum). We assumed 'facebook/bart-large-cnn' was a path, a model identifier, or url to a directory containing vocabulary files named ['vocab.json', 'merges.txt'] but couldn't find such vocabulary files at this path or url.\r\n", "```\r\nfrom transformers import pipeline, AutoTokenizer, AutoModelWithLMHead\r\ntokenizer = AutoTokenizer.from_pretrained(\"bart-large-cnn\")\r\nmodel = AutoModelWithLMHead.from_pretrained(\"bart-large-cnn\")\r\nsummarizer = pipeline(\"summarization\", model = model, tokenizer = tokenizer)\r\n```\r\nworks :-).\r\n\r\nNote that \"bart-large-cnn\" is the default model for the summarization pipeline. The code above is equivalent to: \r\n\r\n```\r\nfrom transformers import pipeline\r\nsummarizer = pipeline(\"summarization\")\r\n```", "I was also able to discover another reason of why the summarization cut off. I believe setting the max_length conflicted with whatever the default min_length was. It looks like max_length takes priority and so the summary was cut off. I think it would be useful if this was managed automatically somehow, or at least display a warning.", "Hi @patrickvonplaten I just found that summarization takes 1024 words into consideration for generating a summary on its default parameters. I would like to know if I can increase the input size in order to consider more words while generating a summary in any case.\r\nI got the following message.\r\n\r\n`Your max_length is set to 1300, but you input_length is only 1024. You might consider decreasing max_length manually, e.g. summarizer('...', max_length=50)\r\n`\r\n\r\n", "As far as I know for `Bart` the `max_length` is 1024 and for `T5` it's 512. So depending on your model, I don't think you can increase the `max_length` higher than its `max_length` value.", "@patrickvonplaten I got your point. I have another question, what is the maximum token ( or words ) we can provide to Bart for a summary generation. Also, what should I do in order to generate a summary from a large text which contains approximately 100k words in it?", "A text that contains 100k words is probably more of a novel than a \"text\" :D. \r\nSo for these kinds of text using Bart you would need to chunk the text. Your memory would explode anyways at such sizes. In a couple of days we will add Reformer which can handle super long input text. We will soon also have an encoder-decoder model for Reformer which you could then use for summarization." ]

[ { "id": 2107554019, "node_id": "MDU6TGFiZWwyMTA3NTU0MDE5", "url": "https://api.github.com/repos/huggingface/transformers/labels/Distributed%20Training%20/%20Models", "name": "Distributed Training / Models", "color": "fef2c0", "default": false, "description": "" } ]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3578?src=pr&el=h1) Report\n> Merging [#3578](https://codecov.io/gh/huggingface/transformers/pull/3578?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/a4ee4da18ad659b196582bbdf40785033ee1d26b?el=desc) will **decrease** coverage by `0.10%`.\n> The diff coverage is `12.90%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3578?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3578 +/- ##\n==========================================\n- Coverage 78.05% 77.94% -0.11% \n==========================================\n Files 100 100 \n Lines 17135 17166 +31 \n==========================================\n+ Hits 13374 13380 +6 \n- Misses 3761 3786 +25 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3578?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_t5.py](https://codecov.io/gh/huggingface/transformers/pull/3578/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190NS5weQ==) | `77.18% <12.90%> (-4.34%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3578/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl91dGlscy5weQ==) | `93.62% <0.00%> (+0.32%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3578?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3578?src=pr&el=footer). Last update [a4ee4da...3bfeebe](https://codecov.io/gh/huggingface/transformers/pull/3578?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\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", "Hello! Do you have plans to merge this feature to master branch? \r\nI tried to make it locally in clonned repo but I got an error while tried to use it:\r\n<ipython\r\n\r\n> -input-22-5591bd8e45c0> in main()\r\n> 143 cache_dir=model_args.cache_dir,\r\n> 144 )\r\n> --> 145 model = model.spread_on_devices(['cpu', 'cpu'])\r\n> 146 \r\n> 147 # Get datasets\r\n> \r\n> /usr/local/lib/python3.6/dist-packages/transformers/modeling_t5.py in spread_on_devices(self, devices)\r\n> 936 return\r\n> 937 \r\n> --> 938 modules_to_move = set(self.modules)\r\n> 939 \r\n> 940 # Evenly spread the blocks on devices\r\n> \r\n> TypeError: 'method' object is not iterable", "Hey @exelents, \r\n\r\nAt the moment I don't think anybody is working on it and I'm not sure what the importance of this PR is at the moment. Feel free to take over the PR and try to make it work. I would be more than happy to help you if you open a PR :-) ", "This is very much related: https://github.com/huggingface/transformers/issues/7526", "This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.\n\nPlease note that issues that do not follow the [contributing guidelines](https://github.com/huggingface/transformers/blob/master/CONTRIBUTING.md) are likely to be ignored.", "This feature was awesome! I think this would be a major improvement to the transformers package!" ]

[]

[]

[ "@Ierezell I am facing same issue. How did you fix this issue? " ]

[]

[]

[ "@patrickvonplaten ", "Not yet :-) For translation you could try to create a `run_train.py` script using the following resources:\r\n\r\n- How to run `T5` for translation: https://github.com/huggingface/transformers/tree/master/examples/translation/t5\r\n- How to train `Bart` for summarization (should be very similar to \"How to train\" T5 for translation): https://github.com/huggingface/transformers/blob/master/examples/summarization/bart/run_train.sh and https://github.com/huggingface/transformers/blob/master/examples/summarization/bart/run_bart_sum.py", "For dialog generation - I would not recommend using T5, but rather a \"decoder-only\" model like gpt2. You could take a look at this script with implements a SOTA dialogue bot using DialoGPT from Microsoft: https://huggingface.co/microsoft/DialoGPT-medium#how-to-use", "Thanks for answering @patrickvonplaten \r\nOne more query: how to create the **data files** and **vocab file** for T5. \r\n\r\nIf I am not wrong, it requires 4 data files and 1 vocab file. And in train.source and val.source files, each instance should have a prefix like \"translate English to German: \". I prepare the data in the same format.\r\nIt gives this error.\r\nTypeError: 'NoneType' object is not iterable\r\n", "@sshleifer ", "in examples/transformer_base.py change line 105 to\r\n```python\r\navg_loss = getattr(self.trainer, \"avg_loss\", 0.0)\r\n```", "@sshleifer Thanks for the reply. It didn't work. and have the same error. Below is the screenshot of the changed line.\r\n\r\n<img width=\"937\" alt=\"Screenshot 2020-04-12 at 9 17 24 PM\" src=\"https://user-images.githubusercontent.com/30004110/79077852-8a308c00-7d04-11ea-92a2-30d9c6fb0c84.png\">", "@prabalbansal I think @sshleifer means line 107, so I added this patch in PR #3768", "@hugoabonizio Thanks for the patch. It works.", "@sshleifer @hugoabonizio When I use the model to predict for test set using the following command:\r\n\r\npython '/content/transformers-master/examples/summarization/bart/run_bart_sum.py' --data_dir='/content/drive/My Drive/two_keywords/' --model_type=t5 --output_dir=/content/t5 --do_predict --model_name_or_path=t5-small\r\n\r\nError generated:\r\n<img width=\"1010\" alt=\"Screenshot 2020-04-13 at 6 18 47 PM\" src=\"https://user-images.githubusercontent.com/30004110/79137728-8a3b9500-7db3-11ea-90e4-218cbc3e1e74.png\">\r\n", "> Thanks for answering @patrickvonplaten\r\n> One more query: how to create the **data files** and **vocab file** for T5.\r\n> \r\n> If I am not wrong, it requires 4 data files and 1 vocab file. And in train.source and val.source files, each instance should have a prefix like \"translate English to German: \". I prepare the data in the same format.\r\n> It gives this error.\r\n> TypeError: 'NoneType' object is not iterable\r\n> \r\n\r\nHi, I didn't understand why you have to prepare the vocab file. I think the pertrained T5 and its default tokenizer will take care of the tokenization? Thanks for your response.", "@MichaelZhouwang yes we didn't need vocab file here.", "> For dialog generation - I would not recommend using T5, but rather a \"decoder-only\" model like gpt2. You could take a look at this script with implements a SOTA dialogue bot using DialoGPT from Microsoft: https://huggingface.co/microsoft/DialoGPT-medium#how-to-use\r\n\r\n\r\nHi @patrickvonplaten,\r\n\r\nI'm trying out T5 and BART for dialogue generation. \r\n\r\nI'm wondering why you say that it's better to just have a decoder. Both FacebookAI's Blender and Google's Meena had encoders in their architectures.\r\n\r\nWhat's the reason for decoder-only systems being better?", "1) @Valdegg I think you are correct that it makes sense to use a seq2seq model.\r\n2) We are also currently working on porting blenderbot from parlai, which was trained on dialogue. 3) 3) We have new forums at https://discuss.huggingface.co/ for discussing higher-level things like which model to use. ", "The same question is there any example about training T5 to translate multiple sentences?" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[]

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

[]

[ "Can any one help me\r\nQAQ", "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 } ]

[]

[ "I meet the same bug when i use bart as an embedding layer.\r\nHave u solve the problem?", "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 meet the same bug when i use bart as an embedding layer.\r\n> Have u solve the problem?\r\n\r\nany luck on this? seeing the same \"RuntimeError: expected device cuda:0 but got device cpu\"" ]

[ { "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" } ]

[]

[ "You could cache the model once in your environment and then load it from there. Just point `from_pretrained` to the directory containing the model and configuration (or tokenizer file if loading the tokenizer) instead of the S3 link.", "@LysandreJik Completely understood and right but this is different case here. I am trying to make use of pipeline and load using pipeline (). And now how this can be achieved in GCP. ", "The pipeline also accepts directories as models and tokenizers. See the [pipeline documentation](https://huggingface.co/transformers/main_classes/pipelines.html#transformers.pipeline):\r\n\r\n- model (str or PreTrainedModel or TFPreTrainedModel, optional, defaults to None) –\r\n\r\n The model that will be used by the pipeline to make predictions. This can be None, **a string checkpoint identifier** or an actual pre-trained model inheriting from PreTrainedModel for PyTorch and TFPreTrainedModel for TensorFlow.\r\n\r\n If None, the default of the pipeline will be loaded.", "Thanks @LysandreJik. The documentation link did helped to get better clarity. Will try and get back. ", "Tried this way ```\r\ntokenizer = AutoTokenizer.from_pretrained(\"distilbert-base-cased-distilled-squad\")\r\n\r\nmodel = AutoModelForQuestionAnswering.from_pretrained(\"https://s3.amazonaws.com/models.huggingface.co/bert/distilbert-base-cased-distilled-squad-pytorch_model.bin\")\r\n\r\nnlp_qa = pipeline('question-answering', model=model, tokenizer=tokenizer)\r\n```\r\nGetting \r\n`UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 0: invalid start byte`", "You can't put a URL like that. It has to be a local file path, like it is shown in the documentation. You can either fetch them and save them to a directory:\r\n\r\n```\r\nmkdir local_path\r\ncd local_path\r\n\r\nwget https://s3.amazonaws.com/models.huggingface.co/bert/distilbert-base-cased-distilled-squad-pytorch_model.bin\r\nwget https://s3.amazonaws.com/models.huggingface.co/bert/distilbert-base-cased-distilled-squad-config.json\r\nwget https://s3.amazonaws.com/models.huggingface.co/bert/bert-large-cased-vocab.txt\r\n```\r\n\r\nor in Python:\r\n\r\n```py\r\nmodel = DistilBertModel.from_pretrained(\"distilbert-base-cased-distilled-squad\")\r\nmodel.save_pretrained(\"local_path\")\r\n\r\ntokenizer = DistilBertTokenizer.from_pretrained(\"distilbert-base-cased-distilled-squad\")\r\ntokenizer.save_pretrained(\"local_path\")\r\n```\r\n\r\nYou can then access this model/tokenizer:\r\n\r\n```py\r\nnlp = pipeline(\"question-answering\", model=\"local_path\", tokenizer=\"local_path\")\r\n```", "Thanks @LysandreJik ", "@vijender412 may I ask how you got a newer version of pytorch to work on cloud function? I'm unable to get mine to build with anything later than torch vesrion 1.0.1 which is stopping me from using pipeline :/", "I tried to run transformer on Cloud Functions v1 but as expected I could not run it due to the lack of it's resources.\r\n\r\n@vijender412 \r\nDid you make it?" ]

[]

[]

[ "@makaveli10 how to solve it ?" ]

[ { "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": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3568?src=pr&el=h1) Report\n> Merging [#3568](https://codecov.io/gh/huggingface/transformers/pull/3568?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/b38d552a92a0a201c005afae0e1b861ae6de9ce0&el=desc) will **increase** coverage by `0.96%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3568?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3568 +/- ##\n==========================================\n+ Coverage 76.90% 77.87% +0.96% \n==========================================\n Files 100 100 \n Lines 17127 17127 \n==========================================\n+ Hits 13172 13338 +166 \n+ Misses 3955 3789 -166 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3568?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/3568/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `81.54% <0.00%> (+1.34%)` | :arrow_up: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/3568/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `75.77% <0.00%> (+2.29%)` | :arrow_up: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/3568/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `98.24% <0.00%> (+2.63%)` | :arrow_up: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/3568/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `95.71% <0.00%> (+10.00%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3568/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `89.93% <0.00%> (+81.20%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3568?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3568?src=pr&el=footer). Last update [b38d552...2298b3b](https://codecov.io/gh/huggingface/transformers/pull/3568?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Can you please add a \r\n\r\n```\r\n---\r\nlanguage: german\r\n---\r\n```\r\nmetadata block at the top of the file? \r\n\r\nAlso, cc'ing @severinsimmler who might be interested in this (if you guys don't know each other already)", "Hey! I added the metablock.\r\n\r\nOne question: We uploaded our models as described in the huggingface documentation and everything looks okay, but when we try to test them with the suggested code, we get a message that the model is not found (OS Error: Model name was not found in tokenizers model name list). Could you please check if we made some error?\r\n\r\nthis is the test code we used:\r\n\r\n> tokenizer = AutoTokenizer.from_pretrained(\"redewiedergabe/bert-base-historical-german-rw-cased\")\r\n ", "Hi @redewiedergabe,\r\n\r\nI can't reproduce your error on `transformers` 2.7.0, Python 3.7.6 and macOS 10.15.4. Does loading the model with `AutoModel` work?", "works for me too", "Thanks! [Model page](https://huggingface.co/redewiedergabe/bert-base-historical-german-rw-cased)" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[]

[]

[]

[ "@singletongue Please give your opinion.", "Great! It looks good to me.", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3566?src=pr&el=h1) Report\n> Merging [#3566](https://codecov.io/gh/huggingface/transformers/pull/3566?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/b38d552a92a0a201c005afae0e1b861ae6de9ce0&el=desc) will **increase** coverage by `0.96%`.\n> The diff coverage is `33.33%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3566?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3566 +/- ##\n==========================================\n+ Coverage 76.90% 77.87% +0.96% \n==========================================\n Files 100 100 \n Lines 17127 17127 \n==========================================\n+ Hits 13172 13338 +166 \n+ Misses 3955 3789 -166 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3566?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/tokenization\\_bert\\_japanese.py](https://codecov.io/gh/huggingface/transformers/pull/3566/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYmVydF9qYXBhbmVzZS5weQ==) | `67.46% <33.33%> (ø)` | |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/3566/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `81.54% <0.00%> (+1.34%)` | :arrow_up: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/3566/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `75.77% <0.00%> (+2.29%)` | :arrow_up: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/3566/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `98.24% <0.00%> (+2.63%)` | :arrow_up: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/3566/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `95.71% <0.00%> (+10.00%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3566/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `89.93% <0.00%> (+81.20%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3566?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3566?src=pr&el=footer). Last update [b38d552...15522ac](https://codecov.io/gh/huggingface/transformers/pull/3566?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Looks good to me. @LysandreJik?" ]

[ { "id": 1834081910, "node_id": "MDU6TGFiZWwxODM0MDgxOTEw", "url": "https://api.github.com/repos/huggingface/transformers/labels/Usage", "name": "Usage", "color": "e28436", "default": false, "description": "General questions about the library" } ]

[]

[ "SciBert is a BERT model.\r\n\r\nPlease be more descriptive. Saying that features are not calculated correctly is not very helpful. Please describe the problem in full.", "Hi @BramVanroy , following is the error stack\r\n```code\r\n04/01/2020 09:10:02 - INFO - __main__ - Training/evaluation parameters Namespace(adam_epsilon=1e-08, block_size=1000000000000, cache_dir=None, config_name=None, device=device(type='cuda'), do_eval=True, do_train=True, eval_all_checkpoints=False, eval_data_file='/media/data1/ravi/covid-challenge/lm_data/dev.txt', evaluate_during_training=False, fp16=False, fp16_opt_level='O1', gradient_accumulation_steps=1, learning_rate=5e-05, line_by_line=False, local_rank=-1, logging_steps=500, max_grad_norm=1.0, max_steps=-1, mlm=True, mlm_probability=0.15, model_name_or_path='allenai/scibert_scivocab_uncased', model_type='bert', n_gpu=1, no_cuda=False, num_train_epochs=1.0, output_dir='./lm_finetune', overwrite_cache=False, overwrite_output_dir=False, per_gpu_eval_batch_size=4, per_gpu_train_batch_size=4, save_steps=500, save_total_limit=None, seed=42, server_ip='', server_port='', should_continue=False, tokenizer_name='allenai/scibert_scivocab_uncased', train_data_file='/media/data1/ravi/covid-challenge/lm_data/train.txt', warmup_steps=0, weight_decay=0.0)\r\n04/01/2020 09:10:02 - INFO - __main__ - Creating features from dataset file at /media/data1/ravi/covid-challenge/lm_data\r\n04/01/2020 09:28:10 - INFO - __main__ - Saving features into cached file /media/data1/ravi/covid-challenge/lm_data/bert_cached_lm_999999999998_train.txt\r\nTraceback (most recent call last):\r\n File \"examples/run_language_modeling.py\", line 781, in <module>\r\n main()\r\n File \"examples/run_language_modeling.py\", line 731, in main\r\n global_step, tr_loss = train(args, train_dataset, model, tokenizer)\r\n File \"examples/run_language_modeling.py\", line 224, in train\r\n train_sampler = RandomSampler(train_dataset) if args.local_rank == -1 else DistributedSampler(train_dataset)\r\n File \"/media/data2/anaconda/envs/covid/lib/python3.6/site-packages/torch/utils/data/sampler.py\", line 94, in __init__\r\n \"value, but got num_samples={}\".format(self.num_samples))\r\nValueError: num_samples should be a positive integer value, but got num_samples=0\r\n```\r\n\r\nIn the `bert_cached_lm_999999999998_train.txt` file there is only one line. `train.txt` file is of size **345MB**.\r\n\r\nThank you for your help!!", "This seems to indicate a problem with the dataset. Can you post the contents of bert_cached_lm_999999999998_train.txt? Maybe you chose a block size that is larger than your data size.", "This is probably the same error as https://github.com/huggingface/transformers/issues/3443#issuecomment-607422291", "@graviraja Had the same issue which is most likely related to \r\n`model = torch.nn.DataParallel(model)` in [trainer.py](https://github.com/huggingface/transformers/blob/8e093e5981e573a0b591dc57e8d52cc3efe82230/src/transformers/trainer.py#L250)\r\n\r\nUncommenting this line or using only one GPU \r\n`export CUDA_VISIBLE_DEVICES=1`\r\nworks in my case:\r\n", "> CUDA_VISIBLE_DEVICES\r\n\r\nNote that `CUDA_VISIBLE_DEVICES=1` does not mean to use \"just one GPU\", but it means to specifically use GPU with ID#1. However, device are zero-indexed, so the first GPU on your system will typically be #0: `CUDA_VISIBLE_DEVICES=0`", "Thanks, @BramVanroy , makes perfect sense.\r\n\r\n@graviraja I've set up a [notebook](https://github.com/Nikoschenk/language_model_finetuning/blob/master/scibert_fine_tuner.ipynb) with the required functionality. Previous comments regarding `block_size` were in fact crucial.\r\n\r\n", "Thank you @Nikoschenk @BramVanroy for the support." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834056635, "node_id": "MDU6TGFiZWwxODM0MDU2NjM1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Tokenization", "name": "Core: Tokenization", "color": "FF4446", "default": false, "description": "Internals of the library; Tokenization." } ]

[ { "login": "sshleifer", "id": 6045025, "node_id": "MDQ6VXNlcjYwNDUwMjU=", "avatar_url": "https://avatars.githubusercontent.com/u/6045025?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sshleifer", "html_url": "https://github.com/sshleifer", "followers_url": "https://api.github.com/users/sshleifer/followers", "following_url": "https://api.github.com/users/sshleifer/following{/other_user}", "gists_url": "https://api.github.com/users/sshleifer/gists{/gist_id}", "starred_url": "https://api.github.com/users/sshleifer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/sshleifer/subscriptions", "organizations_url": "https://api.github.com/users/sshleifer/orgs", "repos_url": "https://api.github.com/users/sshleifer/repos", "events_url": "https://api.github.com/users/sshleifer/events{/privacy}", "received_events_url": "https://api.github.com/users/sshleifer/received_events", "type": "User", "site_admin": false } ]

[ "You can't set the ids, they are set automatically from the sentence piece model.\r\nBut (1) why are you using the T5Tokenizer for a Bart checkpoint and (2) why do you want to tweak the id?", "(1) I used the `T5Tokenizer` in order to make a runnable example that did not require checking out my `mbart` branch.\r\n\r\n(2) Fairseq's MBART logic is split into two stages:\r\n- use `spm_encode --model sentence.bpe.model` to preprocess. (this is like encode_as_pieces in python).\r\n- use a `vocab.json` style lookup to convert each token to an ID.\r\n\r\nI'm trying to do that in one step, using `sp_model.encode_as_ids`, but my ids are off by 1, because the special tokens (sp_model.bos_token, etc) are different than fairseq's dictionary object:\r\n\r\n\r\n\r\n\r\nSo I need to either manipulate the sp_model, retrain it with correct control codes, or try a different approach.\r\n\r\n", "Yes you can check how we do these token index offset stuff (it’s specific to fairseq + sentencepiece) in Camembert and XLMRoberta tokenizers.", "Extremely helpful! Mbart also adds a language code like en_XX and ro_RO to the end of the source and target sentences. So the sentences are like `[tokens]+[<eos>, <language_id>]`\r\n\r\nDo we have any tokenizers that do that? ", "can't find an easy way to generate examples like\r\n```python\r\ninput_ids = [src_tokens]+[<eos>, <src_language_id>]\r\ndecoder_input_ids = [tgt_tokens]+[<eos>, <tgt_language_id>]\r\n```\r\nwhere the special tokens depend on the language.\r\n\r\nMy best idea is to add a method \r\n```python\r\ndef prepare_language_pair_batch(self, source_sentences, source_lang, target_sentences=None, target_lang=None):\r\n\t# encode source sentence\r\n\t# if target_sentence is None ignore it else process it\r\n return {input_ids=encoded_source_ids, attention_mask=attention_mask, decoder_input_ids=processed_target}\r\n```\r\n(Could also overwrite `prepare_inputs_for_model` and add arguments.)\r\n\r\n\r\nTwo other ideas that don't quite work:\r\n- Try to stuff the language codes into the string as text in `prepare_text_for_tokenization`. The problem is this would go before EOS.\r\n- Try to do the magic in `build_inputs_with_special_tokens`. the problem is that you still can't use `prepare_for_model` because it doesn't pass kwargs to `build_inputs_with_special_tokens`.\r\n\r\nWe could also instantiate two tokenizers with different special tokens, but that feels wasteful.\r\n\r\n@LysandreJik @patrickvonplaten ", "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", "> Yes you can check how we do these token index offset stuff (it’s specific to fairseq + sentencepiece) in Camembert and XLMRoberta tokenizers.\r\n\r\nFor posterity, I think Thomas means this:\r\n```\r\nhttps://huggingface.co/transformers/v4.6.0/_modules/transformers/models/camembert/tokenization_camembert.html\r\nhttps://huggingface.co/transformers/v3.5.1/_modules/transformers/tokenization_xlm_roberta.html\r\n```\r\n" ]

[]

[]

[]

[]

[]

[ "Hi, could you specify which version of `transformers` you're running?", "I encountered the same problem when using Albert. @voidful \r\n```\r\nAutoTokenizer.from_pretrained('voidful/albert_chinese_xxlarge')\r\n```\r\nwill raise\r\n```\r\n04/04/2020 14:21:28 - INFO - Model name 'voidful/albert_chinese_xxlarge' not found in model shortcut name list (albert-base-v1, albert-large-v1, albert-xlarge-v1, albert-xxlarge-v1, albert-base-v2, albert-large-v2, albert-xlarge-v2, albert-xxlarge-v2). Assuming 'voidful/albert_chinese_xxlarge' is a path, a model identifier, or url to a directory containing tokenizer files.\r\nTraceback (most recent call last):\r\n File \"preprocess.py\", line 353, in <module>\r\n main()\r\n File \"preprocess.py\", line 303, in main\r\n tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_path, do_lower_case=not args.cased, cache_dir=args.cache_dir)\r\n File \"/data0/username/anaconda3/lib/python3.7/site-packages/transformers/tokenization_auto.py\", line 192, in from_pretrained\r\n return tokenizer_class_py.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs)\r\n File \"/data0/username/anaconda3/lib/python3.7/site-packages/transformers/tokenization_utils.py\", line 393, in from_pretrained\r\n return cls._from_pretrained(*inputs, **kwargs)\r\n File \"/data0/username/anaconda3/lib/python3.7/site-packages/transformers/tokenization_utils.py\", line 496, in _from_pretrained\r\n list(cls.vocab_files_names.values()),\r\nOSError: Model name 'voidful/albert_chinese_xxlarge' was not found in tokenizers model name list (albert-base-v1, albert-large-v1, albert-xlarge-v1, albert-xxlarge-v1, albert-base-v2, albert-large-v2, albert-xlarge-v2, albert-xxlarge-v2). We assumed 'voidful/albert_chinese_xxlarge' was a path, a model identifier, or url to a directory containing vocabulary files named ['spiece.model'] but couldn't find such vocabulary files at this path or url.\r\n```\r\n```\r\n>>> torch.__version__\r\n'1.3.1'\r\n>>> transformers.__version__\r\n'2.7.0'\r\n```\r\n", "@LysandreJik @WiseDoge \r\nthe problem is that the model type is different with the tokenizer type.\r\neg. the model use albert model type and tokenizer use bert tokenizer, so the autoken class won know about it\r\n\r\nyou should let others can specify the tokenizer class or tokenizer model type if nessuary", "waiting for confirm or feature requests", "Thank you. I use BERT tokenizer instead, and it works.", "Since sentencepiece is not used in albert_chinese model\r\nyou have to call BertTokenizer instead of AlbertTokenizer !!! we can eval it using an example on MaskedLM\r\n\r\n[colab trial](https://colab.research.google.com/drive/1Wjz48Uws6-VuSHv_-DcWLilv77-AaYgj) \r\n```python\r\nfrom transformers import *\r\nimport torch\r\nfrom torch.nn.functional import softmax\r\n\r\npretrained = 'voidful/albert_chinese_large'\r\ntokenizer = BertTokenizer.from_pretrained(pretrained)\r\nmodel = AlbertForMaskedLM.from_pretrained(pretrained)\r\n\r\ninputtext = \"今天[MASK]情很好\"\r\n\r\nmaskpos = tokenizer.encode(inputtext, add_special_tokens=True).index(103)\r\n\r\ninput_ids = torch.tensor(tokenizer.encode(inputtext, add_special_tokens=True)).unsqueeze(0) # Batch size 1\r\noutputs = model(input_ids, masked_lm_labels=input_ids)\r\nloss, prediction_scores = outputs[:2]\r\nlogit_prob = softmax(prediction_scores[0, maskpos]).data.tolist()\r\npredicted_index = torch.argmax(prediction_scores[0, maskpos]).item()\r\npredicted_token = tokenizer.convert_ids_to_tokens([predicted_index])[0]\r\nprint(predicted_token,logit_prob[predicted_index])\r\n```\r\nResult: `心 0.9422469735145569`", "close for now", "> Since sentencepiece is not used in albert_chinese model\r\n> you have to call BertTokenizer instead of AlbertTokenizer !!! we can eval it using an example on MaskedLM\r\n> \r\n> [colab trial](https://colab.research.google.com/drive/1Wjz48Uws6-VuSHv_-DcWLilv77-AaYgj)\r\n> \r\n> ```python\r\n> from transformers import *\r\n> import torch\r\n> from torch.nn.functional import softmax\r\n> \r\n> pretrained = 'voidful/albert_chinese_large'\r\n> tokenizer = BertTokenizer.from_pretrained(pretrained)\r\n> model = AlbertForMaskedLM.from_pretrained(pretrained)\r\n> \r\n> inputtext = \"今天[MASK]情很好\"\r\n> \r\n> maskpos = tokenizer.encode(inputtext, add_special_tokens=True).index(103)\r\n> \r\n> input_ids = torch.tensor(tokenizer.encode(inputtext, add_special_tokens=True)).unsqueeze(0) # Batch size 1\r\n> outputs = model(input_ids, masked_lm_labels=input_ids)\r\n> loss, prediction_scores = outputs[:2]\r\n> logit_prob = softmax(prediction_scores[0, maskpos]).data.tolist()\r\n> predicted_index = torch.argmax(prediction_scores[0, maskpos]).item()\r\n> predicted_token = tokenizer.convert_ids_to_tokens([predicted_index])[0]\r\n> print(predicted_token,logit_prob[predicted_index])\r\n> ```\r\n> \r\n> Result: `心 0.9422469735145569`\r\n\r\nI have tried this code\r\nfrom transformers import TFAutoModel, BertTokenizer\r\npretrained = 'voidful/albert_chinese_xlarge'\r\ntokenizer = BertTokenizer.from_pretrained(pretrained)\r\nmodel = TFAutoModel.from_pretrained(pretrained)\r\n\r\ninputs = tokenizer(\"我喜欢你!\", return_tensors=\"tf\")\r\noutputs = model(**inputs)\r\n\r\nprint(outputs)\r\n\r\nit encounters \r\n\r\nOSError: Can't load weights for 'voidful/albert_chinese_xlarge'. Make sure that:\r\n- 'voidful/albert_chinese_xlarge' is a correct model identifier listed on 'https://huggingface.co/models'\r\n- or 'voidful/albert_chinese_xlarge' is the correct path to a directory containing a file named one of tf_model.h5, pytorch_model.bin.", "> > Since sentencepiece is not used in albert_chinese model\r\n> > you have to call BertTokenizer instead of AlbertTokenizer !!! we can eval it using an example on MaskedLM\r\n> > [colab trial](https://colab.research.google.com/drive/1Wjz48Uws6-VuSHv_-DcWLilv77-AaYgj)\r\n> > ```python\r\n> > from transformers import *\r\n> > import torch\r\n> > from torch.nn.functional import softmax\r\n> > \r\n> > pretrained = 'voidful/albert_chinese_large'\r\n> > tokenizer = BertTokenizer.from_pretrained(pretrained)\r\n> > model = AlbertForMaskedLM.from_pretrained(pretrained)\r\n> > \r\n> > inputtext = \"今天[MASK]情很好\"\r\n> > \r\n> > maskpos = tokenizer.encode(inputtext, add_special_tokens=True).index(103)\r\n> > \r\n> > input_ids = torch.tensor(tokenizer.encode(inputtext, add_special_tokens=True)).unsqueeze(0) # Batch size 1\r\n> > outputs = model(input_ids, masked_lm_labels=input_ids)\r\n> > loss, prediction_scores = outputs[:2]\r\n> > logit_prob = softmax(prediction_scores[0, maskpos]).data.tolist()\r\n> > predicted_index = torch.argmax(prediction_scores[0, maskpos]).item()\r\n> > predicted_token = tokenizer.convert_ids_to_tokens([predicted_index])[0]\r\n> > print(predicted_token,logit_prob[predicted_index])\r\n> > ```\r\n> > \r\n> > \r\n> > \r\n> > \r\n> > \r\n> > \r\n> > \r\n> > \r\n> > \r\n> > \r\n> > \r\n> > Result: `心 0.9422469735145569`\r\n> \r\n> I have tried this code\r\n> from transformers import TFAutoModel, BertTokenizer\r\n> pretrained = 'voidful/albert_chinese_xlarge'\r\n> tokenizer = BertTokenizer.from_pretrained(pretrained)\r\n> model = TFAutoModel.from_pretrained(pretrained)\r\n> \r\n> inputs = tokenizer(\"我喜欢你!\", return_tensors=\"tf\")\r\n> outputs = model(**inputs)\r\n> \r\n> print(outputs)\r\n> \r\n> it encounters\r\n> \r\n> OSError: Can't load weights for 'voidful/albert_chinese_xlarge'. Make sure that:\r\n> \r\n> * 'voidful/albert_chinese_xlarge' is a correct model identifier listed on 'https://huggingface.co/models'\r\n> * or 'voidful/albert_chinese_xlarge' is the correct path to a directory containing a file named one of tf_model.h5, pytorch_model.bin.\r\n\r\nYou need to add `from_pt=True` in order to load a pytorch checkpoint.\r\n```python\r\nfrom transformers import TFAutoModel, BertTokenizer\r\npretrained = './albert_chinese_tiny'\r\ntokenizer = BertTokenizer.from_pretrained(pretrained)\r\nmodel = TFAutoModel.from_pretrained(pretrained, from_pt=True)\r\n\r\ninputs = tokenizer(\"我喜欢你!\", return_tensors=\"tf\")\r\noutputs = model(**inputs)\r\n```" ]

[]

[]

[ "Hi, you could use model.eval() or load pre-tuned model and run against test set\r\n\t# Load a trained model and vocabulary that you have fine-tuned\r\n\tmodel = model_class.from_pretrained(output_dir)\r\n\ttokenizer = tokenizer_class.from_pretrained(output_dir)\r\nSee this post: https://mccormickml.com/2019/07/22/BERT-fine-tuning/#a1-saving--loading-fine-tuned-model\r\n", "+1\r\n\r\nIs there something similar for testing to the `--do_train` or `--do_eval` flag in the glue examples?", "@Mahmedturk I tried it like this now:\r\n\r\n```python\r\nfrom transformers import BertTokenizer, BertForSequenceClassification\r\nimport torch\r\nimport pandas as pd\r\n\r\ntokenizer = BertTokenizer.from_pretrained('./my-model')\r\nmodel = BertForSequenceClassification.from_pretrained('./my-model')\r\nlabels = [ ... ]\r\n\r\ndf_test = pd.read_csv('./my-data/test.tsv', sep='\\t', names=['label', 'sentence'])\r\ndf_test['prediction'] = None\r\n\r\nfor row in df_test.itertuples():\r\n inputs = tokenizer.encode_plus(row.sentence, add_special_tokens=True, return_tensors='pt')\r\n pred = model(inputs['input_ids'], token_type_ids=inputs['token_type_ids'])[0].argmax().item()\r\n df_test.loc[row.Index, 'prediction'] = labels[pred]\r\n```\r\n\r\nThen you can filter the pandas dataframe, i.e. `df_test[df_test['label'] == df_test['prediction']]` to see the true positives.", "@olastor is there a way to print confusion matrix?\r\n", "Also for the evaluation set, i get the following three metrics. What is \"acc_and_f1\" in the below?\r\n\r\n\r\nacc = 0.9455445544554455\r\nacc_and_f1 = 0.8709204253758709\r\nf1 = 0.7962962962962963\r\n", "@Mahmedturk Not with a built in function in my example, but manually. Let's say you have the labels `True` and `False`, then the correct way to calculate the absolute values of the confusion matrix would be like this I think:\r\n\r\n- number of true positves: `len(df[(df['label'] == True]) & (df['prediction'] == True)])`\r\n- number of false positves: `len(df[(df['label'] == False]) & (df['prediction'] == True)])`\r\n- number of false negatives: `len(df[(df['label'] == True]) & (df['prediction'] == False)])`\r\n- number of true negatives: `len(df[(df['label'] == False]) & (df['prediction'] == False)])`", "@Mahmedturk From [here](https://github.com/huggingface/transformers/blob/master/src/transformers/data/metrics/__init__.py#L41):\r\n\r\n```python\r\ndef acc_and_f1(preds, labels):\r\n acc = simple_accuracy(preds, labels)\r\n f1 = f1_score(y_true=labels, y_pred=preds)\r\n return {\r\n \"acc\": acc,\r\n \"f1\": f1,\r\n \"acc_and_f1\": (acc + f1) / 2, # <---\r\n }\r\n```", "@sunyangfu \r\nAfter loading the saved model and vocabulary how do i run against the test set?\r\nSorry if this sounds silly. I am very new to PyTorch and deep learning.\r\nThe given link shows how to test on CoLa dataset which has only one sentence. Whereas in QQP dataset there are two sentences. What changes do i need to make in the code to test it with QQP dataset?", "\r\n@Mahmedturk Here you can run against the test set:\r\nAfter loading the model, you can code whatever you want to fit the model.\r\n```python\r\noutput_dir = './saved_model_dir/'\r\n\r\nMODEL_CLASSES = {\r\n 'bert': (BertConfig, BertForSequenceClassification, BertTokenizer),\r\n 'xlnet': (XLNetConfig, XLNetForSequenceClassification, XLNetTokenizer),\r\n 'xlm': (XLMConfig, XLMForSequenceClassification, XLMTokenizer),\r\n 'roberta': (RobertaConfig, RobertaForSequenceClassification, RobertaTokenizer),\r\n 'distilbert': (DistilBertConfig, DistilBertForSequenceClassification, DistilBertTokenizer),\r\n 'albert': (AlbertConfig, AlbertForSequenceClassification, AlbertTokenizer)\r\n}\t\r\n\r\n# Config class and load a trained model and vocabulary \r\nconfig_class, model_class, tokenizer_class = MODEL_CLASSES['bert']\t\r\nmodel = model_class.from_pretrained(output_dir)\r\ntokenizer = tokenizer_class.from_pretrained(output_dir)\r\n\r\n# Copy the model to the GPU.\r\nmodel.to(device)\r\n\r\n# Put model in evaluation mode\r\nmodel.eval()\r\n\r\n#Then you do test data pre-processing and apply the model on test data\r\nwith torch.no_grad():\r\n outputs = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask)\r\n\r\n\r\n```", "Hi @olastor,\r\n\r\nI have tried your method, it returns the below error\r\n\r\n df_test.loc[row.Index, 'prediction'] = labels[pred]\r\nIndexError: list index out of range. ", "@Mahmedturk Did you update the list of labels in my example for your task?", "labels = df_test.label.values", "> labels = df_test.label.values\r\n\r\nIt needs to be the same set of labels used for training.", "@olastor thanks.\r\n" ]

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

[]

[ "I found a workaround solution, just to pass the complete inputs including `input_ids, token_type_ids and attention_mask` to the trace method and invoke forward along with those 3 inputs. \r\nBut it would be great to know how I can just pass in input_ids and token_type_ids", "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 } ]

[]

[ "As you say yourself, the core of the library is to provide you with models (and recently also pipelines and in the future even more exciting things) to build your own projects. The examples show you some implementations that are in themselves usable but that are by no means meant to be exhaustive for all problems. As you indicate yourself, you are invited to adapt these examples to your own use case.\r\n\r\nIf you feel that your changes are useful for the whole community, then feel free to request a PR!", "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": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3557?src=pr&el=h1) Report\n> Merging [#3557](https://codecov.io/gh/huggingface/transformers/pull/3557?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/b38d552a92a0a201c005afae0e1b861ae6de9ce0&el=desc) will **increase** coverage by `0.97%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3557?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3557 +/- ##\n==========================================\n+ Coverage 76.90% 77.88% +0.97% \n==========================================\n Files 100 100 \n Lines 17127 17127 \n==========================================\n+ Hits 13172 13339 +167 \n+ Misses 3955 3788 -167 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3557?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3557/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl91dGlscy5weQ==) | `88.79% <0.00%> (+0.16%)` | :arrow_up: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/3557/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `81.54% <0.00%> (+1.34%)` | :arrow_up: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/3557/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `75.77% <0.00%> (+2.29%)` | :arrow_up: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/3557/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `98.24% <0.00%> (+2.63%)` | :arrow_up: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/3557/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `95.71% <0.00%> (+10.00%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/3557/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `89.93% <0.00%> (+81.20%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3557?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3557?src=pr&el=footer). Last update [b38d552...7e585a0](https://codecov.io/gh/huggingface/transformers/pull/3557?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3556?src=pr&el=h1) Report\n> Merging [#3556](https://codecov.io/gh/huggingface/transformers/pull/3556?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/ae6834e028ecdf7fdbe886c1f86d0e02d5fef6f0&el=desc) will **not change** coverage by `%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3556?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3556 +/- ##\n=======================================\n Coverage 77.80% 77.80% \n=======================================\n Files 100 100 \n Lines 17064 17064 \n=======================================\n Hits 13277 13277 \n Misses 3787 3787 \n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3556?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3556?src=pr&el=footer). Last update [ae6834e...2e14442](https://codecov.io/gh/huggingface/transformers/pull/3556?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "Hi @renatoviolin,\r\n\r\nthe T5 pipeline uses special input arguments for the `generate()` function that have been shown to work well for summarization. If you take a look at `task_specific_params` and under `summarization` in T5's config: https://s3.amazonaws.com/models.huggingface.co/bert/t5-base-config.json. You can see the `generate()` arguments that are used for pipeline.\r\n", "Hi @patrickvonplaten \r\n\r\nThanks for your attention. Now I got the point where I'm doing wrong.\r\nBut the most strange thing that happened is that in third step, after the pipeline was executed, the \r\nmodel.generate() produces similar results (and take long to run) as the pipeline.\r\nAt first glance, given the poor results and how quickly it runs, it seemed to me that the weights were not loaded.", "I think it's probably because beam search is deactivated, no length penalties, no repeat penalties and a very short max length is used " ]

[]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Hi @vsieplus ,\r\n\r\nThis is a known bug and sadly we don't have a solution for this now. TransfoXLLMHead uses adaptive weight embeddings which makes it not very easy to implement this function. Should be implemented in the long run though - I will note it down. @thomwolf @LysandreJik ", "@patrickvonplaten Does the same problem apply to XLNet?", "No it should not. XLNet uses the standard `nn.embedding` - so it should be fine.", "Hi, I faced the same issue and wrote some dirty code as a workaround in `modeling_utils.py`. The main idea is to just operate on the last embedding layer:\r\n```\r\ndef _resize_token_embeddings(self, new_num_tokens):\r\n old_embeddings = self.get_input_embeddings()\r\n\r\n if type(self).__name__ == 'TransfoXLModel':\r\n # since the 'TransfoXLModel' has multiple embedding layers, the last layer is resized\r\n new_num_tokens_last = new_num_tokens\r\n for emb_layer in old_embeddings.emb_layers[:-1]:\r\n new_num_tokens_last -= emb_layer.weight.size(0)\r\n\r\n new_embeddings_last = self._get_resized_embeddings(old_embeddings.emb_layers[-1], new_num_tokens_last)\r\n new_embeddings = old_embeddings\r\n new_embeddings.emb_layers[-1] = new_embeddings_last\r\n else:\r\n new_embeddings = self._get_resized_embeddings(old_embeddings, new_num_tokens)\r\n\t\t\r\n self.set_input_embeddings(new_embeddings)\r\n return self.get_input_embeddings()\r\n```\r\nIt workes for me (at least I get no error). Can someone confirm that this makes sense? Maybe @patrickvonplaten ?\r\n", "Sorry for bothering again @patrickvonplaten, but this is important for me: Can you or someone else comment on my \"fix\" above wether it makes sense?\r\nThanks in advance!", "This looks okay to me, though I think you can patch a custom `_resize_token_embeddings(self, new_num_tokens)` to [`TransfoXLPreTrainedModel`](https://github.com/huggingface/transformers/blob/3e5928c57d57db3071638e6beaec9349a75b6a22/src/transformers/modeling_transfo_xl.py#L451) to avoid making the test (and leave the default behavior for other models).\r\n\r\nActually adding such a method to `TransfoXLPreTrainedModel` would solve this issue AFAICT. Since you wrote it @RafaelWO, you should make a PR with it :-)", "Thanks for your feedback @sgugger ! I will move the logic into the `TransfoXLPreTrainedModel` and make my first PR :)", "Out of curiosity, why do you go with\r\n\r\n```\r\nfor emb_layer in old_embeddings.emb_layers[:-1]:\r\n new_num_tokens_last -= emb_layer.weight.size(0)\r\n```\r\n\r\nWouldn't just `emb_layer = old_embeddings.emb_layers[-1]` work out ? Also are `wug` and `wugs` often used ? If they're syntax tokens, which are frequent, you might want to add them to the corresponding embedding group.", "I think the for loop is to make sure `new_num_tokens_last` is accurate by substracting the other embedding sizes.\r\n\r\nI agree that ideally, the method written on `TransfoXLPreTrainedModel` should have an argument to decide to which embedding layer add the new tokens (with a default to the last one).", "Yes that's correct @sgugger, thanks for answering.\r\n\r\nI understand the idea of your introduced parameter, but for me the question is whether this makes sense? Because if you add the new token into e.g. the first layer, you would have to insert it also at the same position in your tokenizer and shift all tokens after that.\r\n\r\n@TevenLeScao \r\n> Also are wug and wugs often used ?\r\n\r\nIn my case I want to a `cls_token` which is not included in the pretrained tokenizer.", "Ah my bad, misread the `:-1` into `-1:`. I've looked again at the `ProjectedAdaptiveLogSoftmax` and adding elsewhere should be fine if you update the `cutoffs` attribute to make sure it takes into account the changed embedding size.\r\n\r\nAdding at the end is a good baseline; the only issue is that you're going to lose out on some of the benefits of the adaptive softmax as you're often going to have to access the bigger softmax layer whereas you usually want to have the frequent tokens (such as `cls`) on smaller ones.", "> update the cutoffs attribute to make sure it takes into account the changed embedding size.\r\n\r\n> Adding at the end is a good baseline; the only issue is that you're going to lose out on some of the benefits of the adaptive softmax as you're often going to have to access the bigger softmax layer whereas you usually want to have the frequent tokens (such as cls) on smaller ones.\r\n\r\nYes and yes, that's true. \r\n\r\nBut as I mentioned above: if you add such a common token into the first smaller layer and adjust the cutoffs (which would be the preferred way to do), you have a conflict with the tokenizer, because there the new token is at the end and not at position `20001` as in your model (default cutoffs `[20000, 40000, 200000]`).\r\n\r\nOr am I missing something?", "Yes, that is also going to be a problem, but it shouldn't be too hard to solve with a simple conversion function that shifts the other tokens. The cleanest way to do it would probably be to update the tokenizer yourself but I am not sure how easy that would be. ", "Thanks a lot @sgugger for answering here! As @sgugger mentioned, it'd be great if you can add a `_resize_token_embeddings()` function to `TransfoXLPreTrainedModel`. \r\n\r\nThe solution looks great to me @vsieplus :-) \r\n\r\nYou could make it a bit more compact, but that's a nitpick: \r\n\r\n```python \r\n embeddings = self.get_input_embeddings()\r\n new_num_tokens_last = new_num_tokens - sum([emb.shape[0] for emb in embeddings.emb_layers[:-1])\r\n new_embeddings_last = self._get_resized_embeddings(embeddings.emb_layers[-1], new_num_tokens_last)\r\n embeddings.emb_layers[-1] = new_embeddings_last\r\n\r\n self.set_input_embeddings(embeddings)\r\n```", "Hello, I have faced the same issue while using `TFTransfoXLLMHeadModel` class. \r\nAfter initializing `TFTransfoXLLMHeadModel` I attempted to apply `.resize_token_embeddings()`, but it raised `NotImplementedError` error. \r\n\r\nI checked the github [link ](https://github.com/huggingface/transformers/blob/main/src/transformers/models/transfo_xl/modeling_tf_transfo_xl.py) and figured it out that `.resize_token_embeddings()` was really not implemented. \r\n\r\nWhy is that so? Do you guys have any plan to fix this error? " ]

[]

[ { "login": "patrickvonplaten", "id": 23423619, "node_id": "MDQ6VXNlcjIzNDIzNjE5", "avatar_url": "https://avatars.githubusercontent.com/u/23423619?v=4", "gravatar_id": "", "url": "https://api.github.com/users/patrickvonplaten", "html_url": "https://github.com/patrickvonplaten", "followers_url": "https://api.github.com/users/patrickvonplaten/followers", "following_url": "https://api.github.com/users/patrickvonplaten/following{/other_user}", "gists_url": "https://api.github.com/users/patrickvonplaten/gists{/gist_id}", "starred_url": "https://api.github.com/users/patrickvonplaten/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/patrickvonplaten/subscriptions", "organizations_url": "https://api.github.com/users/patrickvonplaten/orgs", "repos_url": "https://api.github.com/users/patrickvonplaten/repos", "events_url": "https://api.github.com/users/patrickvonplaten/events{/privacy}", "received_events_url": "https://api.github.com/users/patrickvonplaten/received_events", "type": "User", "site_admin": false } ]

[ "Hi @dhecloud, \r\n\r\nThanks for you issue :-) \r\nDoes the model still work fine? ", "> Hi @dhecloud,\r\n> \r\n> Thanks for you issue :-)\r\n> Does the model still work fine?\r\n\r\nHi, thanks for your reply.\r\nUsing the examples provided in the doc, the model works fine. \r\nBefore i used `T5WithLMHeadModel` in version `2.5.1` which did not raise this missing keys warning. After i moved to `T5ForConditionalGeneration` in `2.7.0` there was this warning and my training loss diverged so i thought i might raise this issue in case there was some sort of change in naming in the checkpoint", "I'm gonna take a look :-) ", "Hi guys, \r\nAny news on this?\r\nWhen I try to load t5-base I receive this:\r\n\r\n\r\nINFO:transformers.modeling_utils:Weights of T5ForConditionalGeneration not initialized from pretrained model: ['encoder.embed_tokens.weight', 'decoder.embed_tokens.weight', 'lm_head.weight']", "> Hi guys,\r\n> Any news on this?\r\n> When I try to load t5-base I receive this:\r\n> \r\n> INFO:transformers.modeling_utils:Weights of T5ForConditionalGeneration not initialized from pretrained model: ['encoder.embed_tokens.weight', 'decoder.embed_tokens.weight', 'lm_head.weight']\r\n\r\ni think it's mostly a harmless misplaced error. The model should still work fine. You can test it by trying out the examples ", "Yeah this should not be a problem, all these weights are `[encoder.embed_tokens.weight', 'decoder.embed_tokens.weight', 'lm_head.weight']` are weights tied to the input embedding matrix and therefore don't need to be initialized.", "How can we silence the error?", "It should be enough to lower the cli-logger", "@sgugger @patrickvonplaten this seems like a very old post, but I just got this warning:\r\n\r\n```\r\n[WARNING|trainer.py:2231] 2023-11-23 23:20:29,636 >> There were missing keys in the checkpoint model loaded: ['encoder.embed_tokens.weight', 'decoder.embed_tokens.weight']. \r\n```\r\n\r\n on `4.35.2` trying to restart my training from ckpt after my training run crashed. I'm working with FLAN-T5 base. I restarted by passing `resume_from_checkpoint=True` to the `Trainer`.\r\n \r\n I think there was recently an attempt to migrate to `safetensors` so maybe that's why the warning crept in there? Also, rather strangely, the progress training bar started from `0` steps ... I do remember that the `Trainer` used to forward the data loader to the point where you left off when you stopped training and such tricks, but nothing really indicates that this is happening or that indeed I am restarting from my checkpoint ...\r\n \r\n UPDATE: Just reading through the `ignore_data_skip` docs, I'm pretty sure that I haven't started training from the checkpoint, training started rather quickly so I betting the farm on no batch being skipped...\r\n\r\nUPDATE 2: I should expect that to skip batches the trainer state would have to be saved (right?). So given that my code failed at `L2407` (so during `_save_checkpoint`) because I mistakenly passed a metric that I was not calculating as the metric for best model, I presume this elucidates the mystery of not skipping batches. The question remains, has the training started from the checkpoint or not?\r\n\r\n\r\n ", "> Yeah this should not be a problem, all these weights are `[encoder.embed_tokens.weight', 'decoder.embed_tokens.weight', 'lm_head.weight']` are weights tied to the input embedding matrix and therefore don't need to be initialized.\r\n\r\nHello, what do you mean are \"tied to the input\"? \r\n \r\nI am trying to convert a flax finetuned T5 model into pytorch format using these commands:\r\n - `from_flax_model = T5ForConditionalGeneration.from_pretrained(model_path, from_flax=True)`\r\n - `from_flax_model.save_pretrained(folder)`\r\n \r\nand I get the same message: `Some weights of T5ForConditionalGeneration were not initialized from the Flax model and are newly initialized: ['decoder.embed_tokens.weight', 'lm_head.weight', 'encoder.embed_tokens.weight']\r\nYou should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.`" ]

[ { "id": 1838412367, "node_id": "MDU6TGFiZWwxODM4NDEyMzY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/model%20card", "name": "model card", "color": "92d5f4", "default": false, "description": "Related to pretrained model cards" } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3552?src=pr&el=h1) Report\n> Merging [#3552](https://codecov.io/gh/huggingface/transformers/pull/3552?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/83d1fbcff608f84a27234e20d6531b4404dc059e&el=desc) will **decrease** coverage by `0.49%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3552?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3552 +/- ##\n==========================================\n- Coverage 78.31% 77.81% -0.50% \n==========================================\n Files 100 100 \n Lines 17064 17064 \n==========================================\n- Hits 13363 13278 -85 \n- Misses 3701 3786 +85 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3552?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/data/processors/squad.py](https://codecov.io/gh/huggingface/transformers/pull/3552/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL3Byb2Nlc3NvcnMvc3F1YWQucHk=) | `28.89% <0.00%> (-27.60%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3552?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3552?src=pr&el=footer). Last update [83d1fbc...cd4f658](https://codecov.io/gh/huggingface/transformers/pull/3552?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "Just pre-process them as you would for any other NLP task. This may include: normalisation of punctuation, dealing with HTML/XML, white-space normalisation, language verification. Something that people used to do in the age of word2vec is creating new entities for specific items. For instance, any number could be replaced with all 1s (1234 becomes 1111) to have relatively small vocabulary that still takes the number of characters into account. Same with users (e.g @user@) and mentions and urls (@URL@), and so on. This might still be worth the effort, but not necessarily. In such cases you may want to ensure that these tokens are not split in the tokenizer.\r\n\r\nI hope that you understand that this is a very general question not related to this repository, so I am closing this.", "@BramVanroy I think I forgot to mention that I'm using the gpt-2 training models from this repo. I'm not re-training gpt-2. I think I shouldn't create new entities for specific terms if that process has not happened during training step. Am I right? ", "If you are not planning to pretraining the model (and generating a new vocab) then, indeed, you should not try to add new tokens. So in your case you would just need to do some basic normalisation of punctuation, HTML, etc.", "@BramVanroy Thanks for the answer. I'm directly using gpt-2 pretrained models from https://huggingface.co/models . Specifically the ones that are created by Huggingface. I can't find the preprocessing code that was used when those models were trained... so I can replicate the same at inference time. I'm wondering if I should just assume things or it'd be safer to see how it was trained so I can make sure to prepare the sentences in a similar way.", "Has there been any preprocessing during training phase @BramVanroy ? ", "I don't know. Note that HuggingFace didn't train GPT2. They ported the weights to their own architecture. You can try to get into contact with the people that created it, OpenAI. https://github.com/openai/gpt-2", "@Damiox have you found the original preprocessing code?", "@don-prog no, I have not 😢 - I am doing some subset of the preprocessing heuristics that @BramVanroy detailed before when serving the model. But I still think it'd be really good to have a consistency between both preprocessing mechanisms: training (whatever it was) vs inference. I just haven't had the time to identify that original preprocessing code from OpenAI" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3550?src=pr&el=h1) Report\n> Merging [#3550](https://codecov.io/gh/huggingface/transformers/pull/3550?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/6f5a12a5833d1e3783e4b8a42cb556b64085745e&el=desc) will **decrease** coverage by `0.49%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/3550?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #3550 +/- ##\n==========================================\n- Coverage 78.30% 77.80% -0.50% \n==========================================\n Files 100 100 \n Lines 17062 17062 \n==========================================\n- Hits 13360 13275 -85 \n- Misses 3702 3787 +85 \n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/3550?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/data/processors/squad.py](https://codecov.io/gh/huggingface/transformers/pull/3550/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL3Byb2Nlc3NvcnMvc3F1YWQucHk=) | `28.89% <0.00%> (-27.60%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/3550?src=pr&el=continue).\n> **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\n> `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\n> Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/3550?src=pr&el=footer). Last update [6f5a12a...c3ce2fe](https://codecov.io/gh/huggingface/transformers/pull/3550?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[ { "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" ]

[]

[]

[ "What do you think @mfuntowicz?", "[Related issue ](https://github.com/huggingface/transformers/issues/2488)", "Actually, I realized that instead of `index`, even better would be the word piece `offsets`, similar to that returned by `tokenizer.encode` from the huggingface [tokenizers package](https://github.com/huggingface/tokenizers).\r\n\r\nWill get to implementing this within the week if this isn't supported yet!", "This recently merged [PR](https://github.com/huggingface/transformers/pull/3957) should solve this issue 🙂 " ]

[]

[]

[]