Datasets:

---

DanielPFlorian

/

Transformers-Github-Issues

like 0

[]

[]

[ "Awesome. Could you check if the existing @slow tests break for Roberta, and add a new one that hardcodes the fairseq logits from your example and makes sure we also return them. Trying to avoid accidental breakage. Thanks again! ", "@sshleifer Not sure how I would hardcode a tensor of size 1, 12, 50265. Can I just add a small pickled file to the test instead?", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2913?src=pr&el=h1) Report\n> Merging [#2913](https://codecov.io/gh/huggingface/transformers/pull/2913?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/59c23ad9c931ac4fe719abeb3c3851df046ef3a6?src=pr&el=desc) will **decrease** coverage by `<.01%`.\n> The diff coverage is `85.71%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2913?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2913 +/- ##\n==========================================\n- Coverage 75.3% 75.29% -0.01% \n==========================================\n Files 94 94 \n Lines 15424 15424 \n==========================================\n- Hits 11615 11614 -1 \n- Misses 3809 3810 +1\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2913?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2913/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `95.3% <85.71%> (-0.47%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2913?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/2913?src=pr&el=footer). Last update [59c23ad...51514a4](https://codecov.io/gh/huggingface/transformers/pull/2913?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Sorry for the close, had to do some rebasing." ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2912?src=pr&el=h1) Report\n> Merging [#2912](https://codecov.io/gh/huggingface/transformers/pull/2912?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/59c23ad9c931ac4fe719abeb3c3851df046ef3a6?src=pr&el=desc) will **increase** coverage by `0.01%`.\n> The diff coverage is `100%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2912?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2912 +/- ##\n==========================================\n+ Coverage 75.3% 75.32% +0.01% \n==========================================\n Files 94 94 \n Lines 15424 15438 +14 \n==========================================\n+ Hits 11615 11628 +13 \n- Misses 3809 3810 +1\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2912?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/tokenization\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2912/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHMucHk=) | `90.14% <100%> (+0.05%)` | :arrow_up: |\n| [src/transformers/tokenization\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/2912/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYmVydC5weQ==) | `96.99% <100%> (+0.06%)` | :arrow_up: |\n| [src/transformers/tokenization\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2912/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fcm9iZXJ0YS5weQ==) | `100% <100%> (ø)` | :arrow_up: |\n| [src/transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2912/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==) | `61.15% <0%> (-0.18%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2912?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/2912?src=pr&el=footer). Last update [59c23ad...3b7752b](https://codecov.io/gh/huggingface/transformers/pull/2912?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "Got it. \r\nThis parameter is for the context.", "May I know how do you solve this problem? I just ran into this problem.", "You will have to add a \"para\" field for every choice - This is for adding knowledge. To get a baseline you can simply use a dummy text in that field \r\n\r\n`{\r\n \"id\": \"MCAS_2000_4_6\",\r\n \"question\": {\r\n \"stem\": \"Which technology was developed most recently?\",\r\n \"choices\": [\r\n {\r\n \"text\": \"cellular telephone\",\r\n \"label\": \"A\",\r\n \"para\": \"fetched knowledge\"\r\n },\r\n {\r\n \"text\": \"television\",\r\n \"label\": \"B\",\r\n \"para\": \"fetched knowledge\"\r\n },\r\n {\r\n \"text\": \"refrigerator\",\r\n \"label\": \"C\",\r\n \"para\": \"fetched knowledge\"\r\n },\r\n {\r\n \"text\": \"airplane\",\r\n \"label\": \"D\",\r\n \"para\": \"fetched knowledge\"\r\n }\r\n ]\r\n },\r\n \"answerKey\": \"A\"\r\n}\r\n`" ]

[ { "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 @bryant1410, \r\n\r\nThanks for reporting the issue, as a workaround for now, can you try the following:\r\n\r\n```python\r\ntokenizer.tokenize(\"abc\")\r\ntokenizer.tokenizer(\"def\")\r\n```\r\n\r\nIt should do the same, let me know.\r\nIn the meantime I'll have a closer look at the function `tokenizer.build_inputs_with_special_tokens`", "Should be fixed in https://github.com/huggingface/transformers/pull/2912" ]

[]

[]

[ "> Please incorporate my comments on this file in the other PR :)\r\n\r\nis on my radar :-) ", "UPDATE: Changed slow tests for language generation design according to discussion in PR #2885 .\r\nIf this looks alright, I'll add test cases for the other LMModels @LysandreJik & @sshleifer ", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2909?src=pr&el=h1) Report\n> Merging [#2909](https://codecov.io/gh/huggingface/transformers/pull/2909?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/38f5fe9e0277df67a01db80a1c640ac072a2381e?src=pr&el=desc) will **decrease** coverage by `1.03%`.\n> The diff coverage is `100%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2909?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2909 +/- ##\n==========================================\n- Coverage 77.16% 76.12% -1.04% \n==========================================\n Files 98 98 \n Lines 15997 15997 \n==========================================\n- Hits 12344 12178 -166 \n- Misses 3653 3819 +166\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2909?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/2909/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG0ucHk=) | `88.43% <100%> (ø)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2909/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2909/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.71% <0%> (-10%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2909/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `73.48% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2909/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `96.03% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2909/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n| [src/transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2909/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==) | `92.2% <0%> (-0.17%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2909?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/2909?src=pr&el=footer). Last update [38f5fe9...0c5bdef](https://codecov.io/gh/huggingface/transformers/pull/2909?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Updated the language model generation slow tests following Roberta's and Bart's Integration Test style. What do you think? @LysandreJik @sshleifer ", "Finished to add hard-coded tests for all models with LMHead: `GPT2, OpenAI, XLNet, TransfoXL, CTRL and XLM.`\r\n\r\nAll pretrained models generate reasonable results **except** `XLM`. Might need to take a closer look in a future PR.\r\n\r\nAlso future PRs TODO:\r\n\r\n- [ ] Add hardcoded tests for seq-to-seq language generation\r\n- [ ] Add hardcoded tests for DoubleHead language generation", "Don't feel strongly, but I would consider deleting the XLM example so that the tests don't enforce bad generations.\r\nAlso, if it is possible to make shorter examples (or more than one token per line), it would make the code more readable.\r\nOverall, I love this. Makes me feel a lot safer editing generation code!", "We could also add a sentence to `Contributing.md` telling people who change the generation code which command to run to make sure they didn't break stuff\r\n" ]

[]

[]

[]

[ { "id": 1108649053, "node_id": "MDU6TGFiZWwxMTA4NjQ5MDUz", "url": "https://api.github.com/repos/huggingface/transformers/labels/Help%20wanted", "name": "Help wanted", "color": "008672", "default": false, "description": "Extra attention is needed, help appreciated" }, { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834056761, "node_id": "MDU6TGFiZWwxODM0MDU2NzYx", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Modeling", "name": "Core: Modeling", "color": "FF8446", "default": false, "description": "Internals of the library; Models." } ]

[]

[ "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", "h56cho: I'm not sure if you ask about the code or the algorithm.\r\nAs far as I understand from the code, the class MLP is a basic 2-layer neural network (2 consecutive conv1d + gelu activation). This will be used to construct a bigger [network](https://github.com/huggingface/transformers/blob/73028c5df0c28ca179fbe565482a9c2143787f61/src/transformers/modeling_gpt2.py#L215). I hope this (partially) answers your Q2 question.\r\nQ1. For the code, I think the comment is related to the number 3072 mentioned in this [link](https://jalammar.github.io/illustrated-gpt2/)\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" ]

[ { "id": 1834067346, "node_id": "MDU6TGFiZWwxODM0MDY3MzQ2", "url": "https://api.github.com/repos/huggingface/transformers/labels/Documentation", "name": "Documentation", "color": "77cc3b", "default": false, "description": "" } ]

[]

[]

[ { "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." } ]

[]

[ "Solved by #2746 " ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2903?src=pr&el=h1) Report\n> Merging [#2903](https://codecov.io/gh/huggingface/transformers/pull/2903?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/20fc18fbda3669c2f4a3510e0705b2acd54bff07?src=pr&el=desc) will **decrease** coverage by `1.07%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2903?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2903 +/- ##\n==========================================\n- Coverage 75% 73.92% -1.08% \n==========================================\n Files 94 94 \n Lines 15288 15288 \n==========================================\n- Hits 11467 11302 -165 \n- Misses 3821 3986 +165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2903?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2903/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2903/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.91% <0%> (-9.86%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2903/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `70.86% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2903/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `92.07% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2903/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2903?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/2903?src=pr&el=footer). Last update [20fc18f...afa57d9](https://codecov.io/gh/huggingface/transformers/pull/2903?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "[Looks good!](https://huggingface.co/nlpaueb/bert-base-greek-uncased-v1)", "@julien-c Could you perhaps shed some light on how the TF checkpoints should be uploaded? @iliaschalkidis asked about it here https://github.com/huggingface/transformers/issues/2901#issuecomment-588163505", "In the past for another project, this (`from_pt=True`) did the dirty trick for me:\r\n\r\n```python\r\nbert = BERT.from_pretrained(model_path+'pytorch_model.bin',\r\n from_pt=True,\r\n config=BertConfig().from_pretrained(model_path+'config.json'))\r\n```\r\nbut I definitely not recommend this...\r\n\r\nI have already uploaded the TF checkpoint files (`model_ckpt.data-00000-of-00001`, `model_ckpt.index`, `model_ckpt.meta`) in model's folder, so please feel free to troubleshoot.", "see https://github.com/huggingface/transformers/issues/2901#issuecomment-591710959" ]

[ { "id": 1834056761, "node_id": "MDU6TGFiZWwxODM0MDU2NzYx", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Modeling", "name": "Core: Modeling", "color": "FF8446", "default": false, "description": "Internals of the library; Models." }, { "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" } ]

[]

[ "The differences between the BERT and RoBERTa model are the following:\r\n\r\n- Different pre-training (larger batch size for RoBERTa, no NSP, no token type ids ...)\r\n- Different tokenizer\r\n\r\nThe model architecture is exactly the same. The only real difference after the pre-training is the difference in tokenization. Since you're working with a BERT model that was pre-trained, you unfortunately won't be able to change the tokenizer now from a WordPiece (BERT) to a Byte-level BPE (RoBERTa)." ]

[ { "id": 1834056761, "node_id": "MDU6TGFiZWwxODM0MDU2NzYx", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Modeling", "name": "Core: Modeling", "color": "FF8446", "default": false, "description": "Internals of the library; Models." }, { "id": 1843244711, "node_id": "MDU6TGFiZWwxODQzMjQ0NzEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/New%20model", "name": "New model", "color": "fbca04", "default": false, "description": "" } ]

[]

[ "My guess would be that you'd need to load (and then save) with `BertForPreTraining` rather than `BertModel`.\r\n\r\nhttps://github.com/huggingface/transformers/blob/20fc18fbda3669c2f4a3510e0705b2acd54bff07/src/transformers/modeling_bert.py#L806", "@BramVanroy you're my hero for today! Appreciated man! \r\n\r\nTest Cases:\r\n\r\n```python\r\nimport torch\r\nfrom transformers import *\r\n\r\nmodel_path = '/Users/kiddothe2b/Downloads/bert-base-greek-uncased-v2'\r\ntokenizer_greek = AutoTokenizer.from_pretrained(model_path)\r\nlm_model_greek = AutoModelWithLMHead.from_pretrained(model_path)\r\n\r\n# ================ EXAMPLE 1 ================\r\ntext_1 = 'O ποιητής έγραψε ένα [MASK] .'\r\n# EN: The poet wrote a [MASK] . '\r\ninput_ids = tokenizer_greek.encode(text_1)\r\nprint(tokenizer_greek.convert_ids_to_tokens(input_ids))\r\n# ['[CLS]', 'o', 'ποιητης', 'εγραψε', 'ενα', '[MASK]', '.', '[SEP]']\r\noutputs = lm_model_greek(torch.tensor([input_ids]))[0]\r\nprint(tokenizer_greek.convert_ids_to_tokens(outputs[0, 5].max(0)[1].item()))\r\n# the most plausible prediction for [MASK] is \"song\"\r\n\r\n# ================ EXAMPLE 2 ================\r\ntext_2 = 'Είναι ένας [MASK] άνθρωπος.'\r\n# EN: He is a [MASK] person. '\r\ninput_ids = tokenizer_greek.encode(text_1)\r\nprint(tokenizer_greek.convert_ids_to_tokens(input_ids))\r\n# ['[CLS]', 'ειναι', 'ενας', '[MASK]', 'ανθρωπος', '.', '[SEP]']\r\noutputs = lm_model_greek(torch.tensor([input_ids]))[0]\r\nprint(tokenizer_greek.convert_ids_to_tokens(outputs[0, 3].max(0)[1].item()))\r\n# the most plausible prediction for [MASK] is \"good\"\r\n\r\n# ================ EXAMPLE 3 ================\r\ntext_3 = 'Είναι ένας [MASK] άνθρωπος και κάνει συχνά [MASK].'\r\n# EN: He is a [MASK] person he does frequently [MASK]. '\r\ninput_ids = tokenizer_greek.encode(text_3)\r\nprint(tokenizer_greek.convert_ids_to_tokens(input_ids))\r\n# ['[CLS]', 'ειναι', 'ενας', '[MASK]', 'ανθρωπος', 'και', 'κανει', 'συχνα', '[MASK]', '.', '[SEP]']\r\noutputs = lm_model_greek(torch.tensor([input_ids]))[0]\r\nprint(tokenizer_greek.convert_ids_to_tokens(outputs[0, 8].max(0)[1].item()))\r\n# the most plausible prediction for the second [MASK] is \"trips\"\r\n```\r\n\r\nI will release the updated version later today. Although I think this is a very important technical detail and needs to be referred in the examples.\r\n\r\nIs it possible to load the same saved model with TF2 binding?\r\n\r\n", "Tensorflow checkpoints can be loaded when using `from_pretrained`. Have a look at the documentation, particularly this line:\r\n\r\nhttps://github.com/huggingface/transformers/blob/20fc18fbda3669c2f4a3510e0705b2acd54bff07/src/transformers/modeling_utils.py#L317", "Ok, so it is also suggested to upload the initial TF checkpoint files [ `variables.data-00000-of-00001`, `variables.index`, `variables.meta`] through the CLI, otherwise people cannot use the model in TF2 with `AutoModelWith.from_pretrained()`? \r\n", "I am not sure to be honest. Perhaps someone else can help you further along.", "In the past for another project, this (`from_pt=True`) did the dirty trick for me:\r\n\r\n```python\r\nbert = BERT.from_pretrained(model_path+'pytorch_model.bin',\r\n from_pt=True,\r\n config=BertConfig().from_pretrained(model_path+'config.json'))\r\n```\r\nbut I definitely not recommend this...\r\n\r\nI have already uploaded the TF checkpoint files (`model_ckpt.data-00000-of-00001`, `model_ckpt.index`, `model_ckpt.meta`) in model's folder, so please feel free to troubleshoot.", "Looks like you managed to do it in the meantime?\r\n\r\n**For reference here's how to do it:** starting with version 2.5.0 thanks to this commit https://github.com/huggingface/transformers/pull/2765/commits/961c69776f8a2c95b92407a086848ebca037de5d\r\n\r\nYou can now just do \r\n```python\r\ntf_model = TFAutoModelForPreTraining.from_pretrained(\r\n \"nlpaueb/bert-base-greek-uncased-v1\", \r\n from_pt=True\r\n)\r\ntf_model.save_pretrained(dirname)\r\n```\r\n\r\n**And then you can upload the TF weights using the CLI.**\r\n\r\nOf course, if you have the model in a local folder, you don't need to use the remote id.\r\n\r\ncc @LysandreJik ", "> Looks like you managed to do it in the meantime?\r\n> \r\n> **For reference here's how to do it:** starting with version 2.5.0 thanks to this commit [961c697](https://github.com/huggingface/transformers/commit/961c69776f8a2c95b92407a086848ebca037de5d)\r\n> \r\n> You can now just do\r\n> \r\n> ```python\r\n> tf_model = TFAutoModelForPreTraining.from_pretrained(\r\n> \"nlpaueb/bert-base-greek-uncased-v1\", \r\n> from_pt=True\r\n> )\r\n> tf_model.save_pretrained(dirname)\r\n> ```\r\n> \r\n> Of course, if you have the model in a local folder, you don't need to use the remote id.\r\n> \r\n> cc @LysandreJik\r\n\r\nDoes that imply that the CLI upload should only upload PyTorch checkpoints? (I suppose for saving space.) I am asking because the documentation emphasises that loading PT checkpoints to TF and the other way around is quite slow. Also, if all checkpoints are indeed PyTorch only, it might be useful to set from_pt=True automatically when a model is fetched from the HF bucket (since those would then all contain PT checkpoints anyway).", "Thanx @julien-c and @BramVanroy \r\n\r\nIndeed I followed a very similar process:\r\n\r\n```python\r\n\r\nfrom transformers import TFBertForPreTraining\r\nfrom transformers import BertConfig\r\nmodel_path = '/home/ichalkidis/greek_bert/'\r\nbert = TFBertForPreTraining.from_pretrained(model_path + 'pytorch_model.bin',\r\n config=BertConfig().from_pretrained(model_path + 'config.json'), from_pt=True)\r\n\r\nbert.save_pretrained('/home/ichalkidis/bert-base-greek-uncased-v1/')\r\n```\r\n\r\nFrom now on, we also serve the `tf_model.h5` and everyone will be able to load the model in TF2 without any further issue.", "Ah, it seems that I misunderstood, then. In terms of the models themselves, you can upload the tf_model.h5 and the pytorch_model.bin, and when someone requests `/home/ichalkidis/bert-base-greek-uncased-v1/` based on the framework (TF or PT), the appropriate model (.h5 or .bin) is downloaded?", "@BramVanroy, yes that's how it works! You can also explicitely specify `from_pt`/`from_tf` to `from_pretrained` for the model to fetch the other framework's checkpoint and convert it.", "Yep @BramVanroy , I added the line \"And then you can upload the TF weights using the CLI.\" to my comment above to try and make that clearer", "I guess we can close this issue now @iliaschalkidis?", "Of course @julien-c . In the end, this was just a misunderstanding, not a bug at all. Thank all of you for your help!" ]

[]

[]

[]

[ { "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." } ]

[]

[ "Reading more, pretty sure we only expect to have the same results as fairseq when the argument to fairseq starts with a space. Closing but would love verification/knowledge !", "Yes, this comes from #2778, which changes the default behavior to automatically prepending a space when `add_special_tokens=True` for Roberta, since you want a space after `<s>`. Can be overriden with `add_prefix_space=False`. This does deviate from fairseq's encode fn, ~~but reflects the behavior of their `fill_mask` [which also prepends a space](https://github.com/pytorch/fairseq/blob/master/fairseq/models/roberta/hub_interface.py#L149).~~ Nvm, fairseq's `fill_mask` function doesn't prepend a space after all. They expect the user to know that they have to prepend a space to get correctly encoded sequences." ]

[ { "id": 1834053007, "node_id": "MDU6TGFiZWwxODM0MDUzMDA3", "url": "https://api.github.com/repos/huggingface/transformers/labels/Ex:%20LM%20(Pretraining)", "name": "Ex: LM (Pretraining)", "color": "76FFAF", "default": false, "description": "Related to language modeling pre-training" } ]

[]

[ "It is both b) and c) :).", "Hello @LysandreJik and @BramVanroy \r\n\r\nDid you have any results in training run_language_modeling.py for some language from scratch (i mean with and without NSP (next sentence prediction) as is in that script) ?\r\n\r\nDid you get better or relatively the same losses (and perplexity, accuracy) by first doing MLM then \r\nNSP ?\r\n\r\nwhat were your warmup steps and block size (maximum sequence length) for each (MLM , NSP) task , if you have down them separately ? ", "as @LysandreJik mentioned here\r\n \r\nhttps://github.com/huggingface/transformers/issues/2693#issuecomment-589819382\r\n\r\n\"the RoBERTa paper has proven that the NSP objective was not particularly helpful\"\r\n \r\nIs that also right for BERT ? (training for some language from scratch)\r\n", "If it is right , I think we can set block size (maximum sequence length) equal to 64 for MLM\r\n \r\nBecause the original paper used 128 (except for 10% of the last steps) , and it had two sentences (A and B for the sake of NSP and model learns that it shouldn't see B for filling masked words in A, because A and B aren't relevant to each other half of the times) which means average block size of 64 for each one \r\n\r\nAnd it means grate speed up in training BERT, if I say right\r\nbecause, I haven't got a TPU or even enough GPU to do that \r\n\r\nAs mentioned in original paper of BERT \"Longer sequences are disproportionately expensive because attention is quadratic to the sequence length\"", "Perhaps, another idea is that \r\nwords in sentence A do attention on sentence B, anyway\r\nand that attention is very important to get great results in MLM task (here I mean block size of 128 which means 64 for A and another 64 for B)\r\n\r\neven regarding the fact that A and B are related sentences, just half of the times \r\n(and that attentions are the main inputs for task NSP)\r\n\r\nAnd by using block size of 64 (and not doing task NSP during the training) , I will get very bad results " ]

[ { "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": "mfuntowicz", "id": 2241520, "node_id": "MDQ6VXNlcjIyNDE1MjA=", "avatar_url": "https://avatars.githubusercontent.com/u/2241520?v=4", "gravatar_id": "", "url": "https://api.github.com/users/mfuntowicz", "html_url": "https://github.com/mfuntowicz", "followers_url": "https://api.github.com/users/mfuntowicz/followers", "following_url": "https://api.github.com/users/mfuntowicz/following{/other_user}", "gists_url": "https://api.github.com/users/mfuntowicz/gists{/gist_id}", "starred_url": "https://api.github.com/users/mfuntowicz/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/mfuntowicz/subscriptions", "organizations_url": "https://api.github.com/users/mfuntowicz/orgs", "repos_url": "https://api.github.com/users/mfuntowicz/repos", "events_url": "https://api.github.com/users/mfuntowicz/events{/privacy}", "received_events_url": "https://api.github.com/users/mfuntowicz/received_events", "type": "User", "site_admin": false } ]

[ "After upgrading to 2.5.0, the code now throws\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"<stdin>\", line 1, in <module>\r\n File \"/Users/bilal/Documents/transformers/src/transformers/tokenization_utils.py\", line 587, in save_pretrained\r\n return vocab_files + (special_tokens_map_file, added_tokens_file)\r\nTypeError: unsupported operand type(s) for +: 'NoneType' and 'tuple'\r\n```", "Hi @bkkaggle , \r\n\r\nThanks for reporting the issue, it should be fixed through https://github.com/huggingface/transformers/pull/2918 and land very soon on master.\r\n\r\nI'll be included in the first maintenance release following 2.5\r\n\r\nMorgan", "Saving tokenizers works now, but restoring them doesn't\r\n```\r\n> from transformers import *\r\n> tok = GPT2TokenizerFast.from_pretrained('distilgpt2')\r\n> tok.save_pretrained('./')\r\n('./vocab.json-vocab.json', './vocab.json-merges.txt', './special_tokens_map.json', './added_tokens.json')\r\n> tok = GPT2TokenizerFast.from_pretrained('./')\r\nRobertaTokenizerFast has an issue when working on mask language modeling where it introduces an extra encoded space before the mask token.See https://github.com/huggingface/transformers/pull/2778 for more information.\r\n> tok.tokenize('test')\r\n[]\r\n```", "I can't reproduce on my side, using your code:\r\n\r\n```python\r\ntok = GPT2TokenizerFast.from_pretrained('distilgpt2')\r\ntok.save_pretrained('./')\r\n> ('./vocab.json-vocab.json', './vocab.json-merges.txt', './special_tokens_map.json', './added_tokens.json')\r\n\r\ntok = GPT2TokenizerFast.from_pretrained('./')\r\ntok.tokenize('test')\r\n> ['test']\r\n```", "I made a colab notebook to reproduce the error\r\nThe error appears when installing from source on the master branch\r\n\r\ncolab: https://colab.research.google.com/drive/1OJdm6LzVtyb-biVR1ky6joSX7gBgl6St", "Thanks, I'm able to reproduce now, I'll have a look hopefully tomorrow morning.\r\n\r\nI'll keep you posted here 👀 ", "Fixed" ]

[]

[]

[ "Hi, `from_pretrained` appeared in an older version of the library. `pytorch-pretrained-BERT` is a year old, is less robust and lacks certain functionalities (such as the one you mentioned) which are present in `transformers`." ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2895?src=pr&el=h1) Report\n> Merging [#2895](https://codecov.io/gh/huggingface/transformers/pull/2895?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/2ae98336d17fceea7506af9880b862b6252a38f6?src=pr&el=desc) will **decrease** coverage by `1.07%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2895?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2895 +/- ##\n==========================================\n- Coverage 75.06% 73.98% -1.08% \n==========================================\n Files 94 94 \n Lines 15288 15288 \n==========================================\n- Hits 11476 11311 -165 \n- Misses 3812 3977 +165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2895?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/2895/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9fX2luaXRfXy5weQ==) | `98.87% <ø> (ø)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2895/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2895/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.91% <0%> (-9.86%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2895/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `70.86% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2895/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `92.07% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2895/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2895?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/2895?src=pr&el=footer). Last update [2ae9833...014ad24](https://codecov.io/gh/huggingface/transformers/pull/2895?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "(I removed myself because GitHub suggests me as a reviewer on every PR because of a refactoring, and I can't review every PR, not because this is a bad idea. The PR is most likely good.)" ]

[]

[]

[ "Why don't you import it directly via\r\n`from transformers.modeling_tf_albert import TFAlbertMLMHead`\r\n?\r\nIn my opinion it is not good practise to expose everything via `__init__.py` because the autocomplete feature of an IDE will become messy.", "That's true, don't know why that slipped my mind. Thanks for the suggestion!" ]

[ { "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": "mfuntowicz", "id": 2241520, "node_id": "MDQ6VXNlcjIyNDE1MjA=", "avatar_url": "https://avatars.githubusercontent.com/u/2241520?v=4", "gravatar_id": "", "url": "https://api.github.com/users/mfuntowicz", "html_url": "https://github.com/mfuntowicz", "followers_url": "https://api.github.com/users/mfuntowicz/followers", "following_url": "https://api.github.com/users/mfuntowicz/following{/other_user}", "gists_url": "https://api.github.com/users/mfuntowicz/gists{/gist_id}", "starred_url": "https://api.github.com/users/mfuntowicz/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/mfuntowicz/subscriptions", "organizations_url": "https://api.github.com/users/mfuntowicz/orgs", "repos_url": "https://api.github.com/users/mfuntowicz/repos", "events_url": "https://api.github.com/users/mfuntowicz/events{/privacy}", "received_events_url": "https://api.github.com/users/mfuntowicz/received_events", "type": "User", "site_admin": false } ]

[ "Also cc'ing @fmikaelian on this for information :)", "Apologize for the careless mistake @fmikaelian ", "Hi, other than the careless mistake, I'm trying to understand why I cannot load any model from transformers S3 repo. I have tried :\r\n\r\n1) from transformers import FlaubertModel, FlaubertTokenizer\r\n\r\n2) from transformers import CamembertTokenizer\r\n\r\n3)from transformers import CamembertModel\r\n\r\n\r\n4)from transformers import BertModel\r\nmodel = BertModel.from_pretrained('bert-base-uncased') \r\n\r\nOnly the forth option has triggered the download process. All other options return : \r\n`\"ImportError: cannot import name 'CamembertModel'\"`\r\n\r\n i was wondering if there is an issue since I'm using conda in a Windows PC. \r\n\r\nMany thanks for your help.\r\n\r\n\r\n", "I tried to update transformers with conda but that did not work and I also tried to do some pip install but also getting some errors:\r\n\r\n```\r\nFile \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\...\\lib\\site-packages\\transformers\\configuration_utils.py\", line 145, in from_pretrained\r\n raise EnvironmentError(msg)\r\nOSError: Model name 'flaubert-base-uncased-squad' was not found in model name list (bert-base-uncased, bert-large-uncased, bert-base-cased, bert-large-cased, bert-base-multilingual-uncased, bert-base-multilingual-cased, bert-base-chinese, bert-base-german-cased, bert-large-uncased-whole-word-masking, bert-large-cased-whole-word-masking, bert-large-uncased-whole-word-masking-finetuned-squad, bert-large-cased-whole-word-masking-finetuned-squad, bert-base-cased-finetuned-mrpc, bert-base-german-dbmdz-cased, bert-base-german-dbmdz-uncased). We assumed 'flaubert-base-uncased-squad' was a path or url to a configuration file named config.json or a directory containing such a file but couldn't find any such file at this path or url.\r\n```", "As pointed out in my Stackoverflow answer, I suspect a versioning conflict. I successfully managed to load the pipeline in `2.5.0`, but had errors in `2.4.1` (not quite the same as @rcontesti , but similar enough for me to assume problems with an older version).", "Do you have torch installed in your environment? That might explain why you can't import `CamembertModel`.\r\n\r\nThe error \r\n```\r\nOSError: Model name 'flaubert-base-uncased-squad' was not found in model name list (bert-base-uncased, bert-large-uncased, bert-base-cased, bert-large-cased, bert-base-multilingual-uncased, bert-base-multilingual-cased, bert-base-chinese, bert-base-german-cased, bert-large-uncased-whole-word-masking, bert-large-cased-whole-word-masking, bert-large-uncased-whole-word-masking-finetuned-squad, bert-large-cased-whole-word-masking-finetuned-squad, bert-base-cased-finetuned-mrpc, bert-base-german-dbmdz-cased, bert-base-german-dbmdz-uncased). We assumed 'flaubert-base-uncased-squad' was a path or url to a configuration file named config.json or a directory containing such a file but couldn't find any such file at this path or url.\r\n```\r\n\r\nmeans you're trying to load a flaubert checkpoint in BERT. Could you share the code that raised the last error so that we may try to reproduce the error?", "Guyz thank so much for your answers. I was able to solve the version problem but now I'm running into a different problem(Should I open a new thread?):\r\n\r\nI'm currently using:\r\n\r\n```py\r\nmodel_=transformers.FlaubertForQuestionAnswering\r\ntokenizer_ = transformers.FlaubertTokenizer\r\n```\r\n\r\nBut when I place them into pipeline:\r\n\r\n```py\r\nnlp = pipeline('question-answering', \\\r\n model=model, \\\r\n tokenizer=tokenizer)\r\n```\r\n\r\nI'm getting the following error:\r\n\r\n```\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\multiprocessing\\pool.py\", line 119, in worker\r\n result = (True, func(*args, **kwds))\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\multiprocessing\\pool.py\", line 44, in mapstar\r\n return list(map(*args))\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\site-packages\\transformers\\data\\processors\\squad.py\", line 105, in squad_convert_example_to_features\r\n sub_tokens = tokenizer.tokenize(token)\r\nTypeError: tokenize() missing 1 required positional argument: 'text'\r\n\"\"\"\r\n\r\nThe above exception was the direct cause of the following exception:\r\n\r\nTraceback (most recent call last):\r\n File \"question_extraction.py\", line 61, in <module>\r\n answer, score=question_extraction(text, question_, model_, tokenizer_, language_, verbose= True)\r\n File \"question_extraction.py\", line 44, in question_extraction\r\n output=nlp({'question':question, 'context': text})\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\site-packages\\transformers\\pipelines.py\", line 802, in __call__\r\n for example in examples\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\socgen_nlp\\lib\\site-packages\\transformers\\pipelines.py\", line 802, in <listcomp>\r\n for example in examples\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\site-packages\\transformers\\data\\processors\\squad.py\", line 316, in squad_convert_examples_to_features\r\n desc=\"convert squad examples to features\",\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\site-packages\\tqdm\\std.py\", line 1097, in __iter__\r\n for obj in iterable:\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\multiprocessing\\pool.py\", line 320, in <genexpr>\r\n return (item for chunk in result for item in chunk)\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\multiprocessing\\pool.py\", line 735, in next\r\n raise value\r\nTypeError: tokenize() missing 1 required positional argument: 'text'\r\nconvert squad examples to features: 0%|\r\n```", "You need to initialize your model and tokenizer with a checkpoint, for example instead of \r\n```py\r\nmodel_=transformers.FlaubertForQuestionAnswering\r\ntokenizer_ = transformers.FlaubertTokenizer\r\n```\r\nYou would specify a flaubert checkpoint:\r\n```py\r\nmodel_ = transformers.FlaubertModel.from_pretrained(\"fmikaelian/flaubert-base-uncased-squad\")\r\ntokenizer_ = transformers.FlaubertTokenizer.from_pretrained(\"fmikaelian/flaubert-base-uncased-squad\")\r\n```\r\n\r\nI chose a community checkpoint that was trained using question answering. You can check all available FlauBERT models [here](https://huggingface.co/models?search=flaubert).", "Once again many thanks @LysandreJik for the help. I proceed as suggested and now when I'm trying to put both the tokenizer and the model into pipeline I'm running into the following error:\r\n\r\n`Traceback (most recent call last):\r\n File \"question_extraction.py\", line 72, in <module>\r\n answer, score=question_extraction(text, question_, model_, tokenizer_, language_, verbose= True)\r\n File \"question_extraction.py\", line 55, in question_extraction\r\n output=nlp({'question':question, 'context': text})\r\n File \"C:\\Users\\Ruben Contesti\\AppData\\Local\\Continuum\\Anaconda3\\envs\\..\\lib\\site-packages\\transformers\\pipelines.py\", line 818, in __call__\r\n start, end = self.model(**fw_args)\r\nValueError: not enough values to unpack (expected 2, got 1)`\r\n\r\nIt seems like the dictionary of values start and end I'm getting is not a tuple or something like that.", "I updated the code so that it loads a previously saved model\r\n\r\n```python\r\n\r\ntokenizer_ = FlaubertTokenizer.from_pretrained(MODELS)\r\nmodel_ = FlaubertModel.from_pretrained(MODELS)\r\n\r\ndef question_extraction(text, question, model, tokenizer, language=\"French\", verbose=False):\r\n \r\n if language==\"French\":\r\n nlp = pipeline('question-answering', \\\r\n model=model, \\\r\n tokenizer=tokenizer)\r\n else:\r\n nlp=pipeline('question-answering')\r\n\r\n output=nlp({'question':question, 'context': text})\r\n\r\n answer, score = output.answer, output.score \r\n\r\n if verbose==True:\r\n print(\"Q: \", question ,\"\\n\",\\\r\n \"A:\", answer,\"\\n\", \\\r\n \"Confidence (%):\", \"{0:.2f}\".format(str(score*100) )\r\n )\r\n \r\n return answer, score\r\n\r\nif __name__==\"__main__\":\r\n question_=\"Quel est le montant de la garantie?\"\r\n language_=\"French\"\r\n text=\"le montant de la garantie est € 1000\"\r\n\r\n answer, score=question_extraction(text, question_, model_, tokenizer_, language_, verbose= True)\r\n\r\n\r\n```\r\n But now I'm getting an unpacking error:\r\n\r\n```\r\nC:\\...\\NLP\\src>python question_extraction.py\r\nOK\r\nOK\r\nconvert squad examples to features: 100%|████████████████████████████████████████████████| 1/1 [00:00<00:00, 4.66it/s]\r\nadd example index and unique id: 100%|███████████████████████████████████████████████████████████| 1/1 [00:00<?, ?it/s]\r\nTraceback (most recent call last):\r\n File \"question_extraction.py\", line 77, in <module>\r\n answer, score=question_extraction(text, question_, model_, tokenizer_, language_, verbose= True)\r\n File \"question_extraction.py\", line 60, in question_extraction\r\n output=nlp({'question':question, 'context': text})\r\n File \"C:\\...\\transformers\\pipelines.py\", line 818, in __call__\r\n start, end = self.model(**fw_args)\r\nValueError: not enough values to unpack (expected 2, got 1)\r\n\r\n```", "Hi @rcontesti, I've investigated further and found a few issues. First of all, the checkpoint you're trying to load is `fmikaelian/flaubert-base-uncased-squad`, which unfortunately cannot be used by pipelines.\r\n\r\nThis is because this model was fine-tuned with `FlaubertForQuestionAnswering` instead of `FlaubertForQuestionAnsweringSimple`, and only the latter can be used by pipelines. Since it was fine-tuned leveraging a different architecture for the QA head, it, unfortunately, won't be usable by pipelines. The usage example on the [models page](https://huggingface.co/fmikaelian/flaubert-base-uncased-squad) is misleading because of that (cc @fmikaelian).\r\n\r\nUnfortunately, there is no French model that can be used with the pipelines, so you would need to do a custom inference leveraging the model. We don't have any examples showcasing how to leverage `XLNet/XLM/FlaubertForQuestionAnswering`, but it is on our roadmap.", "@LysandreJik many thanks for your answer. It was very clarifying.\r\n\r\nSome follow up questions on my side:\r\n\r\n1. If I use FlaubertForQuestionAnsweringSimple then can I use pipelines? If that is the case would you show me how?\r\n2. Is it also the case that I cannot use CammmBert for QA?\r\n3. I guess that because we have different architectures theres is no quick hack to adapt it to pipelines, am I getting it right?\r\n4. If I were to do custom inferencing, without pipelines and only using pytorch, would you mind showing me the resources to do so?\r\n\r\nMany thanks!!!\r\n\r\n\r\n", "1. You can indeed use `FlaubertForQuestionAnsweringSimple` with pipelines, the issue is that there is currently no model fine-tuned on QA for this model.\r\n2. You could also use the `CamembertForQuestionAnswering` model with pipelines I believe, but unfortunately there is no model fine-tuned on QA for this model either.\r\n3. Indeed, we should add these down the line, but it is not very high on our priority list right now cc @mfuntowicz \r\n4. Yes, I'm currently working on some [examples](https://github.com/huggingface/transformers/pull/2850) that should be merged sometimes today. I'll look into using a `XLNet/XLM/FlaubertForQuestionAnswering` and their differing architecture as well.", "@rcontesti @LysandreJik \r\n\r\nI will fine-tune `FlaubertForQuestionAnsweringSimple` and `CamembertForQuestionAnswering` on French QA in the next days and let you know if we can use the pipeline with those", "@fmikaelian, @LysandreJik \r\n\r\nMany thanks for the help. Eventually I could train it myself, I haven't use Pytorch in a year but if you could point to a good dataset I could do training. Many thanks!", "@rcontesti @LysandreJik \r\n\r\nI fine-tuned `FlaubertForQuestionAnsweringSimple` on [FQuAD](https://fquad.illuin.tech/), by editing `run_squad.py` using the same approach as #2746, but still got `ValueError: not enough values to unpack (expected 2, got 1)` when using the model with a pipeline.\r\n\r\nI also fine-tuned `CamembertForQuestionAnswering` on [FQuAD](https://fquad.illuin.tech/) and [French-SQuAD](https://github.com/Alikabbadj/French-SQuAD), and pipelines are working :-]\r\n\r\n```python3\r\nfrom transformers import pipeline\r\n\r\nnlp = pipeline('question-answering', model='fmikaelian/camembert-base-squad', tokenizer='fmikaelian/camembert-base-squad')\r\n\r\nnlp({\r\n 'question': \"Qui est Claude Monet?\",\r\n 'context': \"Claude Monet, né le 14 novembre 1840 à Paris et mort le 5 décembre 1926 à Giverny, est un peintre français et l’un des fondateurs de l'impressionnisme.\"\r\n})\r\n```\r\n\r\n```\r\n{'answer': 'un peintre français',\r\n 'end': 106,\r\n 'score': 0.498404793881182,\r\n 'start': 87}\r\n```\r\n\r\nModel links:\r\n\r\n- [`fmikaelian/camembert-base-fquad`](https://huggingface.co/fmikaelian/camembert-base-fquad)\r\n- [`fmikaelian/camembert-base-squad`](https://huggingface.co/fmikaelian/camembert-base-squad)\r\n\r\nWill open a PR for models cards (#3089)", "@fmikaelian That's really cool, thanks for taking the time to fine-tune those models! I'll look into the error with the pipeline ASAP, I'm pretty sure I know where it comes from.\r\n\r\nReally cool to have the first community model for question answering in French!", "Hi @fmikaelian \r\n\r\nJust installed transformers from source and it seems the model is still not there\r\n\r\n`Model name 'fmikaelian/camembert-base-squad' was not found in model name list `\r\n\r\nAlso tried to download from S3 but it also does not seem to be there:\r\n\r\n\r\n\r\n`OSError: Model name '../models/fmikaelian/camembert-base-squad' was not found in model name list. We assumed 'https://s3.amazonaws.com/models.huggingface.co/bert/../models/fmikaelian/camembert-base-squad/config.json' was a path, a model identifier, or url to a configuration file named config.json or a directory containing such a file but couldn't find any such file at this path or url.`\r\n\r\nWould you mind sharing the s3 paths? I couldn t get them.", "The models are on the S3. What command did you use? Why is there \"../\" in your model name?\r\n\r\nThe following works:\r\n\r\n```py\r\nfrom transformers import CamembertModel\r\nmodel = CamembertModel.from_pretrained(\"fmikaelian/camembert-base-squad\")\r\n```\r\n\r\nThe following also works:\r\n\r\n```py\r\nfrom transformers import pipeline\r\nnlp = pipeline(\"question-answering\", model=\"fmikaelian/camembert-base-squad\", tokenizer=\"fmikaelian/camembert-base-squad\")\r\n\r\n```", "@LysandreJik, is working now. Many thanks!" ]

[ { "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/2892?src=pr&el=h1) Report\n> Merging [#2892](https://codecov.io/gh/huggingface/transformers/pull/2892?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/2ae98336d17fceea7506af9880b862b6252a38f6?src=pr&el=desc) will **decrease** coverage by `1.07%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2892?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2892 +/- ##\n==========================================\n- Coverage 75.06% 73.98% -1.08% \n==========================================\n Files 94 94 \n Lines 15288 15288 \n==========================================\n- Hits 11476 11311 -165 \n- Misses 3812 3977 +165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2892?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2892/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2892/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.91% <0%> (-9.86%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2892/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `70.86% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2892/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `92.07% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2892/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2892?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/2892?src=pr&el=footer). Last update [2ae9833...25c0467](https://codecov.io/gh/huggingface/transformers/pull/2892?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Thanks for sharing @BinWang28!", "[model page](https://huggingface.co/binwang/xlnet-base-cased)" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2891?src=pr&el=h1) Report\n> Merging [#2891](https://codecov.io/gh/huggingface/transformers/pull/2891?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/2ae98336d17fceea7506af9880b862b6252a38f6?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2891?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2891 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15288 15288 \n=======================================\n Hits 11476 11476 \n Misses 3812 3812\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2891?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/data/processors/utils.py](https://codecov.io/gh/huggingface/transformers/pull/2891/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL3Byb2Nlc3NvcnMvdXRpbHMucHk=) | `21.73% <ø> (ø)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2891?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/2891?src=pr&el=footer). Last update [2ae9833...e0b3974](https://codecov.io/gh/huggingface/transformers/pull/2891?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Thanks!", "Thank you! Easiest review ever! Keep em coming :)" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2890?src=pr&el=h1) Report\n> Merging [#2890](https://codecov.io/gh/huggingface/transformers/pull/2890?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/0dbddba6d2c5b2c6fc08866358c1994a00d6a1ff?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2890?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2890 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15288 15288 \n=======================================\n Hits 11476 11476 \n Misses 3812 3812\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2890?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/2890?src=pr&el=footer). Last update [0dbddba...8f8137f](https://codecov.io/gh/huggingface/transformers/pull/2890?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "@LysandreJik \r\nThis implementation does not work with `pytorch_lightning` > `0.7.1`.\r\n\r\nIt throws the exception `'Trainer' object has no attribute 'avg_loss'` because since version `0.7.2` they removed the `avg_loss` field from the `Trainer` class.\r\n\r\nSee https://github.com/huggingface/transformers/pull/2890/files#diff-d68a6ecfacd8231c59af0ea67d77bb9cR120", "@simonepri can you file an issue? I guess we should just remove that key/value pair. ", "@simonepri did you try 0.7.5? " ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834053813, "node_id": "MDU6TGFiZWwxODM0MDUzODEz", "url": "https://api.github.com/repos/huggingface/transformers/labels/PyTorch", "name": "PyTorch", "color": "a12bef", "default": false, "description": "Anything PyTorch" }, { "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." } ]

[]

[ "Please fix the formatting of your post and use code tags.", "I made the changes still all the text is shown struck off form.\r\nI am new to this bug log not sure how to change to code tag \r\n\r\n", "I have used <code> tag ", "Read how to use tags here: https://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks", "I did the tags as suggested by BramVanroy by using guidelines::https://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks", "You clearly did something wrong because, as you can see yourself, all text is striked through. Likely caused by having tildes (~) around your post.", "Thanks, I cleared it, there was one hiding beside a comment.", "You are using an old version of the library (pytorch_pretrained_bert). You should move to `transformers` instead.", "I upgraded latest ``` transformers ``` still I am getting following error message :\r\n```\r\nERROR:root:Internal Python error in the inspect module.\r\nBelow is the traceback from this internal error.\r\n\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 3319, in run_code\r\n exec(code_obj, self.user_global_ns, self.user_ns)\r\n File \"<ipython-input-13-645c7873d473>\", line 1, in <module>\r\n encoding = tokenizer.encode(raw_text)\r\nAttributeError: 'BertTokenizer' object has no attribute 'encode'\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 2034, in showtraceback\r\n stb = value._render_traceback_()\r\nAttributeError: 'AttributeError' object has no attribute '_render_traceback_'\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\pywrap_tensorflow.py\", line 58, in <module>\r\n from tensorflow.python.pywrap_tensorflow_internal import *\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\pywrap_tensorflow_internal.py\", line 28, in <module>\r\n _pywrap_tensorflow_internal = swig_import_helper()\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\pywrap_tensorflow_internal.py\", line 24, in swig_import_helper\r\n _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\imp.py\", line 242, in load_module\r\n return load_dynamic(name, filename, file)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\imp.py\", line 342, in load_dynamic\r\n return _load(spec)\r\nImportError: DLL load failed: The specified module could not be found.\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\site-packages\\IPython\\core\\ultratb.py\", line 1151, in get_records\r\n return _fixed_getinnerframes(etb, number_of_lines_of_context, tb_offset)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\site-packages\\IPython\\core\\ultratb.py\", line 319, in wrapped\r\n return f(*args, **kwargs)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\site-packages\\IPython\\core\\ultratb.py\", line 353, in _fixed_getinnerframes\r\n records = fix_frame_records_filenames(inspect.getinnerframes(etb, context))\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\inspect.py\", line 1502, in getinnerframes\r\n frameinfo = (tb.tb_frame,) + getframeinfo(tb, context)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\inspect.py\", line 1460, in getframeinfo\r\n filename = getsourcefile(frame) or getfile(frame)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\inspect.py\", line 696, in getsourcefile\r\n if getattr(getmodule(object, filename), '__loader__', None) is not None:\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\inspect.py\", line 733, in getmodule\r\n if ismodule(module) and hasattr(module, '__file__'):\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow\\__init__.py\", line 50, in __getattr__\r\n module = self._load()\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow\\__init__.py\", line 44, in _load\r\n module = _importlib.import_module(self.__name__)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\importlib\\__init__.py\", line 127, in import_module\r\n return _bootstrap._gcd_import(name[level:], package, level)\r\n File \"<frozen importlib._bootstrap>\", line 1006, in _gcd_import\r\n File \"<frozen importlib._bootstrap>\", line 983, in _find_and_load\r\n File \"<frozen importlib._bootstrap>\", line 953, in _find_and_load_unlocked\r\n File \"<frozen importlib._bootstrap>\", line 219, in _call_with_frames_removed\r\n File \"<frozen importlib._bootstrap>\", line 1006, in _gcd_import\r\n File \"<frozen importlib._bootstrap>\", line 983, in _find_and_load\r\n File \"<frozen importlib._bootstrap>\", line 967, in _find_and_load_unlocked\r\n File \"<frozen importlib._bootstrap>\", line 677, in _load_unlocked\r\n File \"<frozen importlib._bootstrap_external>\", line 728, in exec_module\r\n File \"<frozen importlib._bootstrap>\", line 219, in _call_with_frames_removed\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\__init__.py\", line 42, in <module>\r\n from . _api.v2 import audio\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\_api\\v2\\audio\\__init__.py\", line 10, in <module>\r\n from tensorflow.python.ops.gen_audio_ops import decode_wav\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\ops\\gen_audio_ops.py\", line 9, in <module>\r\n from tensorflow.python import pywrap_tensorflow as _pywrap_tensorflow\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow\\__init__.py\", line 50, in __getattr__\r\n module = self._load()\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow\\__init__.py\", line 44, in _load\r\n module = _importlib.import_module(self.__name__)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\importlib\\__init__.py\", line 127, in import_module\r\n return _bootstrap._gcd_import(name[level:], package, level)\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\__init__.py\", line 49, in <module>\r\n from tensorflow.python import pywrap_tensorflow\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\pywrap_tensorflow.py\", line 74, in <module>\r\n raise ImportError(msg)\r\nImportError: Traceback (most recent call last):\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 3319, in run_code\r\n exec(code_obj, self.user_global_ns, self.user_ns)\r\n File \"<ipython-input-13-645c7873d473>\", line 1, in <module>\r\n encoding = tokenizer.encode(raw_text)\r\nAttributeError: 'BertTokenizer' object has no attribute 'encode'\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 2034, in showtraceback\r\n stb = value._render_traceback_()\r\nAttributeError: 'AttributeError' object has no attribute '_render_traceback_'\r\n\r\nDuring handling of the above exception, another exception occurred:\r\n\r\nTraceback (most recent call last):\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\pywrap_tensorflow.py\", line 58, in <module>\r\n from tensorflow.python.pywrap_tensorflow_internal import *\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\pywrap_tensorflow_internal.py\", line 28, in <module>\r\n _pywrap_tensorflow_internal = swig_import_helper()\r\n File \"C:\\Users\\Veeresh\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\pywrap_tensorflow_internal.py\", line 24, in swig_import_helper\r\n _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\imp.py\", line 242, in load_module\r\n return load_dynamic(name, filename, file)\r\n File \"C:\\Users\\Veeresh\\Anaconda3\\lib\\imp.py\", line 342, in load_dynamic\r\n return _load(spec)\r\nImportError: DLL load failed: The specified module could not be found.\r\n\r\n\r\nFailed to load the native TensorFlow runtime.\r\n\r\nSee https://www.tensorflow.org/install/errors\r\n\r\nfor some common reasons and solutions. Include the entire stack trace\r\nabove this error message when asking for help.\r\n---------------------------------------------------------------------------\r\n```", "There's a lot going wrong in that trace. Please recreate your environment from scratch to ensure that all correct dependencies are installed. Particularly, in your first post you were using torch, but your new trace throws Tensorflow errors.", "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" ]

[]

[]

[ "Excited for this!\r\n\r\nA little early for me to have an opinion, but I'd start by adding a bunch of failing tests (e.g. for t5.generate), and some slow tests that verify that T5.generate/another non seq2seq model generate reasonable results. (You have to run those locally). \r\n\r\nStylistically, I'd say `is_seq_to_seq` should probably not be a function, just an attribute. But I think style here much less important than test coverage :)\r\n\r\nBon Chance!", "Meant to just comment, sorry!", "I adapted the language generation according to the newly added Bart file. This is still very much work in progress that's why I left a lot of comments in all files. Would be very happy about some feedback! @sshleifer ", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2888?src=pr&el=h1) Report\n> Merging [#2888](https://codecov.io/gh/huggingface/transformers/pull/2888?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/fc38d4c86fe4bbde91b194880fe38b821a346123?src=pr&el=desc) will **decrease** coverage by `33.74%`.\n> The diff coverage is `2.38%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2888?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2888 +/- ##\n===========================================\n- Coverage 77.12% 43.37% -33.75% \n===========================================\n Files 98 98 \n Lines 15975 15995 +20 \n===========================================\n- Hits 12320 6938 -5382 \n- Misses 3655 9057 +5402\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2888?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19ncHQyLnB5) | `0% <0%> (-86.12%)` | :arrow_down: |\n| [src/transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==) | `0% <0%> (-92.38%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `0% <0%> (-75.78%)` | :arrow_down: |\n| [src/transformers/modeling\\_bart.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19iYXJ0LnB5) | `0% <0%> (-84.69%)` | :arrow_down: |\n| [src/transformers/modeling\\_transfo\\_xl.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190cmFuc2ZvX3hsLnB5) | `0% <0%> (-75.64%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `0% <0%> (-98.24%)` | :arrow_down: |\n| [src/transformers/configuration\\_bart.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX2JhcnQucHk=) | `36.36% <0%> (-63.64%)` | :arrow_down: |\n| [src/transformers/configuration\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX3V0aWxzLnB5) | `96.49% <100%> (+0.03%)` | :arrow_up: |\n| [src/transformers/modeling\\_xlm\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG1fcm9iZXJ0YS5weQ==) | `0% <0%> (-100%)` | :arrow_down: |\n| [src/transformers/modeling\\_camembert.py](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jYW1lbWJlcnQucHk=) | `0% <0%> (-100%)` | :arrow_down: |\n| ... and [29 more](https://codecov.io/gh/huggingface/transformers/pull/2888/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2888?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/2888?src=pr&el=footer). Last update [fc38d4c...ab13956](https://codecov.io/gh/huggingface/transformers/pull/2888?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[ { "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": 1834056761, "node_id": "MDU6TGFiZWwxODM0MDU2NzYx", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Modeling", "name": "Core: Modeling", "color": "FF8446", "default": false, "description": "Internals of the library; Models." } ]

[]

[ "Hi,\r\nwhich model do you use? Can't you simply remove the output of the other heads?", "I am using TFDistilbertmodelforsequenceclassification.\r\nYou are right,i can remove other head attention,but while using tfserving,it's taking a lot time ,since the output attentions has huge dimension.\r\nSo,that's why ,i was asking if i could get just last 2 head attention instead of getting all head attentions while hitting tfserving", "Unfortunately, we don't have a way to do that right now." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834053813, "node_id": "MDU6TGFiZWwxODM0MDUzODEz", "url": "https://api.github.com/repos/huggingface/transformers/labels/PyTorch", "name": "PyTorch", "color": "a12bef", "default": false, "description": "Anything PyTorch" }, { "id": 1834056761, "node_id": "MDU6TGFiZWwxODM0MDU2NzYx", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Modeling", "name": "Core: Modeling", "color": "FF8446", "default": false, "description": "Internals of the library; Models." } ]

[]

[ "```\r\nclass RoBertaMultiwayMatch(nn.Module):\r\n def __init__(self, pretrainedConfigName, num_choices=4):\r\n super(RoBertaMultiwayMatch, self).__init__()\r\n self.num_choices = num_choices\r\n self.RoBerta = RobertaModel.from_pretrained(pretrainedConfigName)\r\n config = self.RoBerta.config\r\n self.dropout = nn.Dropout(config.hidden_dropout_prob)\r\n self.linear_trans = nn.Linear(config.hidden_size, config.hidden_size)\r\n self.linear_fuse_p = nn.Linear(config.hidden_size*2, config.hidden_size)\r\n self.linear_fuse_q = nn.Linear(config.hidden_size*2, config.hidden_size)\r\n self.linear_fuse_a = nn.Linear(config.hidden_size * 2, config.hidden_size)\r\n self.classifier = nn.Linear(config.hidden_size*3, 1)\r\n\r\n #def matching(self, passage_encoded, question_encoded, passage_attention_mask, question_attention_mask): ...\r\n #def fusing_mlp(self, passage_encoded, mp_q, mp_a, mp_qa, question_encoded, ...\r\n #def forward(self, input_ids, token_type_ids=None, attention_mask=None, doc_len=None, ...\r\n```\r\n", "You're inheriting from `BertPreTrainedModel` in your `RoBertaMultiwayMatch`, with an attribute `RoBerta` which contains a `RobertaModel`.\r\n\r\nAs I see it, you want to load your roberta model from a given set of weights, but by calling `from_pretrained` on your class, he's looking to load those weights directly on your model.\r\n\r\nI believe you could override the `from_pretrained` method as such:\r\n\r\n```py\r\ndef from_pretrained(...):\r\n self.RoBerta.from_pretrained(...)\r\n```\r\n\r\nby specifying the correct arguments to your method and to your `RoBerta`'s method. This way when you call `from_pretrained`, it only loads the data for `RoBerta`. You'd have to find a way to save/load your data for your own layers though.", "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/2885?src=pr&el=h1) Report\n> Merging [#2885](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/d490b5d5003654f104af3abd0556e598335b5650?src=pr&el=desc) will **increase** coverage by `1.75%`.\n> The diff coverage is `86.22%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2885 +/- ##\n=========================================\n+ Coverage 75.35% 77.1% +1.75% \n=========================================\n Files 94 98 +4 \n Lines 15444 15971 +527 \n=========================================\n+ Hits 11638 12315 +677 \n+ Misses 3806 3656 -150\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19kaXN0aWxiZXJ0LnB5) | `97.59% <ø> (-0.04%)` | :arrow_down: |\n| [src/transformers/tokenization\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fYmVydC5weQ==) | `96.99% <ø> (ø)` | :arrow_up: |\n| [src/transformers/tokenization\\_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fZ3B0Mi5weQ==) | `96.82% <ø> (-0.03%)` | :arrow_down: |\n| [src/transformers/modeling\\_encoder\\_decoder.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19lbmNvZGVyX2RlY29kZXIucHk=) | `26.66% <ø> (+1.36%)` | :arrow_up: |\n| [src/transformers/data/processors/utils.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL3Byb2Nlc3NvcnMvdXRpbHMucHk=) | `21.73% <ø> (ø)` | :arrow_up: |\n| [src/transformers/utils\\_encoder\\_decoder.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy91dGlsc19lbmNvZGVyX2RlY29kZXIucHk=) | `0% <0%> (ø)` | |\n| [src/transformers/modeling\\_auto.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19hdXRvLnB5) | `75.47% <100%> (+0.23%)` | :arrow_up: |\n| [src/transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==) | `92.2% <100%> (+30.87%)` | :arrow_up: |\n| [src/transformers/tokenization\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHMucHk=) | `90.45% <100%> (-0.14%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `95.71% <100%> (-0.07%)` | :arrow_down: |\n| ... and [18 more](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2885?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/2885?src=pr&el=footer). Last update [d490b5d...80ca73d](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "> Love the tests!\r\n> \r\n> Do we ever test cases:\r\n> \r\n> * eos_token_ids are None, pad_token_id present\r\n> * pad_token_id=None, eos_ids present\r\n> * pad_token_id present, eos_ids None\r\n> \r\n> We should also figure out if those are realistic scenarios. Because if they are not we can delete a lot of code!\r\n\r\neos_token_ids present and pad_token_id = None -> GPT2, so this scenario is tested. This case is the hardest to handle for batch_size > 1, therefore quite a lot of assert statements and a warning in modeling_utils.py\r\n\r\neos_token_ids = None and pad_token_id present -> If eos_token_ids = None, the pad_token_id is somewhat irrelevant for generation because all batches will always be of max. length and therefore never will have to be padded (they can't finish because no eos_token can be generated)\r\n\r\n", "This PR finally implements the following `bos_token_id, pad_token_id, eos_token_ids` logic for lm model generation.\r\n\r\n1. If `bos_token_id` is None, then the input_ids must be defined otherwise, the model cannot generate text, which is checked by the asserts in the beginning. The `bos_token_id` is only relevant for starting a new sentence.\r\n\r\n2. If `eos_token_id` is None, then the length of the generated text will always equal max_length,\r\nno matter how the pad_token_id is defined. Since there is no `eos_token_id` the text will also not \"end\".\r\n\r\n3. If `pad_token_id` is None and `eos_token_ids` is defined (as it is the case for gpt2), then the pad_token_id will be set to the `eos_token_ids[0]` tensor `batches_len` is used to keep track of the first time the sequence generated an eos_token and will later set all tokens following this token to the `pad_token_id`, which is `eos_token_ids[0]` and can thus be handled by the tokenizer (whereas the -1 cannot be handled by the tokenizer).\r\n\r\n4. **No** eos_token_id is appended to sentences that finish due to `max_length`. Instead those sentences are returned with the last token being the last token produced by the model until `max_length` was hit.\r\n\r\nAs an overview, here a table showing which LMModel Tokenizer have which of the tokens `bos_token_id`, `pad_token_id` and `eos_token_ids` is defined:\r\n\r\n\r\n| LM Model | bos_token_id | pad_token_id | eos_token_ids |\r\n| ------------- | ------------- | ------------- | ------------- | \r\n| XLNet | x | x | x |\r\n| OpenAIGPT | o | o | o |\r\n| CTRL | o | o | o |\r\n| GPT2 | x | o | x |\r\n| Transfo-XL | o | o | x |\r\n| XLM | x | x | o |\r\n\r\n## Future PRs:\r\n\r\n- [x] [WIP] adding hard-coded slow tests for pretrained lms in PR #2909\r\n- [ ] [WIP] adapting the `generate` function for Seq-2-Seq and DoubleHeads or other special LM models in PR #2888 \r\n- [x] checking and possibly adapting behavior of `generate_beam_search`\r\n- [x] treat Issues: #2482 and #2415", "> # [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=h1) Report\r\n> > Merging [#2885](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/59c23ad9c931ac4fe719abeb3c3851df046ef3a6?src=pr&el=desc) will **increase** coverage by `1.49%`.\r\n> > The diff coverage is `100%`.\r\n> \r\n> [](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=tree)\r\n> \r\n> ```diff\r\n> @@ Coverage Diff @@\r\n> ## master #2885 +/- ##\r\n> ==========================================\r\n> + Coverage 75.3% 76.79% +1.49% \r\n> ==========================================\r\n> Files 94 94 \r\n> Lines 15424 15448 +24 \r\n> ==========================================\r\n> + Hits 11615 11864 +249 \r\n> + Misses 3809 3584 -225\r\n> ```\r\n> \r\n> [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=tree)\tCoverage Δ\t\r\n> [src/transformers/modeling_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ191dGlscy5weQ==)\t`92.14% <100%> (+30.81%)`\t\r\n> [src/transformers/configuration_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX3V0aWxzLnB5)\t`96.46% <100%> (ø)`\t\r\n> [src/transformers/modeling_transfo_xl.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190cmFuc2ZvX3hsLnB5)\t`75.63% <0%> (+0.84%)`\t\r\n> [src/transformers/modeling_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG0ucHk=)\t`88.43% <0%> (+2.05%)`\t\r\n> [src/transformers/modeling_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==)\t`75.77% <0%> (+2.61%)`\t\r\n> [src/transformers/modeling_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19ncHQyLnB5)\t`86.11% <0%> (+2.83%)`\t\r\n> [src/transformers/modeling_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2885/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5)\t`98.23% <0%> (+3.96%)`\t\r\n> [Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=continue).\r\n> \r\n> > **Legend** - [Click here to learn more](https://docs.codecov.io/docs/codecov-delta)\r\n> > `Δ = absolute <relative> (impact)`, `ø = not affected`, `? = missing data`\r\n> > Powered by [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=footer). Last update [59c23ad...ac2e172](https://codecov.io/gh/huggingface/transformers/pull/2885?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\r\n\r\nI'm a bit confused that the coverage for the file modeling_openai.py did not change. Tests for the openai lm were added but they had seemingly no effect - do you know why? @LysandreJik ", "No need to do all my `torch.Tensor` vs .new comments, TIL that it lets you copy the device and dtype of the first tensor." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834052574, "node_id": "MDU6TGFiZWwxODM0MDUyNTc0", "url": "https://api.github.com/repos/huggingface/transformers/labels/Ex:%20Sequence%20Classification", "name": "Ex: Sequence Classification", "color": "46FFCF", "default": false, "description": "" } ]

[]

[ "Hi! You can already provide `do_eval` to `run_glue` to do the evaluation. If you don't specify `do_train`, it will only do the evaluation, and no training.\r\n\r\nThe inference would be a nice addition.", "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" } ]

[]

[ "👍 " ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834053813, "node_id": "MDU6TGFiZWwxODM0MDUzODEz", "url": "https://api.github.com/repos/huggingface/transformers/labels/PyTorch", "name": "PyTorch", "color": "a12bef", "default": false, "description": "Anything PyTorch" }, { "id": 1834060867, "node_id": "MDU6TGFiZWwxODM0MDYwODY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/Ex:%20Named%20Entity%20Recognition", "name": "Ex: Named Entity Recognition", "color": "06FFD8", "default": false, "description": "" } ]

[]

[ "I tried to make the predictions using CPU, and it worked just fine. But the predictions made by CPU is totally different from the predictions made by GPU. Isn't a model supposed to give the same predictions irrespective of whether it is loaded in GPU or CPU?\r\n\r\nAny help would be appreciated.\r\nThanks in advance", "Can you post a short reproducibility case?", "@cibinjohn \r\n\r\nThe error msg \"Maximum sequence length exceeded\" indicated that the input length (assume 257) was longer than the max_seq_len parameter (assume 256). In this case, the last token will not be predicted (actually be trimmed during the preprocessing). You have to reduce the length of your input or increase max_seq_len whichever works for you.\r\n\r\nFor the question of difference results between CPU and GPU, I cannot repeat a case similar to yours.", "@cibinjohn \r\nI have been facing the same issue yesterday while testing on my own database. I tried doubling the max_seq_length value which was before equal to the MAX_LENGTH env variable. It worked for me.\r\n", "If any of you can post a short reproducible example, we can look into this.", "@BramVanroy what I observed is that the prediction list has a list of different sizes which depends upon the length of the row data before every empty line in test.txt which is created after pre-processing using preprocess.py. \r\nSo, I kept MAX_LENGTH as 128 and max_sequence length in the range of 190-256 and it worked.\r\nAccording to me, it might be because of the length of the word_tokens list or maybe because of the length of the row data before every empty line of the test.txt file.\r\n[test (5).txt](https://github.com/huggingface/transformers/files/4255447/test.5.txt)\r\nIf for this text file if you keep max_sequence_length as 128 then it will show Maximum_sequence_length_exceeded for around 400 tokens.", " @BramVanroy \r\n\r\n%%capture\r\n!pip install -qU transformers==2.4\r\n!pip install -qU pytorch-lightning\r\n!git clone --branch fixlight https://github.com/srush/transformers\r\n!pip install -r transformers/examples/requirements.txt\r\n\r\n%%capture\r\n%%bash\r\ncd transformers/examples/ner/\r\nwget \"https://raw.githubusercontent.com/stefan-it/fine-tuned-berts-seq/master/scripts/preprocess.py\"\r\nexport MAX_LENGTH=128\r\nexport BERT_MODEL=bert-base-multilingual-cased\r\npython3 preprocess.py train.txt.tmp $BERT_MODEL $MAX_LENGTH > train.txt\r\npython3 preprocess.py dev.txt.tmp $BERT_MODEL $MAX_LENGTH > dev.txt\r\npython3 preprocess.py test.txt.tmp $BERT_MODEL $MAX_LENGTH > test.txt\r\ncat train.txt dev.txt test.txt | cut -d \" \" -f 2 | grep -v \"^$\"| sort | uniq > labels.txt\r\n\r\n\r\n!cd transformers/examples/ner/; \\\r\nexport MAX_LENGTH=190; \\\r\nexport BERT_MODEL=bert-base-multilingual-cased; \\\r\nexport OUTPUT_DIR=germeval-model; \\\r\nexport BATCH_SIZE=32; \\\r\nexport NUM_EPOCHS=3; \\\r\nexport SAVE_STEPS=750; \\\r\nexport SEED=42; \\\r\npython3 run_ner.py --data_dir ./ \\\r\n--model_type bert \\\r\n--labels ./labels.txt \\\r\n--model_name_or_path $BERT_MODEL \\\r\n--output_dir $OUTPUT_DIR \\\r\n--max_seq_length $MAX_LENGTH \\\r\n--num_train_epochs $NUM_EPOCHS \\\r\n--per_gpu_train_batch_size $BATCH_SIZE \\\r\n--save_steps $SAVE_STEPS \\\r\n--seed $SEED \\\r\n--do_train \\\r\n--do_eval \\\r\n--do_predict", "I met this question last day ，and i checked all cases but nothing has gone wrong。\r\nso，i made a new dir，then just put ‘train.txt’、“test.txt”、“dev.txt”、“labels.txt” there， start；\r\n\r\nthen，all is ok。\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" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2881?src=pr&el=h1) Report\n> Merging [#2881](https://codecov.io/gh/huggingface/transformers/pull/2881?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/6083c1566e261668a5de73cfe484c171ce232812?src=pr&el=desc) will **decrease** coverage by `1.07%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2881?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2881 +/- ##\n==========================================\n- Coverage 75.06% 73.98% -1.08% \n==========================================\n Files 94 94 \n Lines 15288 15288 \n==========================================\n- Hits 11476 11311 -165 \n- Misses 3812 3977 +165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2881?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2881/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2881/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.91% <0%> (-9.86%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2881/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `70.86% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2881/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `92.07% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2881/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2881?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/2881?src=pr&el=footer). Last update [6083c15...fb4d8d0](https://codecov.io/gh/huggingface/transformers/pull/2881?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Welcome @patrickvonplaten :)" ]

[]

[]

[]

[ { "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/2879?src=pr&el=h1) Report\n> Merging [#2879](https://codecov.io/gh/huggingface/transformers/pull/2879?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/6083c1566e261668a5de73cfe484c171ce232812?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2879?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2879 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15288 15288 \n=======================================\n Hits 11476 11476 \n Misses 3812 3812\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2879?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/2879?src=pr&el=footer). Last update [6083c15...d18f775](https://codecov.io/gh/huggingface/transformers/pull/2879?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Looks good: https://huggingface.co/dbmdz/bert-base-turkish-cased" ]

[]

[]

[ "Sounds like a [sentencepiece](https://github.com/google/sentencepiece) issue?", "I have an inquiry there as well since just a straight install of sentencepiece gives same results - I was just hoping there might be a way of using the HuggingFace Transformers without sentencepiece (although the name sentencepiece suggests to me that it may perform some critical functions...) something like \"pip install transformers --skip sentencepiece\" ? :)\r\n", "@DaveXanatos Even I faced same issues. Issue seems to be with pip package of sentencepiece. I have opened an issue with them.\r\n\r\nAs a word around, I installed sentence piece from conda and it worked. After installing this you can install transformer.\r\n\r\n`conda install -c powerai sentencepiece`\r\n", "Be wary of using conda and pip at the same time, if you don't know _exactly_ what you are doing, this will lead to unexpected complications.\r\n\r\nSentencePiece is required for most recent models, so it is a hard dependency. I advise you to just wait until this is solved in the sentencepiece library, or download and install an earlier version of their library, e.g. https://github.com/google/sentencepiece/releases/tag/v0.1.84", "@tkhan3 Thanks for the conda possibility, I will look into that in the interim.\r\n\r\n@BramVanroy I have heard this before and a couple of years ago I completely hosed my build doing just this :) Where would you suggest, as the most direct route to understanding exactly the differences between pip installs and conda installs in terms of paths, dependencies, etc., such that I could conda install with confidence a package in an otherwise pip installed environment?", "I'm afraid I cannot help with that. I stay away from conda as much as I can. Pipenv is my main driver, falling back to an environment's pip where necessary.", "@BramVanroy I'm with you on that from my experience, although I know some folks swear by it... thanks for the warning. I'll probably flash a backup image and then go and play with the possibilities and see if I can get it to work... I can always reflash back to the backup if I break everything again. If I have success I'll let you know what I found.", "Closing this. I propose that the discussion is moved to the sentencepiece library. https://github.com/google/sentencepiece/issues/452", "I got a similar issue.\r\nWhen I install sentence-transformers on Linux by python, I got an error message: \r\n\r\nERROR: Could not find a version that satisfies the requirement transformers>=3.0.2 (from sentence-transformers) (fr\r\nom versions: 0.1, 2.0.0, 2.1.0, 2.1.1, 2.2.0, 2.2.1, 2.2.2, 2.3.0, 2.4.0, 2.4.1, 2.5.0, 2.5.1)\r\nERROR: No matching distribution found for transformers>=3.0.2 (from sentence-transformers)\r\n\r\nsystem: Linux 4.9.0-12-amd64 #1 SMP Debian 4.9.210-1+deb9u1 (2020-06-07) x86_64 on GCP VM instance.\r\nIs there any suggestion?\r\n", "I had the same issue but on Unbuntu 20.04.1.\r\nMy problem was that i used a pip version to old to install sentencepiece, as it requires pip>=19.3. (https://github.com/google/sentencepiece/issues/572#issuecomment-716890916)\r\nSo my solution was to upgrade my pip installation to 20.2.4\r\n$ pip install --upgrade pip\r\n\r\nSimilar issues has been discussed here https://github.com/google/sentencepiece/issues/572\r\n\r\nHope it helps" ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834053007, "node_id": "MDU6TGFiZWwxODM0MDUzMDA3", "url": "https://api.github.com/repos/huggingface/transformers/labels/Ex:%20LM%20(Pretraining)", "name": "Ex: LM (Pretraining)", "color": "76FFAF", "default": false, "description": "Related to language modeling pre-training" }, { "id": 1834053813, "node_id": "MDU6TGFiZWwxODM0MDUzODEz", "url": "https://api.github.com/repos/huggingface/transformers/labels/PyTorch", "name": "PyTorch", "color": "a12bef", "default": false, "description": "Anything PyTorch" } ]

[]

[ "I ran into this with my own dataset. Following some discussion in #1538 I changed truncation to 256.\r\n\r\n> tokenizer.enable_truncation(max_length=256)\r\n\r\nI also had to make sure that the pad token had index 1, as that seems to be hardcoded in roberta.\r\n\r\nThis appears to work, though the previous error had only appeared fairly deep into training, and reducing context isn't great. So I'm not satisfied by this hack. ", "If you want to see the error clearer, switch it to CPU, then it will print out the real error. I rant into this error in another project, and finally, I found out it is basically `index out of range` error. Fixed it by add some missing words to the vocabulary.txt and resize the model itself.", "> I ran into this with my own dataset. Following some discussion in #1538 I changed truncation to 256.\r\n> \r\n> > tokenizer.enable_truncation(max_length=256)\r\n> \r\n> I also had to make sure that the pad token had index 1, as that seems to be hardcoded in roberta.\r\n> \r\n> This appears to work, though the previous error had only appeared fairly deep into training, and reducing context isn't great. So I'm not satisfied by this hack.\r\n@reidsanders were u able to train language model with ur own dataset eventually?\r\n@binhna have u added missing words inplaceon unk or randomly added them in vocab, ?\r\n", "Hello @reidsanders, @samreenkazi \r\n\r\nI encountered the same error and tried `tokenizer.enable_truncation(max_length=256)` on some BERT models. But it seems that there is no such method:\r\n`'PreTrainedTokenizer' object has no attribute 'enable_truncation'`\r\n\r\nCould you give more details about how you solved the problem?", "I am unable to solve this problem as yet\n\nOn Thu, Mar 5, 2020 at 10:22 PM Jinan Zhou <[email protected]> wrote:\n\n> Hello @reidsanders <https://github.com/reidsanders>, @samreenkazi\n> <https://github.com/samreenkazi>\n>\n> I encountered the same error tried\n> tokenizer.enable_truncation(max_length=256) on some BERT models. But it\n> seems that there is no such method:\n> 'PreTrainedTokenizer' object has no attribute 'enable_truncation'\n>\n> Could you give more details about how you solved the problem?\n>\n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/transformers/issues/2877?email_source=notifications&email_token=ALPPGB4ZSN6BT5PXL6QRWWTRF7NU5A5CNFSM4KWFP7FKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEN6ETGY#issuecomment-595347867>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/ALPPGBYDQHHQTNNOESJOZ33RF7NU5ANCNFSM4KWFP7FA>\n> .\n>\n", "> Hello @reidsanders, @samreenkazi\r\n> \r\n> I encountered the same error and tried `tokenizer.enable_truncation(max_length=256)` on some BERT models. But it seems that there is no such method:\r\n> `'PreTrainedTokenizer' object has no attribute 'enable_truncation'`\r\n> \r\n> Could you give more details about how you solved the problem?\r\n\r\nenable_truncation is not a method in PreTrainedTokenizers (we are training from scratch, not pretrained). I'm using ByteLevelBPETokenizer imported from tokenizers module as in the example blog post (and op).", "Thanks for your comment @reidsanders @samreenkazi \r\n\r\nAccording to my observation, the error is indeed caused by the data samples longer than 512 tokens. A conservative solution is fixing `block_size=512` in `TextDataset` and `LineByLineTextDataset` class. Or if you worry about the truncation of data, you can go through `self.examples` in these two classes, check whether each sample is shorter than 512. If not, split it into multiple chunks with length 512. \r\n ", "For me it helped to just specify the argument `--block_size=512` (it's -1 by default)", "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" } ]

[]

[ "The file path should be `model_cards/mrm8488/bert-spanish-cased-finedtuned-ner/README.md` @mrm8488 " ]

[ { "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/2875?src=pr&el=h1) Report\n> Merging [#2875](https://codecov.io/gh/huggingface/transformers/pull/2875?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/73028c5df0c28ca179fbe565482a9c2143787f61?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2875?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2875 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15288 15288 \n=======================================\n Hits 11476 11476 \n Misses 3812 3812\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2875?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/2875?src=pr&el=footer). Last update [73028c5...cb1cba9](https://codecov.io/gh/huggingface/transformers/pull/2875?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Thanks for updating! GIFs on Giphy are a bit over-compressed so you can also host them \r\n- on one of your GitHub repos\r\n- or even directly in the model card's folder (see severinsimmler/literary-german-bert/README.md as an example)", "Also, great results :)", "Thank you!!" ]

[]

[]

[ "TFBERT model can only be loaded in tf>2.0 version.And,tf 2.0 onwards,eager_execution() is on by default" ]

[]

[]

[ "Just click on 'List all files in model' and you will see the upload date [1].\r\n\r\n[1] https://huggingface.co/xlnet-base-cased" ]

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

[]

[ "@gofimofi You might want to look at https://github.com/jessevig/bertviz which is compatible with transformers", "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": 1834056761, "node_id": "MDU6TGFiZWwxODM0MDU2NzYx", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Modeling", "name": "Core: Modeling", "color": "FF8446", "default": false, "description": "Internals of the library; Models." } ]

[]

[ "This is something that we're looking at with @LysandreJik and @thomwolf – In the meantime, feel free to open a draft PR.", "As discussed in the other issues, it would be great if a lot of care is taken in maximising the compatibility between a tokenizer and its corresponding model, as I discussed in https://github.com/huggingface/transformers/issues/2702#issuecomment-581480669. In other words, the tokenizer encode methods should only return those values that are accepted by its model's forward method.", "RoBERTa has only one possible token type id (`0`), but the embedding for that still needs to be there. That embedding is added to all word piece embeddings, and that's how it was trained. If you suddenly stop doing that, the model will stop working.\r\n\r\nYou could just add that embedding to all word piece embeddings and store the model under a new name to achieve the same effect. But you can't just take it away.\r\n\r\nIf you take it away, you have to retrain the whole thing.", "In my opinion you dont have to retrain it as the weights of this layer are zero. Have a look at the example below:\r\n```from transformers.modeling_roberta import RobertaForSequenceClassification\r\nmodel = RobertaForSequenceClassification.from_pretrained('roberta-base')\r\nprint(model.state_dict()['roberta.embeddings.token_type_embeddings.weight'])\r\n##Output truncated by me:\r\n##tensor([[0., 0., 0., 0., 0., .....0., 0., 0., 0.]])\r\n```\r\nSo what happens when we remove this layer:\r\n```\r\n##Defining our own roberta class without the token_type_embeddings layer\r\nfrom torch import nn\r\nfrom transformers.modeling_bert import BertLayerNorm, BertModel, BertPreTrainedModel\r\nfrom transformers.modeling_roberta import RobertaClassificationHead, RobertaConfig, ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP, create_position_ids_from_input_ids, CrossEntropyLoss\r\nclass MyBertEmbeddings(nn.Module):\r\n def __init__(self, config):\r\n super().__init__()\r\n self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=0)\r\n self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)\r\n #self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)\r\n\r\n # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load\r\n # any TensorFlow checkpoint file\r\n self.LayerNorm = BertLayerNorm(config.hidden_size, eps=config.layer_norm_eps)\r\n self.dropout = nn.Dropout(config.hidden_dropout_prob)\r\n\r\n def forward(self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None):\r\n if input_ids is not None:\r\n input_shape = input_ids.size()\r\n else:\r\n input_shape = inputs_embeds.size()[:-1]\r\n\r\n seq_length = input_shape[1]\r\n device = input_ids.device if input_ids is not None else inputs_embeds.device\r\n if position_ids is None:\r\n position_ids = torch.arange(seq_length, dtype=torch.long, device=device)\r\n position_ids = position_ids.unsqueeze(0).expand(input_shape)\r\n if token_type_ids is None:\r\n token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)\r\n\r\n if inputs_embeds is None:\r\n inputs_embeds = self.word_embeddings(input_ids)\r\n position_embeddings = self.position_embeddings(position_ids)\r\n #token_type_embeddings = self.token_type_embeddings(token_type_ids)\r\n\r\n embeddings = inputs_embeds + position_embeddings #+ token_type_embeddings\r\n embeddings = self.LayerNorm(embeddings)\r\n embeddings = self.dropout(embeddings)\r\n return embeddings\r\n\r\nclass MyRobertaEmbeddings(MyBertEmbeddings):\r\n\r\n def __init__(self, config):\r\n super().__init__(config)\r\n self.padding_idx = 1\r\n self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=self.padding_idx)\r\n self.position_embeddings = nn.Embedding(\r\n config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx\r\n )\r\n\r\n def forward(self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None):\r\n if position_ids is None:\r\n if input_ids is not None:\r\n # Create the position ids from the input token ids. Any padded tokens remain padded.\r\n position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx).to(input_ids.device)\r\n else:\r\n position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)\r\n\r\n return super().forward(\r\n input_ids, token_type_ids=token_type_ids, position_ids=position_ids, inputs_embeds=inputs_embeds\r\n )\r\n\r\n def create_position_ids_from_inputs_embeds(self, inputs_embeds):\r\n input_shape = inputs_embeds.size()[:-1]\r\n sequence_length = input_shape[1]\r\n\r\n position_ids = torch.arange(\r\n self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device\r\n )\r\n return position_ids.unsqueeze(0).expand(input_shape)\r\n\r\nclass MyRobertaModel(BertModel):\r\n config_class = RobertaConfig\r\n pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP\r\n base_model_prefix = \"roberta\"\r\n\r\n def __init__(self, config):\r\n super().__init__(config)\r\n\r\n self.embeddings = MyRobertaEmbeddings(config)\r\n self.init_weights()\r\n\r\n def get_input_embeddings(self):\r\n return self.embeddings.word_embeddings\r\n\r\n def set_input_embeddings(self, value):\r\n self.embeddings.word_embeddings = value\r\n\r\nclass MyRobertaForSequenceClassification(BertPreTrainedModel):\r\n config_class = RobertaConfig\r\n pretrained_model_archive_map = ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP\r\n base_model_prefix = \"roberta\"\r\n\r\n def __init__(self, config):\r\n super().__init__(config)\r\n self.num_labels = config.num_labels\r\n\r\n self.roberta = MyRobertaModel(config)\r\n self.classifier = RobertaClassificationHead(config)\r\n\r\n def forward(\r\n self,\r\n input_ids=None,\r\n attention_mask=None,\r\n token_type_ids=None,\r\n position_ids=None,\r\n head_mask=None,\r\n inputs_embeds=None,\r\n labels=None,\r\n ):\r\n outputs = self.roberta(\r\n input_ids,\r\n attention_mask=attention_mask,\r\n token_type_ids=token_type_ids,\r\n position_ids=position_ids,\r\n head_mask=head_mask,\r\n inputs_embeds=inputs_embeds,\r\n )\r\n sequence_output = outputs[0]\r\n logits = self.classifier(sequence_output)\r\n\r\n outputs = (logits,) + outputs[2:]\r\n if labels is not None:\r\n if self.num_labels == 1:\r\n # We are doing regression\r\n loss_fct = MSELoss()\r\n loss = loss_fct(logits.view(-1), labels.view(-1))\r\n else:\r\n loss_fct = CrossEntropyLoss()\r\n loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))\r\n outputs = (loss,) + outputs\r\n\r\n return outputs # (loss), logits, (hidden_states), (attentions)\r\n```\r\n\r\n```\r\nmymodel = MyRobertaForSequenceClassification.from_pretrained('roberta-base')\r\n\r\n##We need to set the weights of the randomly initialized layers to the same values\r\nimport torch\r\nfor name, param in mymodel.named_parameters():\r\n if not(torch.all(param.data.eq(model.state_dict()[name]))):\r\n print('{} is not identical'.format(name))\r\n param.data = model.state_dict()[name]\r\n##Output:\r\n##classifier.dense.weight is not identical\r\n##classifier.dense.bias is not identical\r\n##classifier.out_proj.weight is not identical\r\n##classifier.out_proj.bias is not identical\r\n```\r\nNow we can compare mymodel with model:\r\n```\r\nfrom transformers import RobertaTokenizer\r\n\r\ntokenizer = RobertaTokenizer.from_pretrained('roberta-base')\r\n\r\nfor m in [mymodel, model]:\r\n input_ids = torch.tensor(tokenizer.encode(\"Hello, my dog is cute\", add_special_tokens=True)).unsqueeze(0) # Batch size 1\r\n labels = torch.tensor([1]).unsqueeze(0) # Batch size 1\r\n outputs = m(input_ids, labels=labels)\r\n print(outputs)\r\n##(tensor(0.8658, grad_fn=<NllLossBackward>), tensor([[ 0.0927, -0.2271]], grad_fn=<AddmmBackward>))\r\n##(tensor(0.8658, grad_fn=<NllLossBackward>), tensor([[ 0.0927, -0.2271]], grad_fn=<AddmmBackward>))\r\n```\r\nand see that the output is the same.", "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/2870?src=pr&el=h1) Report\n> Merging [#2870](https://codecov.io/gh/huggingface/transformers/pull/2870?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/73028c5df0c28ca179fbe565482a9c2143787f61?src=pr&el=desc) will **increase** coverage by `<.01%`.\n> The diff coverage is `100%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2870?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2870 +/- ##\n==========================================\n+ Coverage 75.06% 75.06% +<.01% \n==========================================\n Files 94 94 \n Lines 15288 15290 +2 \n==========================================\n+ Hits 11476 11478 +2 \n Misses 3812 3812\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2870?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/tokenization\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2870/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHMucHk=) | `86.1% <ø> (ø)` | :arrow_up: |\n| [src/transformers/tokenization\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/2870/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fZGlzdGlsYmVydC5weQ==) | `100% <100%> (ø)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2870?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/2870?src=pr&el=footer). Last update [73028c5...db509eb](https://codecov.io/gh/huggingface/transformers/pull/2870?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "obsolet since 2.6 release" ]

[]

[]

[ "I suggest to close this topic and keep the discussion over at https://github.com/ThilinaRajapakse/simpletransformers/issues/229." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834060867, "node_id": "MDU6TGFiZWwxODM0MDYwODY3", "url": "https://api.github.com/repos/huggingface/transformers/labels/Ex:%20Named%20Entity%20Recognition", "name": "Ex: Named Entity Recognition", "color": "06FFD8", "default": false, "description": "" } ]

[]

[ "+1 \r\nSame for [run_language_modeling.py](https://github.com/huggingface/transformers/blob/master/examples/run_language_modeling.py)?", "What do you think about moving the examples to `AutoModels` (in this case `AutoModelForTokenClassification`) @srush @LysandreJik @julien-c ?", "@thomwolf Indeed, that would be nice.", "Yup, sounds good to me (it will make things much simpler). ", "Yes, makes sense", "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": 1834052129, "node_id": "MDU6TGFiZWwxODM0MDUyMTI5", "url": "https://api.github.com/repos/huggingface/transformers/labels/High-Level%20feature", "name": "High-Level feature", "color": "f7c9a3", "default": false, "description": "" } ]

[]

[ "I might be mistaken, but it seems that `s3_etag` verifies that the etag of a cached (downloaded) file is the same as the one that is in the S3 bucket, to ensure that you have the right files (in terms of versions, or corruption). If those files are not in the cached folder, they are downloaded.\r\n\r\nSee \r\n\r\nhttps://github.com/huggingface/transformers/blob/0dbddba6d2c5b2c6fc08866358c1994a00d6a1ff/src/transformers/file_utils.py#L330-L336", "Is there any way to turn that off? ", "Not as far as I can see. What is your use-case? Why do you need this?", "A use case where validating against external servers is not ideal is if the network is behind a firewall and/or is a containerized microservice, and you want to avoid pinging outside the firewall as much as possible.\r\n\r\nI would appreciate a config flag that disables all external pinging.", "It's not comfortable for development - I'm doing many tests with the pretrained model and it's pretty annoying as it slows down my experiments considerably. I quess I could just save and load the model myself but I was curious why `from_pretrained` takes so long.", "I think it should be possible by skipping this block (and setting `etag=None`)\r\n\r\nhttps://github.com/huggingface/transformers/blob/0dbddba6d2c5b2c6fc08866358c1994a00d6a1ff/src/transformers/file_utils.py#L399-L409\r\n\r\nwhich will then fallback to\r\n\r\nhttps://github.com/huggingface/transformers/blob/0dbddba6d2c5b2c6fc08866358c1994a00d6a1ff/src/transformers/file_utils.py#L418-L430\r\n\r\nA flag should be added to the signature, something like: `disable_outgoing=False`. When `True`, it will skip the lookup and possible download.\r\n\r\nI might be able to work on this in the future, but it's not high on my priority list.\r\n\r\nOpinions? @minimaxir @Swarzkopf314 ", "Yeah that would be great :)", "@Swarzkopf314 Can you tell me how you made the graphs in OP? (some library, I presume) So I can use them for testing.", "I made a wrapper for `pyinstrument`, feel free to use it:\r\n\r\n```python\r\nimport pyinstrument\r\n\r\n# with TreeProfiler(show_all=True):\r\n# # code to profie...\r\nclass TreeProfiler(object):\r\n\r\n def __init__(self, show_all=False):\r\n self.profiler = pyinstrument.Profiler()\r\n self.show_all = show_all # verbose output of pyinstrument profiler\r\n\r\n def __enter__(self):\r\n print(\"WITH TREE_PROFILER:\")\r\n self.profiler.start()\r\n\r\n def __exit__(self, *args):\r\n self.profiler.stop()\r\n print(self.profiler.output_text(unicode=True, color=True, show_all=self.show_all))\r\n\r\n```", "You can try out my PR https://github.com/huggingface/transformers/pull/2930 if you want.\r\n\r\n```python\r\nimport pyinstrument\r\nfrom transformers import DistilBertConfig, DistilBertModel, DistilBertTokenizer\r\n\r\n\r\nclass TreeProfiler():\r\n def __init__(self, show_all=False):\r\n self.profiler = pyinstrument.Profiler()\r\n self.show_all = show_all # verbose output of pyinstrument profiler\r\n\r\n def __enter__(self):\r\n print(\"WITH TREE_PROFILER:\")\r\n self.profiler.start()\r\n\r\n def __exit__(self, *args):\r\n self.profiler.stop()\r\n print(self.profiler.output_text(unicode=True, color=True, show_all=self.show_all))\r\n\r\n\r\ndef main():\r\n with TreeProfiler(show_all=True):\r\n config = DistilBertConfig.from_pretrained('distilbert-base-uncased', disable_outgoing=True)\r\n model = DistilBertModel.from_pretrained('distilbert-base-uncased', disable_outgoing=True)\r\n tokenizer = DistilBertTokenizer.from_pretrained('distilbert-base-uncased', disable_outgoing=True)\r\n\r\n\r\nif __name__ == '__main__':\r\n main()\r\n```\r\n\r\nThe above snippet will throw an error message when the expected files are not present in the cache. When they are, though, everything is loaded fine without the need of any additional lookups.", "Amazing, thanks a lot! <3", "No problem. Note that I have not written tests for this functionality yet. I don't think it should break the library, but if you do find some inconsistencies, please let me know.", "Excellent! :D", "Note that the parameter name has been changed to `local_files_only`.", "Note that in practice, I find some parameter \"local_files_first\" which will resolve this issue even further. As named, it will first check if the model is cached. If not, it will make internet connection and download that model. I find this useful for production and testing, thus might write some pull requests for this new feature." ]

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

[]

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

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834053813, "node_id": "MDU6TGFiZWwxODM0MDUzODEz", "url": "https://api.github.com/repos/huggingface/transformers/labels/PyTorch", "name": "PyTorch", "color": "a12bef", "default": false, "description": "Anything PyTorch" } ]

[]

[ "It is likely that your SO question was downvoted because it is a lot of unreproducible code, and not a lot of explanation. In other words: when someone reads your qusetion, it is almost impossible to answer because we cannot try your code ourselves. Try reducing it to a minimal, verifiable, executable example.\r\n\r\nThat being said: you are mixing conda and pip installations, which is a drag. Also you don't need to install pytorch-transformers AND transformers. The latter is the successor to the former, so you should only install one or the other (preferably only transformers), and fix your imports accordingly. Just install everything with pip, is my advice.", "Thanks for your answer. I am following your suggestions. However when I replace\r\n```\r\nfrom pytorch_transformers import AdamW, WarmupLinearSchedule\r\n```\r\nwith \r\n\r\n```\r\nfrom transformers import AdamW, WarmupLinearSchedule\r\n\r\n```\r\n\r\nI get this error\r\n\r\n```\r\nImportError Traceback (most recent call last)\r\n<ipython-input-7-fc8519a4dbdc> in <module>()\r\n 19 RobertaConfig, RobertaForSequenceClassification, RobertaTokenizer)\r\n 20 \r\n---> 21 from transformers import AdamW, WarmupLinearSchedule\r\n 22 \r\n 23 from utils import (convert_examples_to_features,output_modes, processors)\r\n\r\nImportError: cannot import name 'WarmupLinearSchedule'\r\n```\r\nCan you help me out?\r\nThanks\r\n", "You are probably looking for\r\n\r\nhttps://github.com/huggingface/transformers/blob/20fc18fbda3669c2f4a3510e0705b2acd54bff07/src/transformers/optimization.py#L47-L59", "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" ]

[]

[]

[ "looks good!", "On fire! :D", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2864?src=pr&el=h1) Report\n> Merging [#2864](https://codecov.io/gh/huggingface/transformers/pull/2864?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/92e974196fc35eb826f64808ae82d20c4380e3eb?src=pr&el=desc) will **increase** coverage by `1.1%`.\n> The diff coverage is `90.9%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2864?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2864 +/- ##\n=========================================\n+ Coverage 73.95% 75.06% +1.1% \n=========================================\n Files 93 94 +1 \n Lines 15272 15288 +16 \n=========================================\n+ Hits 11295 11476 +181 \n+ Misses 3977 3812 -165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2864?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/configuration\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9jb25maWd1cmF0aW9uX2Rpc3RpbGJlcnQucHk=) | `100% <ø> (ø)` | :arrow_up: |\n| [src/transformers/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9fX2luaXRfXy5weQ==) | `98.87% <ø> (ø)` | :arrow_up: |\n| [src/transformers/data/processors/squad.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9kYXRhL3Byb2Nlc3NvcnMvc3F1YWQucHk=) | `28.89% <0%> (-0.29%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `73.15% <100%> (+2.21%)` | :arrow_up: |\n| [src/transformers/modeling\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19kaXN0aWxiZXJ0LnB5) | `97.62% <100%> (-0.01%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG0ucHk=) | `86.37% <100%> (-0.04%)` | :arrow_down: |\n| [src/transformers/modeling\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19iZXJ0LnB5) | `88.16% <100%> (+0.25%)` | :arrow_up: |\n| [src/transformers/tokenization\\_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fZ3B0Mi5weQ==) | `96.26% <100%> (+0.05%)` | :arrow_up: |\n| [src/transformers/modeling\\_flaubert.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19mbGF1YmVydC5weQ==) | `83.82% <100%> (ø)` | :arrow_up: |\n| [src/transformers/tokenization\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy90b2tlbml6YXRpb25fdXRpbHMucHk=) | `86.1% <100%> (+0.41%)` | :arrow_up: |\n| ... and [14 more](https://codecov.io/gh/huggingface/transformers/pull/2864/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2864?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/2864?src=pr&el=footer). Last update [92e9741...a2925e9](https://codecov.io/gh/huggingface/transformers/pull/2864?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[]

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

[]

[ "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" ]

[]

[]

[ " Should be fixed with ee5a6856caec83e7f2f305418f3199b87ea6cc2d. I can execute your code without an error with the latest version from github.", "> Should be fixed with [ee5a685](https://github.com/huggingface/transformers/commit/ee5a6856caec83e7f2f305418f3199b87ea6cc2d). I can execute your code without an error with the latest version from github.\r\n\r\n@cronoik I appreciate the prompt response. **I didn't compile from git, rather installed via** \r\n\r\n`pip install transformer --upgrade`\r\n\r\nIt upgraded to transformers==2.4.1 -- post-upgrade though the error code changed to below:\r\n\r\n`ValueError: Can't find a vocabulary file at path /root/.cache/torch/transformers/37cc1eaaea18a456726fc28ecb438852f0ca1d9e7d259e6e3747ee33065936f6'. To load the vocabulary from a Google pretrained model use tokenizer = BertTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`\r\n", "The mentioned commit is not part of 2.4.1. You have to wait for the next release or pull transformers from git.", "Ok, I will pull from the git for the time being. Thank you!", "Please close the issure when your problem is solved.", "We can close this since we have a workaround, and the team is aware of the issue to be rolled out in the next release. Thanks!!", "v2.5.0 was released a few days ago, `distilbert-base-cased` is now accessible via the pip release! :)", "The vocab file is missing here: \r\nhttps://huggingface.co/distilbert-base-cased#list-files\r\nWhile the auto-downloaded model has one.", "I'm still having the same problem. Using transformers version 2.8.0, neither `distilbert-base-cased` or `distilbert-base-uncased` are available. I also ran the following command:\r\n\r\n```\r\nimport pytorch_pretrained_bert as ppb\r\nassert 'distilbert-base-uncased' in ppb.modeling.PRETRAINED_MODEL_ARCHIVE_MAP\r\n```\r\n\r\nWhich results in `AssertionError`. Any thoughts on what might be going on here?", "Are you really using the `transformers` [1] package? The code you have showed contains only the `pytorch_pretrained_bert` [2] package which doesn't contain distilbert. While `pytorch_pretrained_bert` [2] and `transformers` [1] are both packages from huggingface, they are not the same. `pytorch_pretrained_bert` last release is from april 2019. Please use the `transformers` package [1].\r\n\r\n[1] https://pypi.org/project/transformers/\r\n[2] https://pypi.org/project/pytorch-pretrained-bert/#description", "Thanks for the quick reply: I am using transformers, I picked up that code snippet from another issue, must have been for that package. \r\n\r\nI realize what I did wrong: I was using BertTokenizer/BertModel to load, and I should have been using DistilBertTokenizer/DistilBertModel. It's working now, thanks!" ]

[]

[]

[ "Hi, please look into the documentation of the [attention mask](https://huggingface.co/transformers/glossary.html#attention-mask).", "Actually, it was a valid question. The output will be numerically different for sure since there are extra positions to attend and even if those are paddings -> there is a difference among the float values. However, the real question is whether the (cosine/dot) similarity among the resulting vectors have changed at all." ]

[ { "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/2859?src=pr&el=h1) Report\n> Merging [#2859](https://codecov.io/gh/huggingface/transformers/pull/2859?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/925a13ced1e155ea7e55e14e177a7b5ae7ad174c?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2859?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2859 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15287 15287 \n=======================================\n Hits 11475 11475 \n Misses 3812 3812\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2859?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/2859?src=pr&el=footer). Last update [925a13c...917aa8d](https://codecov.io/gh/huggingface/transformers/pull/2859?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "@yvespeirsman Thanks for sharing! I can't push to your fork so I'll merge this and tweak it (languages have to be in a list)" ]

[ { "id": 1834053007, "node_id": "MDU6TGFiZWwxODM0MDUzMDA3", "url": "https://api.github.com/repos/huggingface/transformers/labels/Ex:%20LM%20(Pretraining)", "name": "Ex: LM (Pretraining)", "color": "76FFAF", "default": false, "description": "Related to language modeling pre-training" } ]

[]

[ "Hi @ARDUJS can you update your issue title to something more descriptive? Thanks!", "Should be correct -> 80% masked, that means 20% is left. Using this 20% in 50 % the random word is used, 50% original token is kept. So both random word and original has an overall prob. of 10%.\r\n\r\nOriginal BERT is using the same logic, see [here](https://github.com/google-research/bert/blob/master/create_pretraining_data.py#L391)." ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2857?src=pr&el=h1) Report\n> Merging [#2857](https://codecov.io/gh/huggingface/transformers/pull/2857?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/925a13ced1e155ea7e55e14e177a7b5ae7ad174c?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2857?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2857 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15287 15287 \n=======================================\n Hits 11475 11475 \n Misses 3812 3812\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2857?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/2857?src=pr&el=footer). Last update [925a13c...acca7c4](https://codecov.io/gh/huggingface/transformers/pull/2857?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/2855?src=pr&el=h1) Report\n> Merging [#2855](https://codecov.io/gh/huggingface/transformers/pull/2855?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/925a13ced1e155ea7e55e14e177a7b5ae7ad174c?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2855?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2855 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15287 15287 \n=======================================\n Hits 11475 11475 \n Misses 3812 3812\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2855?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/2855?src=pr&el=footer). Last update [925a13c...c86fc74](https://codecov.io/gh/huggingface/transformers/pull/2855?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

[]

[]

[ "Thanks!", "Welcome, Julien!\nThis one won't be my last contribution! :)\nNot so easy :P\n\nEl vie., 14 feb. 2020 5:05, Julien Chaumond <[email protected]>\nescribió:\n\n> Thanks!\n>\n> —\n> You are receiving this because you authored the thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/transformers/pull/2854?email_source=notifications&email_token=AA34BHPFYBDFG2UBR7FZQHDRCYJ6ZA5CNFSM4KVADOQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOELXP3CI#issuecomment-586087817>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/AA34BHIPBUNF5MRYBVO6ZPTRCYJ6ZANCNFSM4KVADOQQ>\n> .\n>\n", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2854?src=pr&el=h1) Report\n> Merging [#2854](https://codecov.io/gh/huggingface/transformers/pull/2854?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/4d36472b96d144887cbe95b083f0d2091fd5ff03?src=pr&el=desc) will **decrease** coverage by `25.28%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2854?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2854 +/- ##\n===========================================\n- Coverage 75.06% 49.77% -25.29% \n===========================================\n Files 94 94 \n Lines 15287 15287 \n===========================================\n- Hits 11475 7609 -3866 \n- Misses 3812 7678 +3866\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2854?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_camembert.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9jYW1lbWJlcnQucHk=) | `0% <0%> (-100%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_xlm\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl94bG1fcm9iZXJ0YS5weQ==) | `0% <0%> (-100%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9yb2JlcnRhLnB5) | `0% <0%> (-100%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9jdHJsLnB5) | `0% <0%> (-97.83%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_t5.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl90NS5weQ==) | `0% <0%> (-96.55%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9iZXJ0LnB5) | `0% <0%> (-96.06%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9vcGVuYWkucHk=) | `0% <0%> (-95.85%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9kaXN0aWxiZXJ0LnB5) | `0% <0%> (-95.12%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9ncHQyLnB5) | `0% <0%> (-94.67%)` | :arrow_down: |\n| [src/transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl91dGlscy5weQ==) | `0% <0%> (-92.79%)` | :arrow_down: |\n| ... and [18 more](https://codecov.io/gh/huggingface/transformers/pull/2854/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2854?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/2854?src=pr&el=footer). Last update [4d36472...3643bb8](https://codecov.io/gh/huggingface/transformers/pull/2854?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/2853?src=pr&el=h1) Report\n> Merging [#2853](https://codecov.io/gh/huggingface/transformers/pull/2853?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/1eec69a90007b8f4a7af10805dab4904ea5dea77?src=pr&el=desc) will **decrease** coverage by `1.07%`.\n> The diff coverage is `100%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2853?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2853 +/- ##\n==========================================\n- Coverage 75.06% 73.98% -1.08% \n==========================================\n Files 94 94 \n Lines 15287 15288 +1 \n==========================================\n- Hits 11475 11311 -164 \n- Misses 3812 3977 +165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2853?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/2853/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9fX2luaXRfXy5weQ==) | `98.87% <ø> (ø)` | :arrow_up: |\n| [src/transformers/pipelines.py](https://codecov.io/gh/huggingface/transformers/pull/2853/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9waXBlbGluZXMucHk=) | `71.5% <100%> (+0.07%)` | :arrow_up: |\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2853/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2853/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.91% <0%> (-9.86%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2853/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `70.86% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2853/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `92.07% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2853/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2853?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/2853?src=pr&el=footer). Last update [1eec69a...549ce87](https://codecov.io/gh/huggingface/transformers/pull/2853?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/2852?src=pr&el=h1) Report\n> Merging [#2852](https://codecov.io/gh/huggingface/transformers/pull/2852?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/1eec69a90007b8f4a7af10805dab4904ea5dea77?src=pr&el=desc) will **decrease** coverage by `1.07%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2852?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2852 +/- ##\n==========================================\n- Coverage 75.06% 73.98% -1.08% \n==========================================\n Files 94 94 \n Lines 15287 15287 \n==========================================\n- Hits 11475 11310 -165 \n- Misses 3812 3977 +165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2852?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2852/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2852/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.91% <0%> (-9.86%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2852/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `70.86% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2852/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `92.07% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2852/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2852?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/2852?src=pr&el=footer). Last update [1eec69a...59baea0](https://codecov.io/gh/huggingface/transformers/pull/2852?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "See #2851.\r\n\r\nThanks!" ]

[ { "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" } ]

[]

[ "Thanks for sharing!\r\n\r\nHow did you pre-train this model (infrastructure, number of epochs, etc.)?\r\nDo you have eval results on downstream tasks?\r\n\r\nAlso you can add a \r\n```\r\n---\r\nlanguage: greek\r\n---\r\n```\r\ntag to the top of the file\r\n\r\nI'll merge this in the meantime, thanks for sharing!", "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2851?src=pr&el=h1) Report\n> Merging [#2851](https://codecov.io/gh/huggingface/transformers/pull/2851?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/8744402f1eb51c7ae6b86cae1015983096beb655?src=pr&el=desc) will **decrease** coverage by `1.07%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2851?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2851 +/- ##\n==========================================\n- Coverage 75.06% 73.98% -1.08% \n==========================================\n Files 94 94 \n Lines 15287 15287 \n==========================================\n- Hits 11475 11310 -165 \n- Misses 3812 3977 +165\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/2851?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [src/transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/2851/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ190Zl9weXRvcmNoX3V0aWxzLnB5) | `8.72% <0%> (-81.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/2851/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19yb2JlcnRhLnB5) | `85.91% <0%> (-9.86%)` | :arrow_down: |\n| [src/transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/2851/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ194bG5ldC5weQ==) | `70.86% <0%> (-2.3%)` | :arrow_down: |\n| [src/transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/2851/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19jdHJsLnB5) | `92.07% <0%> (-2.21%)` | :arrow_down: |\n| [src/transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/2851/diff?src=pr&el=tree#diff-c3JjL3RyYW5zZm9ybWVycy9tb2RlbGluZ19vcGVuYWkucHk=) | `80.2% <0%> (-1.35%)` | :arrow_down: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2851?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/2851?src=pr&el=footer). Last update [8744402...6aa9688](https://codecov.io/gh/huggingface/transformers/pull/2851?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "I just amended the initial model card with extra information on the pre-training process. No evaluation yet, I hope will have some experiments, pretty soon. Thanks!" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/2850?src=pr&el=h1) Report\n> Merging [#2850](https://codecov.io/gh/huggingface/transformers/pull/2850?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/7f98edd7e362a64c947b083cfc0c401c4d0ffe91?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/2850?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #2850 +/- ##\n=======================================\n Coverage 75.06% 75.06% \n=======================================\n Files 94 94 \n Lines 15287 15287 \n=======================================\n Hits 11475 11475 \n Misses 3812 3812\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/2850?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/2850?src=pr&el=footer). Last update [7f98edd...51830ef](https://codecov.io/gh/huggingface/transformers/pull/2850?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "I've added a way to switch between `PyTorch` and `TensorFlow` implementations. I didn't want to have a wall of code that had the two frameworks, so now there's a toggle to show which framework you would like to see.\r\n\r\nIt works as follows: a javascript method parses the documentation page shown and looks for `.highlight` classes, which are the code blocks. In there, it looks for `## PYTORCH CODE`, which represents the beginning of a `PyTorch` snippet and `## TENSORFLOW CODE` which represents the beginning of a `TensorFlow` snippet.\r\n\r\nWould love an opinion on the Javascript code as well. Would love to convert this to TS down the road.\r\n\r\nHere's a gif of the result\r\n\r\n\r\n", "@LysandreJik reviewing the JS now and had a question. Depending on your thoughts, would it make sense from a UX standpoint for all of the buttons to toggle together? So, if a user selects \"Tensorflow\" all of the code blocks would switch to \"Tensorflow\". ", "I guess this would be cool and makes sense from a UX standpoint. Do you think it's necessary or can it wait for the second version?", "It can 100% wait for a second version. " ]

[ { "id": 1834056761, "node_id": "MDU6TGFiZWwxODM0MDU2NzYx", "url": "https://api.github.com/repos/huggingface/transformers/labels/Core:%20Modeling", "name": "Core: Modeling", "color": "FF8446", "default": false, "description": "Internals of the library; Models." }, { "id": 1845609017, "node_id": "MDU6TGFiZWwxODQ1NjA5MDE3", "url": "https://api.github.com/repos/huggingface/transformers/labels/seq2seq", "name": "seq2seq", "color": "fef2c0", "default": false, "description": "" }, { "id": 1862634478, "node_id": "MDU6TGFiZWwxODYyNjM0NDc4", "url": "https://api.github.com/repos/huggingface/transformers/labels/Should%20Fix", "name": "Should Fix", "color": "FF0000", "default": false, "description": "This has been identified as a bug and should be fixed." } ]

[]

[ "I put this as a bug because the code as-is does not hint that Model2LSTM does not work. \r\nhttps://github.com/huggingface/transformers/blob/90ab15cb7a8fcf8bf58c05453ddf1aa6a4fa00c1/src/transformers/modeling_encoder_decoder.py\r\nIt would be great to say that LSTM is not currently supported there. ", "Indeed, my initial comment was a mistake. I'm looking into it now." ]

[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 1834054694, "node_id": "MDU6TGFiZWwxODM0MDU0Njk0", "url": "https://api.github.com/repos/huggingface/transformers/labels/TensorFlow", "name": "TensorFlow", "color": "FF6F00", "default": false, "description": "Anything TensorFlow" } ]

[]

[ "Hi! This feature would be great to have.\r\n\r\nI'm curious how `TFBertMaskedLM` (and the like) are supposed to be used with the keras `fit()` functionality?\r\n\r\nIt seems like one is supposed to loop through the training data and calculate the cross-entropy loss for each batch (#2926). I see there was related discussion also here #1999.\r\n\r\nHappy for any input!", "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", "Closing because this feature was added in #4530 " ]

[ { "id": 1845609017, "node_id": "MDU6TGFiZWwxODQ1NjA5MDE3", "url": "https://api.github.com/repos/huggingface/transformers/labels/seq2seq", "name": "seq2seq", "color": "fef2c0", "default": false, "description": "" } ]

[]

[ "We are hard at work on this! I'd estimate 6 weeks out.", "Looking forward to this for the T5 model :)", "@sshleifer any updates? ", "The example doesn't seem to show training/fine-tuning, only evaluation of already fine-tuned models.", "@sshleifer Hello, any updates for training/fine-tuning on text generation for T5 model ?", "`summarization/bart/finetune.py` supports T5." ]