Datasets:

---

DanielPFlorian

/

Transformers-Github-Issues

like 0

[]

[]

[ "Noticed the same thing on the Salesforce version as well. ", "Do you test with [this url](https://github.com/salesforce/ctrl/blob/master/generation.py)?\r\n@tanselmi If not, please specify which script do you try out\r\n\r\nN.B. if you see this code, the lines 40-41-42-43 assign the seed passed as argument (default value is 1337) for NumPy, PyTorch and CUDA. It's strange that the internal seed doesn't modify the CTRL's model output.", "> Noticed the same thing on the Salesforce version as well.\r\n\r\nIn general, is it a bug in the code implementation of the CTRL model or a \"model property\" (i.e. the CTRL model is not affect by a random seed set at the start of the process) ?", "Hi, this is not due to a bug, this is due to your argument `temperature=0`, which implies greedy sampling via argmax, rather than polynomial sampling like it is usually done.\r\n\r\nYou can try it out with a temperature of 0.2:\r\n\r\nWith seed 42 - Links Transformers movie is **a huge hit in the box office and has been**\r\nWith seed 22 - Links Transformers movie is **a big deal for the studio. The first film was**", "Thanks for the update and explanation. I will give it a go.\n\nOn Mon, Nov 4, 2019, 08:37 Lysandre Debut <[email protected]> wrote:\n\n> Hi, this is not due to a bug, this is due to your argument temperature=0,\n> which implies greedy sampling via argmax, rather than polynomial sampling\n> like it is usually done.\n>\n> You can try it out with a temperature of 0.2:\n>\n> With seed 42 - Links Transformers movie is *a huge hit in the box office\n> and has been*\n> With seed 22 - Links Transformers movie is *a big deal for the studio.\n> The first film was*\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/1710?email_source=notifications&email_token=ABO5AYW55TLAXKGM2PODQJTQSAXSZA5CNFSM4JIMFK42YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC7OTJY#issuecomment-549382567>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/ABO5AYWK6JXQUIHOLWNNXG3QSAXSZANCNFSM4JIMFK4Q>\n> .\n>\n", "@LysandreJik I confirm setting temperature to non-zero makes the seed argument work fine.\r\nThank you!" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1709?src=pr&el=h1) Report\n> Merging [#1709](https://codecov.io/gh/huggingface/transformers/pull/1709?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/8a62835577a2a93642546858b21372e43c1a1ff8?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1709?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1709 +/- ##\n=======================================\n Coverage 85.14% 85.14% \n=======================================\n Files 94 94 \n Lines 13920 13920 \n=======================================\n Hits 11852 11852 \n Misses 2068 2068\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1709?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/1709?src=pr&el=footer). Last update [8a62835...e5b1048](https://codecov.io/gh/huggingface/transformers/pull/1709?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Ok thanks!" ]

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

[]

[ "Hi, in the `bert-base-uncased` vocab.txt file, the id of the token \"the\" is 1996. You can see it [inside the file, where \"the\" is on the row 1996](https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-vocab.txt). You can also check it out directly in the tokenizer with the following command: \r\n\r\n```py\r\ntokenizer = BertTokenizer.from_pretrained(\"bert-base-uncased\")\r\ntokenizer.vocab.get(\"the\") # 1996\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": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null } ]

[]

[ "The summarization/generation scripts are still work in progress. They should be included in the next release.", "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 } ]

[]

[ "That is a good way to load the labels! \r\n\r\nPlease be aware, however, that by loading the configuration this way you will not load the weights for your model associated with the `bert-base-uncased` checkpoint as you would have done had you used the `BertForSequenceClassification.from_pretrained(...)` method.", "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" ]

[]

[]

[ "try this \r\n\r\n```\r\n%%bash\r\nrm -r hugging-face-squad\r\nmkdir hugging-face-squad\r\necho > hugging-face-squad/__init__.py\r\ncd hugging-face-squad\r\nwget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/run_squad.py'\r\nwget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/utils_squad.py'\r\nwget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/utils_squad_evaluate.py'\r\nsed -i 's/utils_squad_evaluate/.utils_squad_evaluate/g' utils_squad.py\r\nsed -i 's/utils_squad/.utils_squad/g' run_squad.py\r\n\r\nimport importlib\r\nhfs = importlib.import_module('.run_squad', package='hugging-face-squad')\r\n\r\nexamples = hfs.read_squad_examples(input_file=\"train-v2.0.json\",\r\n is_training=True,\r\n version_2_with_negative=True)\r\n\r\n\r\n\r\n```\r\n\r\n", "For me, just adding\r\n\r\n```\r\n!wget 'https://raw.githubusercontent.com/huggingface/transformers/master/examples/run_squad.py'\r\n!wget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/utils_squad.py'\r\n!wget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/utils_squad_evaluate.py'\r\n```\r\n\r\nIn the same directory as my script worked", "> try this\r\n> \r\n> ```\r\n> %%bash\r\n> rm -r hugging-face-squad\r\n> mkdir hugging-face-squad\r\n> echo > hugging-face-squad/__init__.py\r\n> cd hugging-face-squad\r\n> wget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/run_squad.py'\r\n> wget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/utils_squad.py'\r\n> wget 'https://raw.githubusercontent.com/huggingface/pytorch-transformers/master/examples/utils_squad_evaluate.py'\r\n> sed -i 's/utils_squad_evaluate/.utils_squad_evaluate/g' utils_squad.py\r\n> sed -i 's/utils_squad/.utils_squad/g' run_squad.py\r\n> \r\n> import importlib\r\n> hfs = importlib.import_module('.run_squad', package='hugging-face-squad')\r\n> \r\n> examples = hfs.read_squad_examples(input_file=\"train-v2.0.json\",\r\n> is_training=True,\r\n> version_2_with_negative=True)\r\n>\r\n@manishiitg ...For this also getting errors....utils_squad.py, & utils_squad_evaluate.py, was not found.\r\n", "Hi, when I was trying to wget the scripts, it says 404 not found... Probably it is been removed somewhere? \r\n\r\n`!wget 'https://raw.githubusercontent.com/huggingface/transformers/master/examples/run_squad.py'` \r\n`Proxy request sent, awaiting response... 404 Not Found\r\n2020-05-21 20:18:30 ERROR 404: Not Found.\r\n`", "I believe the code has been refactored a while ago, maybe try the Dec version of transformers, or update your code to get the location of the new data processors", "Hi, did you solve the problem ? For me it is the same issue and I just cant find a solution how to get the proper python scripts ? It would be super kind if you could help me :)", "I was facing this issue and finally, after some thinking, I tried finding utils_squad.py and utils_squad_evaluate.py from this [repo](https://github.com/nlpyang/pytorch-transformers) and copied the link of the raw version of those two files. So these commands won't give a 404 error and import then worked for me:)\r\n`\r\n!wget 'https://raw.githubusercontent.com/nlpyang/pytorch-transformers/master/examples/utils_squad.py'\r\n`\r\n`\r\n!wget 'https://raw.githubusercontent.com/nlpyang/pytorch-transformers/master/examples/utils_squad_evaluate.py'\r\n`", "Thanks @lazyCodes7!", "> Thanks @lazyCodes7!\r\n\r\ncool!" ]

[]

[]

[ "extra details:\r\n\r\nexample = list(bert_validation_matched_dataset.__iter__())[0]\r\nexample\r\n```\r\n{'hypothesis': <tf.Tensor: id=4914985, shape=(64,), dtype=string, numpy=\r\n array([b'The St. Louis Cardinals have always won.',\r\n b'The fortress was built a number of years after the caravanserai.',\r\n b'Mastihohoria is a collection of twenty mastic villages built be the genoese.',\r\n b'Reggae is the most popular music style in Jamaica.',\r\n b'I am able to receive mail at my workplace.',\r\n b'Men have lower levels of masculinity than in the decade before now.',\r\n b'Clinton has several similarities to Whitewater or Flytrap.',\r\n b'A search has been conducted for an AIDS vaccine.',\r\n b'We can acknowledge there is fallout from globalization around the world.',\r\n b'My feelings towards pigeons are filled with animosity.',\r\n b'She could see through the ghosts with ease.',\r\n b'Leading organizations want to be sure their processes are successful.',\r\n b'The Postal Service spends considerable sums on cost analysis.',\r\n b'Indeed we got away from the original subject.',\r\n b'Economic growth continued apace, with many people employed by the railroad repair shop.',\r\n b'Neither side is actually interested in a settlement at this time.',\r\n b'The rooms are opulent, and used for formal, elegant events.',\r\n b\"The East side of the square is where the Old King's House stands.\",\r\n b'The islands are part of France now instead of just colonies.',\r\n b'A stoichiometry of 1.03 is typical when the FGD process is not producing gypsum by-product',\r\n b'You can hire the equipment needed for windsurfing at Bat Galim Beach. ',\r\n b\"There isn't enough room for an airport on the island.\",\r\n b'The setup rewards good business practices.',\r\n b'People sacrifice their lives for farmers and slaves.',\r\n b\"She doesn't like people like me. \",\r\n b\"It's nothing like a drug hangover.\",\r\n b'The philosophy was to seize opportunities when the economy is doing poorly.',\r\n b\"Bill Clinton isn't a rapist\",\r\n b'Various episodes depict that he is a member.',\r\n b\"Bellagio's water display was born from this well received show.\",\r\n b'Fannie Mae had terrible public-relations.',\r\n b\"Gododdin's accomplishments have been recorded in a Welsh manuscript.\",\r\n b'I can imagine how you are troubled by insects up there',\r\n b'Howard Berman is a Democrat of the House.',\r\n b'Gore dodged the draft.', b\"Jon was glad that she wasn't. \",\r\n b'Section 414 helps balance allowance allocations for units.',\r\n b'Reducing HIV is important, but there are also other worthy causes.',\r\n b'I think there are some small colleges that are having trouble.',\r\n b'The best hotels in the region are in Hassan. ',\r\n b'She mentioned approaching the law with a holistic approach/',\r\n b\"Select this option for Finkelstein's understanding of why this logic is expedient.\",\r\n b\"It's impossible to have a plate hand-painted to your own design in Hong Kong.\",\r\n b\"We could annex Cuba, but they wouldn't like that.\",\r\n b'She really needs to mention it',\r\n b\"The basics don't need to be right first.\",\r\n b'Standard Costing was applied to the ledger.',\r\n b'The exhibition was too bare and too boring. ',\r\n b'The uncle had no match in administration; certainly not in his inefficient and careless nephew, Charles Brooke.',\r\n b'The Legacy Golf Club is just inside city limits.',\r\n b'They do not give money to legal services.',\r\n b'Do you want some coffee?',\r\n b'In 1917, the Brittish General Allenby surrendered the city using a bed-sheet.',\r\n b'Daniel explained what was happening.',\r\n b'That never happened to me.', b'You would have a prescription.',\r\n b\"It's lovely speaking with you. \",\r\n b'Each Pokemon card pack is filled with every rare card a kid could want.',\r\n b'He generally reports very well on all kinds of things.',\r\n b'The final rule was declared not to be an economically significant regulator action.',\r\n b'Dana, this conversation bored me.',\r\n b'Andratx is on the northwest coast and the Cape of Formentor is further east.',\r\n b'Sunblock is an unnecessary precaution if you are in the water.',\r\n b'U.S. consumers and factories in East Asia benefit from imports.'],\r\n dtype=object)>,\r\n 'idx': <tf.Tensor: id=4914986, shape=(64,), dtype=int32, numpy=\r\n array([3344, 3852, 5009, 5398, 2335, 647, 7823, 8927, 2302, 4800, 8628,\r\n 637, 7756, 2189, 3146, 8990, 4759, 2592, 96, 5144, 2373, 7698,\r\n 2862, 1558, 7639, 3860, 416, 5768, 9299, 3149, 2927, 5914, 4960,\r\n 2880, 8203, 7787, 7556, 6465, 9781, 4053, 1217, 7178, 39, 8885,\r\n 6666, 8157, 3995, 1758, 5552, 4476, 3325, 7537, 7940, 8409, 7899,\r\n 4104, 2874, 4845, 3934, 5351, 2982, 5235, 2614, 6318], dtype=int32)>,\r\n 'label': <tf.Tensor: id=4914987, shape=(64,), dtype=int64, numpy=\r\n array([2, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 2, 0, 2, 0, 1,\r\n 2, 1, 1, 2, 2, 1, 0, 0, 2, 0, 0, 0, 2, 0, 0, 0, 0, 1, 0, 0, 2, 0,\r\n 2, 2, 1, 2, 2, 2, 2, 2, 2, 0, 2, 0, 0, 2, 1, 2, 2, 1, 2, 0])>,\r\n 'premise': <tf.Tensor: id=4914988, shape=(64,), dtype=string, numpy=\r\n array([b\"yeah well losing is i mean i'm i'm originally from Saint Louis and Saint Louis Cardinals when they were there were uh a mostly a losing team but\",\r\n b'Beside the fortress lies an 18th-century caravanserai, or inn, which has been converted into a hotel, and now hosts regular folklore evenings of Turkish dance and music.',\r\n b'The twenty mastic villages known collectively as mastihohoria were built by the Genoese in the 14 15th centuries.',\r\n b'Jamaican music ska and, especially, reggae has since the 1970s been exported and enjoyed around the world.',\r\n b'for me now the address is the same you know my my office address',\r\n b'[W]omen mocking men by calling into question their masculinity is also classified as sexual harassment, the paper added.',\r\n b\"Watergate remains for many an unhealed wound, and Clinton's critics delight in needling him with Watergate comparisons--whether to Whitewater or Flytrap.\",\r\n b'The search for an AIDS vaccine currently needs serious help, with the U.S. government, the biggest investor in the effort, spending less than 10 percent of its AIDS-research budget on the problem.',\r\n b'First, we can acknowledge, and maybe even do something about, some of the disaffecting fallout from globalization, such as pollution and cultural dislocation.',\r\n b'I hate pigeons.',\r\n b'From that spot she could see all of them and, should she need to, she could see through them as well.',\r\n b'We also have found that leading organizations strive to ensure that their core processes efficiently and effectively support mission-related outcomes.',\r\n b'Also, considerable sums are spent by the Postal Service analyzing the costs associated with worksharing, and mailers/competitors incur considerable expense litigating their positions on worksharing before the Postal Rate Commission.',\r\n b'yeah well we veered from the subject',\r\n b'Growth continued for ten years, and by 1915 the town had telephones, round-the-clock electricity, and a growing population many of whom worked in the railroad repair shop.',\r\n b\"And if, as ultimately happened, no settlement resulted, we could shrug our shoulders, say, 'Hey, we tried,' and act like unsuccessful brokers to an honorable peace.\",\r\n b'Lavishly furnished and decorated, with much original period furniture, the rooms are used for ceremonial events, visits from foreign dignitaries, and EU meetings.',\r\n b\"On the west side of the square is Old King's House (built in 1762), which was the official residence of the British governor; it was here that the proclamation of emancipation was issued in 1838.\",\r\n b'All of the islands are now officially and proudly part of France, not colonies as they were for some three centuries.',\r\n b'8 A stoichiometry of 1.03 is typical when the FGD process is producing gypsum by-product, while a stoichiometry of 1.05 is needed to produce waste suitable for a landfill.',\r\n b' The equipment you need for windsurfing can be hired from the beaches at Tel Aviv (marina), Netanya, Haifa (at Bat Galim beach), Tiberias, and Eilat.',\r\n b'Since there is no airport on the island, all visitors must arrive at the port, Skala, where most of the hotels are located and all commercial activity is carried out.',\r\n b'The entire setup has an anti-competitive, anti-entrepreneurial flavor that rewards political lobbying rather than good business practices.',\r\n b'Why bother to sacrifice your lives for dirt farmers and slavers?',\r\n b'She hates me.\"',\r\n b'and the same is true of the drug hangover you know if you',\r\n b'In the meantime, the philosophy is to seize present-day opportunities in the thriving economy.',\r\n b'Most of the Clinton women were in their 20s at the time of their Clinton encounter',\r\n b'On various episodes he is a member, along with Bluebeard and the Grim Reaper, of the Jury of the Damned; he takes part in a snake-bludgeoning (in a scandal exposed by a Bob Woodward book); his enemies list is used for dastardly purposes; even his dog Checkers is said to be bound for hell.',\r\n b'This popular show spawned the aquatic show at the Bellagio.',\r\n b\"Not surprisingly, then, Fannie Mae's public-relations operation is unparalleled in Washington.\",\r\n b'Little is recorded about this group, but they were probably the ancestors of the Gododdin, whose feats are told in a seventh-century Old Welsh manuscript.',\r\n b'i understand i can imagine you all have much trouble up there with insects or',\r\n b'Howard Berman of California, an influential Democrat on the House International Relations Committee.',\r\n b'An article explains that Al Gore enlisted for the Vietnam War out of fealty to his father and distaste for draft Gore deplored the inequity of the rich not having to serve.',\r\n b\"I am glad she wasn't, said Jon.\",\r\n b'If necessary to meeting the restrictions imposed in the preceding sentence, the Administrator shall reduce, pro rata, the basic Phase II allowance allocations for each unit subject to the requirements of section 414.',\r\n b\"Second, reducing the rate of HIV transmission is in any event not the only social goal worth If it were, we'd outlaw sex entirely.\",\r\n b\"yes well yeah i am um actually actually i think that i at the higher level education i don't think there's so much of a problem there it's pretty much funded well there are small colleges that i'm sure are struggling\",\r\n b'The most comfortable way to see these important Hoysala temples is to visit them on either side of an overnight stay at Hassan, 120 km (75 miles) northwest of Mysore.',\r\n b'We saw a whole new model develop - a holistic approach to lawyering, one-stop shopping, she said. ',\r\n b\"Click here for Finkelstein's explanation of why this logic is expedient.\",\r\n b'In Hong Kong you can have a plate, or even a whole dinner service, hand-painted to your own design.',\r\n b\"of course you could annex Cuba but they wouldn't like that a bit\",\r\n b'She hardly needs to mention it--the media bring it up anyway--but she invokes it subtly, alluding (as she did on two Sunday talk shows) to women who drive their daughters halfway across the state to shake my hand, a woman they dare to believe in.',\r\n b'First, get the basics right, that is, the blocking and tackling of financial reporting.',\r\n b'STANDARD COSTING - A costing method that attaches costs to cost objects based on reasonable estimates or cost studies and by means of budgeted rates rather than according to actual costs incurred.',\r\n b'NEH-supported exhibitions were distinguished by their elaborate wall panels--educational maps, photomurals, stenciled treatises--which competed with the objects themselves for space and attention.',\r\n b'More reserved and remote but a better administrator and financier than his uncle, Charles Brooke imposed on his men his own austere, efficient style of life.',\r\n b'Also beyond city limits is the Legacy Golf Club in the nearby suburb of Henderson.',\r\n b'year, they gave morethan a half million dollars to Western Michigan Legal Services.',\r\n b\"'Would you like some tea?'\",\r\n b'On a December day in 1917, British General Allenby rode up to Jaffa Gate and dismounted from his horse because he would not ride where Jesus walked; he then accepted the surrender of the city after the Ottoman Turks had fled (the flag of surrender was a bed-sheet from the American Colony Hotel).',\r\n b'Daniel took it upon himself to explain a few things.',\r\n b'yep same here',\r\n b\"because then they'll or you have a prescription\",\r\n b\"well it's a pleasure talking with you\",\r\n b'By seeding packs with a few high-value cards, the manufacturer is encouraging kids to buy Pokemon cards like lottery tickets.',\r\n b\"He reported masterfully on the '72 campaign and the Hell's Angels.\",\r\n b'The final rule was determined to be an economically significant regulatory action by the Office of Management and Budget and was approved by OMB as complying with the requirements of the Order on March 26, 1998.',\r\n b\"well Dana it's been really interesting and i appreciate talking with you\",\r\n b'The dramatic cliffs of the Serra de Tramuntana mountain range hug the coastline of the entire northwest and north, from Andratx all the way to the Cape of Formentor.',\r\n b'Keep young skins safe by covering them with sunblock or a T-shirt, even when in the water.',\r\n b'In the short term, U.S. consumers will benefit from cheap imports (as will U.S. multinationals that use parts made in East Asian factories).'],\r\n dtype=object)>}\r\n```\r\n\r\nand\r\nexample1 = list(bert_train_dataset.__iter__())[0]\r\nexample1\r\n```\r\n({'attention_mask': <tf.Tensor: id=4915606, shape=(32, 128), dtype=int32, numpy=\r\n array([[1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0],\r\n ...,\r\n [1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0],\r\n [1, 1, 1, ..., 0, 0, 0]], dtype=int32)>,\r\n 'input_ids': <tf.Tensor: id=4915607, shape=(32, 128), dtype=int32, numpy=\r\n array([[ 101, 1105, 1128, ..., 0, 0, 0],\r\n [ 101, 1448, 2265, ..., 0, 0, 0],\r\n [ 101, 17037, 20564, ..., 0, 0, 0],\r\n ...,\r\n [ 101, 178, 1274, ..., 0, 0, 0],\r\n [ 101, 6249, 1107, ..., 0, 0, 0],\r\n [ 101, 146, 1354, ..., 0, 0, 0]], dtype=int32)>,\r\n 'token_type_ids': <tf.Tensor: id=4915608, shape=(32, 128), dtype=int32, numpy=\r\n array([[0, 0, 0, ..., 0, 0, 0],\r\n [0, 0, 0, ..., 0, 0, 0],\r\n [0, 0, 0, ..., 0, 0, 0],\r\n ...,\r\n [0, 0, 0, ..., 0, 0, 0],\r\n [0, 0, 0, ..., 0, 0, 0],\r\n [0, 0, 0, ..., 0, 0, 0]], dtype=int32)>},\r\n <tf.Tensor: id=4915609, shape=(32,), dtype=int64, numpy=\r\n array([1, 0, 1, 1, 0, 1, 0, 0, 2, 0, 0, 0, 2, 2, 2, 1, 0, 1, 2, 0, 2, 2,\r\n 0, 1, 0, 1, 1, 2, 1, 1, 2, 2])>)\r\n```", "In the above seems that ```bert_validation_matched_dataset``` 's format is wrong. I would expect to be similar to ```bert_train_dataset```. \r\n``` bert_validation_matched_dataset``` is produced with the following code:\r\n bert_validation_matched_dataset = glue_convert_examples_to_features(validation_matched_dataset, bert_tokenizer, 128, 'mnli', label_list=['0', '1', '2']) \r\nbert_validation_matched_dataset = validation_matched_dataset.batch(64)\r\n\r\nAny idea why that didn't work?", "OK, I found out. \r\nbert_validation_matched_dataset = glue_convert_examples_to_features(validation_matched_dataset, bert_tokenizer, 128, 'mnli', label_list=['0', '1', '2'])\r\nbert_validation_matched_dataset = **validation_matched_dataset**.batch(64)\r\nI have to write bert_validation_matched_dataset there" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1703?src=pr&el=h1) Report\n> Merging [#1703](https://codecov.io/gh/huggingface/transformers/pull/1703?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/8a62835577a2a93642546858b21372e43c1a1ff8?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1703?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1703 +/- ##\n=======================================\n Coverage 85.14% 85.14% \n=======================================\n Files 94 94 \n Lines 13920 13920 \n=======================================\n Hits 11852 11852 \n Misses 2068 2068\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1703?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/1703?src=pr&el=footer). Last update [8a62835...7a8f3ca](https://codecov.io/gh/huggingface/transformers/pull/1703?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Instead of merging 5 commits I pushed a commit on master 68f7064 of which you are an author." ]

[]

[]

[ "My bad! Should have read the documentation for BertForMaskedLM code:\r\n```\r\nOutputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:\r\n **masked_lm_loss**: (`optional`, returned when ``masked_lm_labels`` is provided) ``torch.FloatTensor`` of shape ``(1,)``:\r\n Masked language modeling loss.\r\n **ltr_lm_loss**: (`optional`, returned when ``lm_labels`` is provided) ``torch.FloatTensor`` of shape ``(1,)``:\r\n Next token prediction loss.\r\n **prediction_scores**: ``torch.FloatTensor`` of shape ``(batch_size, sequence_length, config.vocab_size)``\r\n Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).\r\n **hidden_states**: (`optional`, returned when ``config.output_hidden_states=True``)\r\n list of ``torch.FloatTensor`` (one for the output of each layer + the output of the embeddings)\r\n of shape ``(batch_size, sequence_length, hidden_size)``:\r\n Hidden-states of the model at the output of each layer plus the initial embedding outputs.\r\n **attentions**: (`optional`, returned when ``config.output_attentions=True``)\r\n list of ``torch.FloatTensor`` (one for each layer) of shape ``(batch_size, num_heads, sequence_length, sequence_length)``:\r\n Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.\r\n```\r\n\r\nThe second tuple of tensors is the hidden_states from each layer + initial embedding layer. (num_hidden_layers=12) in config file. \r\nCan someone please confirm the layer order? Should I extract output from top or bottom?\r\n\r\nI have seen code examples using features from last 4 layers by indices like -1, -2, -3, -4, but just want to sure that it is what I should be doing as well.\r\n\r\nThanks!", "Okay. I understand the output now and selecting only the last hidden representation, but my gpu get filled if I set config.output_hidden_states=True for the fine-tuned LM model (as it returns the output from all hidden layers). How can I make it to return output from the selected layers only?", "Hi, thank you for sharing. Just want to ensure that the **last hidden state** could be derived by querying from indices = -1? I am super confused that there are 13 hidden states, I do not know whether the first one or last one is the last hidden states. Hope for your reply. Thanks!", "Yeah, I have the same confusion too. But considering the convention people used in other examples and/or issues, I am considering index = -1 belongs to the last hidden layer. \r\nAnyone is welcome to correct me if I am wrong. ", "@aurooj, @mingbocui you're both right: the last hidden state corresponds to the index -1. The index 0 corresponds to the output of the embedding layer, the index 1 corresponds to the output of the first layer, etc.", "@LysandreJik Thanks for your reply. Can you please help me with this?\r\n\r\n> Okay. I understand the output now and selecting only the last hidden representation, but my gpu get filled if I set config.output_hidden_states=True for the fine-tuned LM model (as it returns the output from all hidden layers). How can I make it to return output from the selected layers only?\r\n\r\n", "We currently have no way of returning the output of the selected layers only. You would have to manually re-implement the forward method of BERT and output only the layers you're interested in.\r\n\r\nIf you're looking to fine-tune BERT with the `bert-base-uncased` checkpoint and are concerned about the memory usage, maybe you could use DistilBERT with the `distilbert-base-uncased` checkpoint, which is based on BERT but has a much lower memory usage.", "Oh okay. Good to have some pointers to get what I want. \r\nThanks!", "@LysandreJik using the same approach as @aurooj mentioned, I'm getting tuple of 1 length, specifically this one\r\n> First one is of shape (B, SEQ_L, 30522)\r\n\r\nAlso, there are so many checkpoint-### folders created (starting from 500 to 8500).\r\nIs it the expected behavior?\r\n", "I'm not sure I understand your first question.\r\n\r\nYou can manage when the checkpoints are saved using the `--save_steps` argument. You can check the full list of arguments in the script (now renamed `run_language_modeling`) available [here](https://github.com/huggingface/transformers/blob/master/examples/run_language_modeling.py#L478-L613).", "Looking at the code was extremely helpful, I was able to continue fine-tuning from the saved checkpoints and understanding the arguments, though complete documentation of it would be ideal. \r\nAbout my first question, I did the same things @aurooj mentioned, but instead of 2 length tuple, I got 1 length tuple from `BertForMaskedLM` of shape `(B, SEQ_L, 30522)`", "You can show the arguments of that script alongside their documentation by running `python run_language_modeling.py --help`.\r\nConcerning what was returned by `BertForMaksedLM`, that seems normal as this is what is mentioned in the [documentation](https://huggingface.co/transformers/model_doc/bert.html#bertformaskedlm). The output tuple can contain up to 5 items, 4 of which are optional.", "My bad, even after reading the documentation. somehow it got in my mind that 3 were optional not 4.\r\nThank you. \r\nNote to self, \r\n\r\n> Reading documentation is paramount too.", "Hey @aurooj @nauman-chaudhary - I am very interested to see how you have done the fine-tuning for the masked LM task. Would really appreciate if you could share some insights. Thank you very much in advance! " ]

[]

[]

[ "We had a temporary issue with our `gpt2` model, should be fixed now.", "I reinstalled `transformers` (`pip install transformers`), and the error remains.", "> We had a temporary issue with our `gpt2` model, should be fixed now.\r\nI reinstalled transformers (pip install transformers), and the error remains.\r\n", "> We had a temporary issue with our `gpt2` model, should be fixed now.\r\n\r\n??", "use `force_download` option" ]

[ { "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" ]

[]

[]

[ "The first element is the embedding output.\r\n\r\nYes, use `hidden_layers[-1]`." ]

[]

[]

[ "I guess that's why add_prefix_space=True would be needed here." ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1697?src=pr&el=h1) Report\n> Merging [#1697](https://codecov.io/gh/huggingface/transformers/pull/1697?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/e85855f2c408f65a4aaf5d15baab6ca90fd26050?src=pr&el=desc) will **increase** coverage by `0.03%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1697?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1697 +/- ##\n=========================================\n+ Coverage 83.97% 84% +0.03% \n=========================================\n Files 105 97 -8 \n Lines 15570 14340 -1230 \n=========================================\n- Hits 13075 12047 -1028 \n+ Misses 2495 2293 -202\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1697?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3B5dG9yY2hfdXRpbHMucHk=) | `92.95% <0%> (-2.19%)` | :arrow_down: |\n| [transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3V0aWxzLnB5) | `89.64% <0%> (-1.22%)` | :arrow_down: |\n| [transformers/configuration\\_auto.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2NvbmZpZ3VyYXRpb25fYXV0by5weQ==) | `59.45% <0%> (-0.55%)` | :arrow_down: |\n| [transformers/modeling\\_tf\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX2JlcnQucHk=) | `96.08% <0%> (-0.24%)` | :arrow_down: |\n| [transformers/configuration\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2NvbmZpZ3VyYXRpb25fdXRpbHMucHk=) | `92.1% <0%> (-0.11%)` | :arrow_down: |\n| [transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3V0aWxzLnB5) | `92.68% <0%> (-0.05%)` | :arrow_down: |\n| [transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX2N0cmwucHk=) | `96.46% <0%> (-0.02%)` | :arrow_down: |\n| [transformers/configuration\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2NvbmZpZ3VyYXRpb25fZGlzdGlsYmVydC5weQ==) | `89.74% <0%> (ø)` | :arrow_up: |\n| [transformers/tests/tokenization\\_tests\\_commons.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rlc3RzL3Rva2VuaXphdGlvbl90ZXN0c19jb21tb25zLnB5) | `100% <0%> (ø)` | :arrow_up: |\n| [transformers/modeling\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX2Rpc3RpbGJlcnQucHk=) | `95.87% <0%> (ø)` | :arrow_up: |\n| ... and [20 more](https://codecov.io/gh/huggingface/transformers/pull/1697/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1697?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/1697?src=pr&el=footer). Last update [e85855f...226d137](https://codecov.io/gh/huggingface/transformers/pull/1697?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n" ]

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

[]

[ "I have encountered this embedding error and I found that wrapping the model in a tf.distribute strategy fixes it (extremely odd...)\r\n\r\nYou can wrap the model in `tf.distribute.OneDeviceStrategy` and it will work.", "Did you ever figure it out? ", "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/1695?src=pr&el=h1) Report\n> Merging [#1695](https://codecov.io/gh/huggingface/transformers/pull/1695?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/68f7064a3ea979cdbdadfed62ad655eac4c53463?src=pr&el=desc) will **increase** coverage by `0.07%`.\n> The diff coverage is `98.17%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1695?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1695 +/- ##\n==========================================\n+ Coverage 83.95% 84.03% +0.07% \n==========================================\n Files 94 94 \n Lines 13951 14021 +70 \n==========================================\n+ Hits 11713 11782 +69 \n- Misses 2238 2239 +1\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1695?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/modeling\\_tf\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX2Rpc3RpbGJlcnQucHk=) | `98.59% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_gpt2.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX2dwdDIucHk=) | `94.79% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX2JlcnQucHk=) | `96.6% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3hsbmV0LnB5) | `87.82% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX29wZW5haS5weQ==) | `96.04% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3V0aWxzLnB5) | `92.4% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX2N0cmwucHk=) | `97.75% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_transfo\\_xl.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3RyYW5zZm9feGwucHk=) | `92.21% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3hsbS5weQ==) | `90.39% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3JvYmVydGEucHk=) | `89.9% <ø> (ø)` | :arrow_up: |\n| ... and [11 more](https://codecov.io/gh/huggingface/transformers/pull/1695/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1695?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/1695?src=pr&el=footer). Last update [68f7064...00337e9](https://codecov.io/gh/huggingface/transformers/pull/1695?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "LGTM.\r\nFeel free to add the TF version or merge if you don't want to add them now." ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1694?src=pr&el=h1) Report\n> Merging [#1694](https://codecov.io/gh/huggingface/transformers/pull/1694?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/93d2fff0716d83df168ca0686d16bc4cd7ccb366?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1694?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1694 +/- ##\n=======================================\n Coverage 85.14% 85.14% \n=======================================\n Files 94 94 \n Lines 13920 13920 \n=======================================\n Hits 11852 11852 \n Misses 2068 2068\n```\n\n\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1694?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/1694?src=pr&el=footer). Last update [93d2fff...5b50eec](https://codecov.io/gh/huggingface/transformers/pull/1694?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Thank you for your PR. As I said in the issue, we’re still working on the summarization; you can follow the changes in the ˋexample-summarization` branch." ]

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

[]

[ "I have a fix for this in #1763 ", "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" ]

[]

[]

[ "Problem can be addressed by updating line 542 in transformers/modeling_tf_xlnet.py to :\r\n```python\r\n input_mask = 1.0 - tf.cast(attention_mask, dtype=dtype_float)\r\n```", "The above solution is not present in the current version. Why ?\r\nIt is still showing the same error as a result.", "@PradyumnaGupta the last version of transformers (v2.3.0) was released the 20th of december. The above solution was merged the 21st of december, and is therefore not in the latest pypi release.\r\n\r\nUntil we release a new version on pypi, feel free to install from source to obtain the most up-to-date bugfixes:\r\n\r\n```py\r\npip install git+https://github.com/huggingface/transformers\r\n```", "@LysandreJik \r\ni tried !pip install git+https://github.com/huggingface/transformers.git\r\ni was trying xlnet,here is the code used for modeling : \r\n\r\n`def create_model():\r\n q_id = tf.keras.layers.Input((MAX_SEQUENCE_LENGTH,), dtype=tf.int32) #, dtype=tf.int32\r\n a_id = tf.keras.layers.Input((MAX_SEQUENCE_LENGTH,), dtype=tf.int32)\r\n \r\n q_mask = tf.keras.layers.Input((MAX_SEQUENCE_LENGTH,), dtype=tf.float32)\r\n a_mask = tf.keras.layers.Input((MAX_SEQUENCE_LENGTH,), dtype=tf.float32) #, dtype=tf.float32\r\n \r\n q_atn = tf.keras.layers.Input((MAX_SEQUENCE_LENGTH,), dtype=tf.int32)\r\n a_atn = tf.keras.layers.Input((MAX_SEQUENCE_LENGTH,), dtype=tf.int32)\r\n from_tf=True\r\n #config = BertConfig() # print(config) to see settings\r\n config = XLNetConfig()\r\n config.d_inner = 3072\r\n config.n_head = 12\r\n config.d_model = 768\r\n config.n_layer = 12\r\n config.output_hidden_states = False # Set to True to obtain hidden states\r\n # caution: when using e.g. XLNet, XLNetConfig() will automatically use xlnet-large config\r\n \r\n # normally \".from_pretrained('bert-base-uncased')\", but because of no internet, the \r\n # pretrained model has been downloaded manually and uploaded to kaggle. \r\n #bert_model = TFBertModel.from_pretrained(BERT_PATH+'bert-base-uncased-tf_model.h5', config=config)\r\n #bert_model = TFBertModel.from_pretrained('xlnet-base-cased')\r\n #bert_model = XLNetModel(config)\r\n bert_model = TFXLNetModel.from_pretrained('xlnet-base-cased',config = config)\r\n #bert_model = XLNetForMultipleChoice.from_pretrained('xlnet-base-cased')\r\n \r\n # if config.output_hidden_states = True, obtain hidden states via bert_model(...)[-1]\r\n q_embedding = bert_model(q_id, attention_mask=q_mask, token_type_ids=q_atn)[0]\r\n a_embedding = bert_model(a_id, attention_mask=a_mask, token_type_ids=a_atn)[0]\r\n \r\n q = tf.keras.layers.GlobalAveragePooling1D()(q_embedding)\r\n a = tf.keras.layers.GlobalAveragePooling1D()(a_embedding)\r\n \r\n x = tf.keras.layers.Concatenate()([q, a])\r\n \r\n x = tf.keras.layers.Dropout(0.2)(x)\r\n \r\n x = tf.keras.layers.Dense(30, activation='sigmoid')(x)\r\n\r\n model = tf.keras.models.Model(inputs=[q_id, q_mask, q_atn, a_id, a_mask, a_atn,], outputs=x)\r\n \r\n return model`\r\n\r\ni was doing group k fold,so after 1st fold training i get TypeError: Expected int32, got 1.0 of type 'float' instead.\r\n\r\nhow can i solve this issue?" ]

[]

[]

[]

[]

[]

[ "duplicate of #1617 " ]

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

[]

[ "Yes, this is a known issue, `TransformerXL` is not traceable.\r\n\r\nFixing this is not on our short-term roadmap (cc @LysandreJik) but feel free to investigate and propose a solution in a PR if you want.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Any update? Does it work successfully?", "Same problem here.", "`hids = [t.transpose(0, 1).contiguous() for t in hids]` maybe work." ]

[]

[]

[ "There is an example in [the documentation](https://huggingface.co/transformers/examples.html#roberta-bert-and-masked-language-modeling)." ]

[]

[]

[ "As with any model hosted on our S3, you can do as follows to load one of the checkpoints:\r\n\r\n```py\r\nfrom transformers import BertModel\r\n\r\nmodel = BertModel.from_pretrained(\"bert-base-uncased\")\r\n```\r\n\r\nYou can find the list of pre-trained models in [the documentation](https://huggingface.co/transformers/pretrained_models.html).", "can we use bert-based-uncased model for QA(question-answer) if yes then how\r\n\r\nbecause model.predict(doc,q) giving error( **BertModel has no attribute predict**)\r\n\r\n\r\n", "The usage u need to reference on examples/run_squad.py and u will know\neverything on that code\n\nOn Mon, Nov 4, 2019 at 02:53 rhl2k <[email protected]> wrote:\n\n> can we use bert-based-uncased model for QA(question-answer) if yes then how\n>\n> because model.predict(doc,q) giving error( *BertModel has no attribute\n> predict*)\n>\n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/transformers/issues/1687?email_source=notifications&email_token=AIEAE4FNAI7STEJQG4J6GYDQR4M2XA5CNFSM4JHYQUQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC52AAA#issuecomment-549167104>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/AIEAE4GZOEPRXAX6NLWSAX3QR4M2XANCNFSM4JHYQUQQ>\n> .\n>\n", "You can find the documentation [here](https://huggingface.co/transformers/). The [quickstart](https://huggingface.co/transformers/quickstart.html) may be especially useful for you. As @pohanchi said, looking at the examples can also help in understanding the usage." ]

[]

[]

[ "Indeed thanks, fixed" ]

[]

[]

[ "Closing this in favor of #1684" ]

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

[]

[ "I'm having the same error", "There is a known (temporary) issue with our `gpt2` model – can you guys use `gpt2-medium` or `distilgpt2` instead for now?\r\n\r\ncc @LysandreJik @thomwolf @n1t0 @clmnt ", "Sure thing! Thanks for letting us know :)", "(should be fixed now)", "Thank you!", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1683?src=pr&el=h1) Report\n> Merging [#1683](https://codecov.io/gh/huggingface/transformers/pull/1683?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/fa735208c96c18283b8d2f3fcbfc3157bbd12b1e?src=pr&el=desc) will **increase** coverage by `0.9%`.\n> The diff coverage is `87.16%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1683?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1683 +/- ##\n=========================================\n+ Coverage 85.08% 85.99% +0.9% \n=========================================\n Files 94 98 +4 \n Lines 13920 14713 +793 \n=========================================\n+ Hits 11844 12652 +808 \n+ Misses 2076 2061 -15\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1683?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/tokenization\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rva2VuaXphdGlvbl94bG5ldC5weQ==) | `90.24% <ø> (ø)` | :arrow_up: |\n| [transformers/configuration\\_albert.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2NvbmZpZ3VyYXRpb25fYWxiZXJ0LnB5) | `100% <100%> (ø)` | |\n| [transformers/modeling\\_albert.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX2FsYmVydC5weQ==) | `81.73% <81.73%> (ø)` | |\n| [transformers/modeling\\_tf\\_albert.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX2FsYmVydC5weQ==) | `84.46% <84.46%> (ø)` | |\n| [transformers/tokenization\\_albert.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rva2VuaXphdGlvbl9hbGJlcnQucHk=) | `89.74% <89.74%> (ø)` | |\n| [transformers/tests/modeling\\_tf\\_albert\\_test.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rlc3RzL21vZGVsaW5nX3RmX2FsYmVydF90ZXN0LnB5) | `94.39% <94.39%> (ø)` | |\n| [transformers/tests/modeling\\_albert\\_test.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rlc3RzL21vZGVsaW5nX2FsYmVydF90ZXN0LnB5) | `95.04% <95.04%> (ø)` | |\n| [transformers/tests/tokenization\\_albert\\_test.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rlc3RzL3Rva2VuaXphdGlvbl9hbGJlcnRfdGVzdC5weQ==) | `97.43% <97.43%> (ø)` | |\n| [transformers/tests/modeling\\_bert\\_test.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rlc3RzL21vZGVsaW5nX2JlcnRfdGVzdC5weQ==) | `96.38% <0%> (-0.54%)` | :arrow_down: |\n| [transformers/configuration\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2NvbmZpZ3VyYXRpb25fdXRpbHMucHk=) | `98.66% <0%> (-0.02%)` | :arrow_down: |\n| ... and [20 more](https://codecov.io/gh/huggingface/transformers/pull/1683/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1683?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/1683?src=pr&el=footer). Last update [fa73520...afef0ac](https://codecov.io/gh/huggingface/transformers/pull/1683?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Will there be models for classification?", "Yes, at the moment there is AlbertForSequenceClassification and there may be more soon", "@LysandreJik Thanks for adding this :+1: \r\n\r\nI've one question: the ALBERT team did release version 2 of their models yesterday, see: \r\n\r\nhttps://github.com/google-research/google-research/commit/2ba150bef51fcedcfda31f16321264300f201a8d\r\n\r\nAre these updated models available on S3 yet 🤔", "V2 just use 0 dropout and lr to 1e-5, the architecture didn’t change, so\nmaybe it just need time to transfer model to here.\n\nOn Sat, Nov 2, 2019 at 21:06 Stefan Schweter <[email protected]>\nwrote:\n\n> @LysandreJik <https://github.com/LysandreJik> Thanks for adding this 👍\n>\n> I've one question: the ALBERT team did release version 2 of their models\n> yesterday, see:\n>\n> google-research/google-research@2ba150b\n> <https://github.com/google-research/google-research/commit/2ba150bef51fcedcfda31f16321264300f201a8d>\n>\n> Are these updated models available on S3 yet 🤔\n>\n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/transformers/pull/1683?email_source=notifications&email_token=AIEAE4FY2TZTBZ5YV6P5ARLQRV3ONA5CNFSM4JHPMB22YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC43PFI#issuecomment-549042069>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/AIEAE4BGB37TAZRZI6CQ3HLQRV3ONANCNFSM4JHPMB2Q>\n> .\n>\n", "I'm not sure how heavily you want to take advantage of Apex when available, but Apex does provide a fused implementation of Lamb. https://nvidia.github.io/apex/optimizers.html#apex.optimizers.FusedLAMB", "@stefan-it the ALBERT v2 models are now available on the S3. You can access them using `albert-{base,large,xlarge,xxlarge}-v2` identifiers!\r\n\r\n@BramVanroy Indeed, thanks! For now we're focusing more on the model implementation rather than the optimizers; the optimizers can be obtained from other libraries (such as apex) and used with the models from `transformers` so it is not a priority right now.", "Hi @LysandreJik thanks for the model versioning :)\r\n\r\nJust a few notes from my (early) experiments with this ALBERT implementation.\r\n\r\nI used a feature-based approach in Flair for NER on English CoNLL dataset. More precisely I used embeddings from all layers (incl. word embedding layer) + scalar mix over all layers to get an embedding for the first subtoken of each token. Results for the base model are \"ok\": 93.13 (dev) and 89.17 (test) compared to BERT base: 94.74 (dev) and 91.38 (test).\r\n\r\nAfter work I implemented an `AlbertForTokenClassification` class and added it to the `run_ner.py` example script. With default parameters 88.06 (dev) and 82.94 (test) could be achieved (so there's large room for improvement in my implementation 😅).\r\n\r\nBut: I also tested the `large` and `xlarge` models. Using Flair (and all 24 + 1 layers with scalar mix) the F-score dropped to 45% on test set?! The fine-tuning experiment (with `run_ner.py`) yields 0% for F-score 😂\r\n\r\nI'm not sure what's going on with the > `large` models 🤔 (I did experiments for NER only)", "HI @stefan-it,\r\nI won't suggest ALBERT for NER task. As of now, all the released weights are trained using lowering the sentence. NER model is usually built using Cased models. BERT NER is based on bert-base/large-cased. ", "For BERT the difference was ~ 0.2 to 0.3% on CoNLL (base and large model, feature-base approach) - but I'll further investigate the large ALBERT models 😅", "Hi @stefan-it, thanks for your study! I could indeed replicate the issue with ALBERT-large which has very bad results on SQuAD after being fine-tuned on it. I'm looking into it today and I'll update you on the progress.", "I curious about how worse for squad\n\nOn Wed, Nov 6, 2019 at 21:57 Lysandre Debut <[email protected]>\nwrote:\n\n> Hi @stefan-it <https://github.com/stefan-it>, thanks for your study! I\n> could indeed replicate the issue with ALBERT-large which has very bad\n> results on SQuAD after being fine-tuned on it. I'm looking into it today\n> and I'll update you on the progress.\n>\n> —\n> You are receiving this because you commented.\n>\n>\n> Reply to this email directly, view it on GitHub\n> <https://github.com/huggingface/transformers/pull/1683?email_source=notifications&email_token=AIEAE4G7RYHSXRPZ5Y2MWJ3QSLENDA5CNFSM4JHPMB22YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDGTPTQ#issuecomment-550320078>,\n> or unsubscribe\n> <https://github.com/notifications/unsubscribe-auth/AIEAE4AOIEITNSQM6EO55FDQSLENDANCNFSM4JHPMB2Q>\n> .\n>\n", "I've looked into it and there seems to be an error with the models `large`, `xlarge` and `xxlarge` version 2. The `base` models of both versions as well as the larger models of version 1 seem to work correctly (there was an issue that was fixed this morning).\r\n\r\n@pohanchi based on a single epoch just to check the models were learning, `albert-base-v2` obtains 82.5 exact and 89.9 F1 and `albert-large-v1` obtains 82.8 exact and 90 F1\r\n\r\nI'm looking into the V2 models now.", "Side question: for how long are you planning to support Python 2? Considering it's as good as EOL and all that.", "@BramVanroy, as long as Google outputs models in Python 2 we'll continue to maintain it, and probably for a few months after that!", "(That's only for the core code though. Examples and scripts are already Python 3 only AFAIK)", "@LysandreJik Great job! Could you elaborate on why you added and removed the Lamb optimizer? Is there any issue with this implementation?", "Great work all. Tried it and noticed a few things, that may or may not be issues, but I'll post the details here just in case.:\r\n\r\n- doesn't work in pytorch 1.1.0, does in 1.2.0. This is probobly OK as 1.2.0 is the version listed in requirements.dev.txt\r\n - The error is for line [`w = self.dense.weight.T` \"Parameter self.dense.weight has not attribute T\"](https://github.com/huggingface/transformers/blob/06fc337815/transformers/modeling_albert.py#L206)\r\n- You may be aware of this but it doesn't work with fp16 O1 yet \r\n - `RuntimeError: Expected object of scalar type Half but got scalar type Float for argument #2 'mat2'`\r\n - reffering to line [`projected_context_layer = torch.einsum(\"bfnd,ndh->bfh\", context_layer, w) + b`](https://github.com/huggingface/transformers/blob/06fc337815/transformers/modeling_albert.py#L209). Specifically context_layer is half, w while b are float.\r\n - these changes fix fp16 O1:\r\n - `w = self.dense.weight.T.view(self.num_attention_heads, self.attention_head_size, self.hidden_size).to(context_layer.dtype)`\r\n - `b = self.dense.bias.to(context_layer.dtype)`\r\n- it does run without fp16 :)", "Thank you very much for your great work!@LysandreJik\r\n\r\nI have tried running with the run_glue.py file to obtain the test accuracy for MNLI task. \r\n**Without training, just evaluation**. Using the **albert-base-v1** model from the S3, I have obtained **31.8% accuracy** for MNLI, which differs greatly from the ALBERT paper. However, after training with the default hyperparameters specified in the run_glue.py file, I obtained an accuracy which is similar to the paper. \r\n\r\nI am a new guy to NLP, previously working in CV. I am wondering does the S3 model contains the pretrained weight for ALBERT? Since without training, the result differs greatly from the papers.\r\n\r\n", "@panaali LAMB was first added but I couldn't manage to make it work immediately, so as the authors said that there was not a huge difference between ADAM and LAMB, I removed it and fine-tuned with ADAM instead. As I told Bram a few messages ago: \"For now we're focusing more on the model implementation rather than the optimizers; the optimizers can be obtained from other libraries (such as apex) and used with the models from transformers so it is not a priority right now.\". I believe you can use existing LAMB implementations with our models and it will work out of the box, such as [cybertronai's implementation.](https://github.com/cybertronai/pytorch-lamb), or from apex.\r\n\r\n@wassname Thank you for your comments, I'm looking into that.\r\n\r\n@astrongstorm the model as it is saved on our S3 only contains the base model, without the classification head (similarly to most of the models hosted on our S3). Before using them, it is essential to fine-tune them so that the classification head may be trained on the actual dataset.", "@LysandreJik Thanks for your reply! There is still one confusing point about the S3 model. I am wondering in S3 model, does it contain both hyperparameter and the parameters for the model, or it only contains one of them. ", "The S3 holds several files: \r\n- The configuration files which holds what you might call the hyper-parameters: number of inner group, hidden size, vocabulary size, etc.\r\n- The model files which contain parameters for pytorch (albert-xxx-pytorch_model.bin) and for tensorflow (albert-xxx-tf_model.h5)\r\n- the tokenizer vocabulary files", "Thanks @LysandreJik for the great work! I am looking forward to use it. When will this branch be merged to the master or is there a timeline?", "Hi @jimmycode, it should be merged at the end of the week.", "I think the v1 models are looking good (v2 are currently very bad) - I did some comparisons for NER (CoNLL-2003):\r\n\r\n| Model | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Avg.\r\n| ------------------------ | ----- | ----- | ----- | ----- | ----- | ---------\r\n| BERT large, cased (Dev) | 95.69 | 95.47 | 95.77 | 95.86 | 95.91 | 95.74\r\n| BERT large, cased (Test) | 91.73 | 91.17 | 91.77 | 91.22 | 91.46 | **91.47**\r\n| ALBERT xxlarge, uncased, v1 (Dev) | 95.35 | 95.42 | 95.17 | 95.16 | 95.39 | 95.30\r\n| ALBERT xxlarge, uncased, v1 (Test) | 91.49 | 91.60 | 91.69 | 90.88 | 91.27 | 91.39\r\n\r\n(although cased vs. uncased is not really a fair comparison)\r\n\r\nI'll prepare a PR when the ALBERT code was merged to support a \"for-token-classification\" interface that can be used in the `run_ner.py` example.", "Hi Thanks for the quick addition.\r\n\r\nDoes ALBERT require the usage of AlbertTokenizer? or can we simply use BERTTokenizer?\r\n\r\nBecause otherwise, there might be a need to re-process all data using AlbertTokenizer.", "Hi @jshin49. yes the `AlbertTokenizer` should be used: BERT uses word pieces, ALBERT uses a sentence piece model. The output of the tokenizer implementations is totally different:\r\n\r\n```python\r\nfrom transformers import AlbertTokenizer, BertTokenizer\r\n\r\nbert_tokenizer = BertTokenizer.from_pretrained(\"bert-base-uncased\")\r\nalbert_tokenizer = AlbertTokenizer.from_pretrained(\"albert-base-v1\")\r\n\r\nsentence = \"neuschwanstein near munich\"\r\n```\r\n\r\nOutputs:\r\n\r\n```python\r\nIn [9]: bert_tokenizer.tokenize(sentence)\r\nOut[9]: ['ne', '##us', '##ch', '##wan', '##stein', 'near', 'munich']\r\n\r\nIn [10]: albert_tokenizer.tokenize(sentence)\r\nOut[10]: ['▁neu', 'sch', 'wan', 'stein', '▁near', '▁munich']\r\n```", "Thank you @LysandreJik for the great work! , Do you have any plans to add multilingual ALBERT?", "> After work I implemented an `AlbertForTokenClassification` class and added it to the `run_ner.py` example script.\r\n\r\n@stefan-it could you add this as PR?\r\n", "Oh, I totally forgot that 😅\r\n\r\nI can look into it the next days :)" ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1682?src=pr&el=h1) Report\n> Merging [#1682](https://codecov.io/gh/huggingface/transformers/pull/1682?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/7daacf00df433621e3d3872a9f3bb574d1b00f5a?src=pr&el=desc) will **increase** coverage by `1.67%`.\n> The diff coverage is `35.84%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1682?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1682 +/- ##\n=========================================\n+ Coverage 84.03% 85.7% +1.67% \n=========================================\n Files 94 92 -2 \n Lines 14021 13704 -317 \n=========================================\n- Hits 11782 11745 -37 \n+ Misses 2239 1959 -280\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1682?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/configuration\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2NvbmZpZ3VyYXRpb25fZGlzdGlsYmVydC5weQ==) | `89.74% <ø> (ø)` | :arrow_up: |\n| [transformers/tokenization\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rva2VuaXphdGlvbl9kaXN0aWxiZXJ0LnB5) | `100% <ø> (ø)` | :arrow_up: |\n| [transformers/modeling\\_distilbert.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX2Rpc3RpbGJlcnQucHk=) | `95.8% <ø> (-0.03%)` | :arrow_down: |\n| [transformers/data/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2RhdGEvX19pbml0X18ucHk=) | `100% <100%> (ø)` | :arrow_up: |\n| [transformers/data/processors/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2RhdGEvcHJvY2Vzc29ycy9fX2luaXRfXy5weQ==) | `100% <100%> (ø)` | :arrow_up: |\n| [transformers/tokenization\\_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rva2VuaXphdGlvbl94bG0ucHk=) | `83.6% <100%> (+0.39%)` | :arrow_up: |\n| [transformers/data/metrics/\\_\\_init\\_\\_.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2RhdGEvbWV0cmljcy9fX2luaXRfXy5weQ==) | `34.04% <25%> (-0.85%)` | :arrow_down: |\n| [transformers/data/processors/xnli.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2RhdGEvcHJvY2Vzc29ycy94bmxpLnB5) | `31.11% <31.11%> (ø)` | |\n| [transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX2N0cmwucHk=) | `95.45% <0%> (-1.01%)` | :arrow_down: |\n| [transformers/modeling\\_transfo\\_xl.py](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RyYW5zZm9feGwucHk=) | `75.16% <0%> (-0.75%)` | :arrow_down: |\n| ... and [39 more](https://codecov.io/gh/huggingface/transformers/pull/1682/diff?src=pr&el=tree-more) | |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1682?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/1682?src=pr&el=footer). Last update [7daacf0...828058a](https://codecov.io/gh/huggingface/transformers/pull/1682?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Great addition! As you may have seen, we've been slowly moving the utils from our examples folder to the actual transformer library. \r\nWe've done so with GLUE and we have put the processors directly in `transformers/data/processors/glue.py`. This way the processors may be used as a component of the library rather than as a utility class/function.\r\n\r\nDo you think you could do the same for XNLI? It would require you to create a file `transformers/data/processors/xnli.py` and put the `XnliProcessor` there.", "Concerning the documentation, if you choose to add `XnliProcessor` to the processors it would be great to add it to the processors documentation in `docs/source/main_classes/processors.rst`", "This one looks ready to be merged. @thomwolf ?" ]

[]

[]

[ "You are right the release is wrong, it should be `<s> SEQUENCE_0 </s></s> SEQUENCE_1 </s>`. I just updated it; thank you!" ]

[]

[]

[ "It's not in a pip released version yet so you need to pull from master if you want to use it for now.\r\n\r\nWe'll do a release soon.", "Thanks for the info. Do you have an estimation when that pip release would be, @julien-c ?", "Reviving this thread. I just cloned 2.2.2 from the master and updated `transformers`. `distilroberta-base` is still not available. Am I missing something? Thanks, you all!..\r\n\r\n> OSError: Model name 'distilroberta-base' was not found in model name list (distilbert-base-uncased, distilbert-base-uncased-distilled-squad, distilbert-base-german-cased, distilbert-base-multilingual-cased)\r\n ", "What are the exact commands you typed @oersoy1? ", "@julien-c OK, I found out what is happening and will document here just in case someone else falls into the same trap. I wrote my custom script similar to run_glue.py. I was passing `distilbert` to the `args.model_type` argument and the `model_name` got `distilroberta-base` assigned. I assumed that it was a subclass of `distilbert` models so the type intuitively looked as if it should have been distilbert. [This list](https://huggingface.co/transformers/pretrained_models.html) certainly gives me that impression.\r\n\r\nHowever, looking at the distillation [examples](https://github.com/huggingface/transformers/tree/master/examples/distillation), I realized the model type needs to be `roberta` not `distilbert`. \r\n\r\nIt is a little bit confusing but regardless, I got `distilroberta-base` working and it gave me great results. \r\n\r\nThanks a lot!", "Ah, yeah, you are correct. Moved the model shortcut in the table in ac1b449\r\n", "No worries. I see that you have corrected the pretrained models list and moved `distilroberta-base` under `roberta` which was the main problem for me. Updating all documentation when you make the changes could be difficult, especially when the gatekeepers for a specific document is different than the ones making the change. " ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1679?src=pr&el=h1) Report\n> Merging [#1679](https://codecov.io/gh/huggingface/transformers/pull/1679?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/fa735208c96c18283b8d2f3fcbfc3157bbd12b1e?src=pr&el=desc) will **increase** coverage by `0.05%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1679?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1679 +/- ##\n==========================================\n+ Coverage 85.08% 85.14% +0.05% \n==========================================\n Files 94 94 \n Lines 13920 13920 \n==========================================\n+ Hits 11844 11852 +8 \n+ Misses 2076 2068 -8\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1679?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/modeling\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1679/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3V0aWxzLnB5) | `92.44% <ø> (+1.45%)` | :arrow_up: |\n| [transformers/tests/modeling\\_common\\_test.py](https://codecov.io/gh/huggingface/transformers/pull/1679/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rlc3RzL21vZGVsaW5nX2NvbW1vbl90ZXN0LnB5) | `76.49% <0%> (+0.59%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1679?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/1679?src=pr&el=footer). Last update [fa73520...ac29353](https://codecov.io/gh/huggingface/transformers/pull/1679?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "This looks good to me!", "Yes indeed, thanks!" ]

[]

[]

[ "I have read the comment part of the reference code:\r\n```\r\n# Download to temporary file, then copy to cache dir once finished.\r\n# Otherwise you get corrupt cache entries if the download gets interrupted.\r\n```\r\nSo I would change my proposal:\r\n* Either let it be configurable to skip the tmp folder and download directly to the cache folder -> the user will know what he is doing and will know that the asset could get corrupted\r\n* Or check the file in the cache before usage - e.g. using checksums\r\n* Or write \"download has started\" and \"download has finished\" information to the meta data file that can be checked before asset usage.\r\n", "I would propose to download to the cache_dir with a specific temporary name (like a `.part` suffix) and copy + rename at the end.\r\n\r\nProbably best to activate that with an option `use_cache_dir_as_tmp`. To not clutter the cache dir with temporary files in the default settings.\r\n\r\nDo you want to submit a PR for that? Would be happy to review it", "Yes that is also a good approach. For now, we seem to be okay with this limitation, but I'll do a pr if we face this as an issue or have some free time.", "Same problem here. On cluster, /tmp folder is small. Keep getting no space on device.", "I fixed this in b67fa1a8d2302d808ecb9d95355181eaf21ee3b6.", "Until there's a release with this fix, you can set $TMPDIR to an appropriate location if /tmp is too small.", "Cool, thank you @aaugustin !" ]

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

[]

[ "run_multiple_choice.py will be a good choice.\r\nin case ‘bert’, it uses \r\nhttps://github.com/huggingface/transformers/blob/master/transformers/modeling_bert.py#L1021\r\n\r\nbut i think your problem formulation is odd.\r\nwhat about classifying ‘request_*’ intent as normal classification problem and slot tagging ‘inform_*’ as sequence classification?\r\n", "hi. i just start to learn bert for a very short time,so i could not have learned it clearly. \r\n‘request_’ intent classifying can be very different with the normal classification problem?\r\ni have not understand this.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

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

[ "@thomwolf another encoder-decoder", "Was released today: https://github.com/pytorch/fairseq/tree/master/examples/bart 🎉", "Let me know if you guys plan to add xsum/eli5/cnn-dm ft with our released bart into hugging face. ", "Is there any news on this?", "any progress on this one? also thanks :)", "I'm getting started on this Feb 4!" ]

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

[]

[ "im asking for the same thing", "I also need this. experts, please help/guide.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

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

[]

[ "Hi!\r\n\r\nThanks for pointing these out. The summarization is still work in progress and should be included in the next release. Latest changes are in the `example-summarization` branch.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[]

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

[]

[ "Hi @h56cho,\r\n\r\nThe loss is actually returned if labels are present.\r\n\r\nCheck https://github.com/huggingface/transformers/blob/master/transformers/modeling_transfo_xl.py#L793", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

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

[]

[ "Hi, I think this was fixed by #1601, could you try now by cloning and installing from master?", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Thanks @stefan-it, feel free to give your opinion on the explanation/templates as well, always happy to have your feedback" ]

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

[]

[ "Hello @ANSHUMAN87,\r\n\r\nCould you share the command you're using (and the error you get)?\r\nYou should have at least these arguments: `--model_type distilbert --model_name_or_path <your_model_path>`.\r\n\r\nVictor", "I have mentioned below the steps i followed.\r\n\r\nStep 1: python3 train.py --student_type distilbert --student_config training_configs/distilbert-base-uncased.json --teacher_type bert --teacher_name bert-base-uncased --alpha_ce 5.0 --alpha_mlm 2.0 --alpha_cos 1.0 --mlm --freeze_pos_embs --dump_path serialization_dir/my_first_training --data_file binarized_text.bert-base-uncased.pickle --token_counts token_counts.bert-base-uncased.pickle --force\r\n\r\nResult: Successful\r\n\r\nStep 2: python3 run_glue.py --model_type distilbert --model_name_or_path distillation/serialization_dir/my_first_training/ --task_name CoLA --do_eval --do_lower_case --data_dir /home/anshuman/3/GLUE-Dataset/glue_data/CoLA/ --max_seq_length 128 --output_dir distillation/serialization_dir/my_first_training/\r\n\r\nError: \r\n\r\n\r\n\r\n", "Thank you, I understand what's happening now.\r\n\r\nIt happens because you're not launching any training (`--do_train`) before evaluating. What happens in `run_glue.py` is that when you do the evaluation, the tokenizer is loaded from the one saved in `output_dir` (see [here](https://github.com/huggingface/transformers/blob/master/examples/run_glue.py#L524)). The latter has been saved a few lines before ([here](https://github.com/huggingface/transformers/blob/master/examples/run_glue.py#L510)) in `do_train`... So basically, you're trying to load something that doesn't exist yet...\r\n\r\nOne way to quickly bypass this is: a/ adding `--do_train --num_train_epochs 0.0`, b/ set the return to `global_step, tr_loss / 1` (see [here](https://github.com/huggingface/transformers/blob/master/examples/run_glue.py#L207)) to avoid division by 0.\r\n\r\nOf course, by doing that, you're evaluating on a GLUE task a model that hasn't been finetuned for the GLUE task in question (i.e. you're doing zero-shot).\r\n\r\nAlso, I recommend to use a different `output_dir` in the `run_glue.py` command: run_glue will overwrite your pre-training (step 1) when saving the model under the name `pytorch_model.bin`.", "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", "> Thank you, I understand what's happening now.\r\n> \r\n> It happens because you're not launching any training (`--do_train`) before evaluating. What happens in `run_glue.py` is that when you do the evaluation, the tokenizer is loaded from the one saved in `output_dir` (see [here](https://github.com/huggingface/transformers/blob/master/examples/run_glue.py#L524)). The latter has been saved a few lines before ([here](https://github.com/huggingface/transformers/blob/master/examples/run_glue.py#L510)) in `do_train`... So basically, you're trying to load something that doesn't exist yet...\r\n> \r\n> One way to quickly bypass this is: a/ adding `--do_train --num_train_epochs 0.0`, b/ set the return to `global_step, tr_loss / 1` (see [here](https://github.com/huggingface/transformers/blob/master/examples/run_glue.py#L207)) to avoid division by 0.\r\n> \r\n> Of course, by doing that, you're evaluating on a GLUE task a model that hasn't been finetuned for the GLUE task in question (i.e. you're doing zero-shot).\r\n> \r\n> Also, I recommend to use a different `output_dir` in the `run_glue.py` command: run_glue will overwrite your pre-training (step 1) when saving the model under the name `pytorch_model.bin`.\r\n\r\nHello,@VictorSanh,\r\nI have completed model train in pytorch. But how can I use the trained model to do some new test on a new test.tvs? my run.sh is:\r\nexport TASK_NAME=mytask\r\n\r\npython src/run_glue.py \\\r\n --model_name_or_path ch/ \\\r\n --task_name $TASK_NAME \\\r\n --do_predict \\\r\n --data_dir data/ \\\r\n --max_seq_length 128 \\\r\n --output_dir saved_test_moels/ \\\r\n --overwrite_cache\r\nit doesn't work. What should I change? Thank you." ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1668?src=pr&el=h1) Report\n> Merging [#1668](https://codecov.io/gh/huggingface/transformers/pull/1668?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/079bfb32fba4f2b39d344ca7af88d79a3ff27c7c?src=pr&el=desc) will **not change** coverage.\n> The diff coverage is `75%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1668?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1668 +/- ##\n======================================\n Coverage 85.9% 85.9% \n======================================\n Files 91 91 \n Lines 13653 13653 \n======================================\n Hits 11728 11728 \n Misses 1925 1925\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1668?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/modeling\\_tf\\_xlm.py](https://codecov.io/gh/huggingface/transformers/pull/1668/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3hsbS5weQ==) | `90.39% <75%> (ø)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1668?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/1668?src=pr&el=footer). Last update [079bfb3...842f3bf](https://codecov.io/gh/huggingface/transformers/pull/1668?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Awesome, thanks a lot @tlkh " ]

[]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1667?src=pr&el=h1) Report\n> Merging [#1667](https://codecov.io/gh/huggingface/transformers/pull/1667?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/079bfb32fba4f2b39d344ca7af88d79a3ff27c7c?src=pr&el=desc) will **increase** coverage by `0.02%`.\n> The diff coverage is `n/a`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1667?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1667 +/- ##\n==========================================\n+ Coverage 85.9% 85.92% +0.02% \n==========================================\n Files 91 91 \n Lines 13653 13653 \n==========================================\n+ Hits 11728 11732 +4 \n+ Misses 1925 1921 -4\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1667?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/file\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1667/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2ZpbGVfdXRpbHMucHk=) | `76.37% <0%> (+2.19%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1667?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/1667?src=pr&el=footer). Last update [079bfb3...2669079](https://codecov.io/gh/huggingface/transformers/pull/1667?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "Great, thank you @tlkh ! Feel free to add a link to your spreadsheet in the documentation." ]

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

[]

[ "Going through the source code, the sequence is actually truncated.\r\n\r\nhttps://github.com/huggingface/transformers/blob/079bfb32fba4f2b39d344ca7af88d79a3ff27c7c/transformers/tokenization_utils.py#L846-L853\r\n\r\nThe warning occurs because `encode_plus` calls `convert_tokens_to_ids` _first_ and only then the IDs are truncated. The warning originates from `convert_tokens_to_ids` before truncation has happened. This is quite confusing indeed, since in the end result the IDs _are_ truncated.\r\n\r\nPerhaps one of the maintainers can chip in.", "> Going through the source code, the sequence is actually truncated.\r\n> \r\n> https://github.com/huggingface/transformers/blob/079bfb32fba4f2b39d344ca7af88d79a3ff27c7c/transformers/tokenization_utils.py#L846-L853\r\n> \r\n> The warning occurs because `encode_plus` calls `convert_tokens_to_ids` _first_ and only then the IDs are truncated. The warning originates from `convert_tokens_to_ids` before truncation has happened. This is quite confusing indeed, since in the end result the IDs _are_ truncated.\r\n> \r\n> Perhaps one of the maintainers can chip in.\r\n\r\nSo, it safe to use or not?", "This should have been patched in release 2.2.0.", "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" ]

[]

[]

[ "Please don't post issues like this. I'm sure the maintainers work as hard as they can. Asking them to _work faster_ doesn't help. In fact, adding these kind of non-issues only distract the maintainers from actually working on the actual issues at hand. Please close this question." ]

[]

[]

[ "`.to()` is not an in-place operation. You should use `model = model.to('cpu')`. If that doesn't work, it might be that you need to access the module as part of the DataParallel object, like this:\r\n\r\n```python\r\nmodel = model.module.to('cpu')\r\n```", "Ahh gotcha. Thanks for the quick reply!" ]

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

[]

[ "I found a bug, it's TPU related. For some reason, after I move the mode to TPU, using `model = model.to(device)`, the weights become decoupled. Then I save decoupled weights and during restore it ties them again. It loads correctly, it just doesn't expect tied weights to be different.\r\n\r\nThe workaround is to tie weights again after moving model to the TPU.\r\n\r\n```\r\n model = model.to(args.device)\r\n model.tie_weights()\r\n```\r\n\r\n", "I'm sorry to reopen this issue, Davide Libenzi suggesting this is the model issue, not the PyTorch XLA issue. I'm a bit of a tired from debugging and I'm happy with the workaround. \r\n\r\nYou can find details here https://github.com/pytorch/xla/issues/1245\r\n", "Ok, do you think we should fix this upstream in our library?\r\nI'm not super excited about overwriting PyTorch built-in `nn.Module.apply()` method.", "It feels to me, the Pytorch/XLA is more appropriate place for fix, since Pytorch/Cuda have that behavior and the fix will make two libraries consistent. \r\nBut I don't feel competent neither in Pytorch/XLA nor in Transformers to insist. \r\nIt would be great to have somebody from Transformers to talk to Pytorch/XLA over this issue. ", "Ok, we'll try to push this up-stream", "The PyTorch community decided it's more appropriate to tie weights after moving the model to the device (TPU/GPU/CPU). I believe it's worth to fix the model accordingly. \r\n\r\nhttps://github.com/pytorch/xla/issues/1245#issuecomment-552559970\r\nhttps://github.com/pytorch/xla/pull/1335", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "Hello, stale bot, it would be great to keep it open.", "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" ]

[]

[]

[ "We've seen this issues also with other tokenizers, like XLNet. \r\n\r\nIt would be awesome to have a unified tokenization strategy (across all `Tokenizer` classes) that return `unk_token` in these cases. And of course we should discuss other possibilities here :)\r\n", "@voidful this behavior arises because `bert-base-multilingual-uncased` is lower-casing the input (as the name indicates) and as such remove accents.\r\n\r\nYour character is classified as an accent in the Unicode category database (see \"Mn\" [here](https://www.fileformat.info/info/unicode/category/index.htm)).\r\n\r\nTo fix this behavior, use the recommended multilingual model for Bert: `bert-base-multilingual-cased` instead of the one you are using (see the list of models and the recommended ones [here](https://huggingface.co/transformers/pretrained_models.html))\r\n\r\n@stefan-it I think the other issues you are referring to are likely different from this one.\r\nFeel free to open another issue if you want us to investigate them in detail.", "Thank you for your help! It really solve the problem !" ]

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

[]

[ "I think the [BERTology](https://huggingface.co/transformers/bertology.html) section could help, especially the [run_bertology.py](https://github.com/huggingface/transformers/blob/master/examples/run_bertology.py) script can perform pruning and includes other useful functions :)", "I am beginner to BERT, can you please tell me to turn off the second head in the ninth layer for example.\r\nHere is my model config\r\n\r\nconfig = BertConfig.from_pretrained(\"bert-base-uncased\",output_attentions=True,output_hidden_states=True, num_labels=2)\r\n\r\nmodel = BertForSequenceClassification.from_pretrained(\"bert-base-uncased\", config= config)\r\nmodel.cuda()", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

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

[]

[ "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n", "@zhongpeixiang do you have any info about finetuning ctrl?", "@saippuakauppias No, in the end I chose the original CTRL repo from Salesforce to finetune.", "Hey @zhongpeixiang, could you share some more information on how you fine-tuned the CTRL? I am also struggling to fine tune it using transformers. ", "@ludoro I followed this repo to fine-tune the CTRL: https://github.com/salesforce/ctrl", "also struggling through the fine-tuning of CTRL, if someone can show a notebook or just the code to do that, it will help a lot!", "> also struggling through the fine-tuning of CTRL, if someone can show a notebook or just the code to do that, it will help a lot!\r\n\r\nhttps://github.com/salesforce/ctrl/tree/master/training_utils" ]

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

[]

[ "Hi, the models are not fine-tuned on the fly. Language models like GPT-2 are very context-aware and are strong at generating words related to the inputs they were given. \r\n\r\nWe are not training the models in that demo, we are only using them for inference.", "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 } ]

[]

[ "If you're looking to fine-tune it on an MLM task you could simply re-use some parts of the `run_lm_finetuning.py` script to do it. ", "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 } ]

[]

[ "# [Codecov](https://codecov.io/gh/huggingface/transformers/pull/1657?src=pr&el=h1) Report\n> Merging [#1657](https://codecov.io/gh/huggingface/transformers/pull/1657?src=pr&el=desc) into [master](https://codecov.io/gh/huggingface/transformers/commit/079bfb32fba4f2b39d344ca7af88d79a3ff27c7c?src=pr&el=desc) will **decrease** coverage by `1.39%`.\n> The diff coverage is `100%`.\n\n[](https://codecov.io/gh/huggingface/transformers/pull/1657?src=pr&el=tree)\n\n```diff\n@@ Coverage Diff @@\n## master #1657 +/- ##\n=========================================\n- Coverage 85.9% 84.51% -1.4% \n=========================================\n Files 91 91 \n Lines 13653 13654 +1 \n=========================================\n- Hits 11728 11539 -189 \n- Misses 1925 2115 +190\n```\n\n\n| [Impacted Files](https://codecov.io/gh/huggingface/transformers/pull/1657?src=pr&el=tree) | Coverage Δ | |\n|---|---|---|\n| [transformers/modeling\\_bert.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX2JlcnQucHk=) | `88.2% <100%> (+0.02%)` | :arrow_up: |\n| [transformers/modeling\\_tf\\_pytorch\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3RmX3B5dG9yY2hfdXRpbHMucHk=) | `9.85% <0%> (-83.1%)` | :arrow_down: |\n| [transformers/tests/modeling\\_tf\\_common\\_test.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL3Rlc3RzL21vZGVsaW5nX3RmX2NvbW1vbl90ZXN0LnB5) | `79.78% <0%> (-17.03%)` | :arrow_down: |\n| [transformers/modeling\\_roberta.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3JvYmVydGEucHk=) | `58.68% <0%> (-12.58%)` | :arrow_down: |\n| [transformers/modeling\\_xlnet.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX3hsbmV0LnB5) | `70.82% <0%> (-2.47%)` | :arrow_down: |\n| [transformers/modeling\\_ctrl.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX2N0cmwucHk=) | `93.18% <0%> (-2.28%)` | :arrow_down: |\n| [transformers/modeling\\_openai.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL21vZGVsaW5nX29wZW5haS5weQ==) | `80.4% <0%> (-1.36%)` | :arrow_down: |\n| [transformers/file\\_utils.py](https://codecov.io/gh/huggingface/transformers/pull/1657/diff?src=pr&el=tree#diff-dHJhbnNmb3JtZXJzL2ZpbGVfdXRpbHMucHk=) | `76.37% <0%> (+2.19%)` | :arrow_up: |\n\n------\n\n[Continue to review full report at Codecov](https://codecov.io/gh/huggingface/transformers/pull/1657?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/1657?src=pr&el=footer). Last update [079bfb3...cbd0696](https://codecov.io/gh/huggingface/transformers/pull/1657?src=pr&el=lastupdated). Read the [comment docs](https://docs.codecov.io/docs/pull-request-comments).\n", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

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

[]

[ "What is your suggestion, then? Adding a mp_encode function? Perhaps this is something that should stay at the user's side. ", "Hello @jianwolf,\r\nYes indeed, I've never taken the time to do it (mainly because most of the I do pre-processing are one-shot: I launch it before leaving the office 😴).\r\nIf you feel like opening a pull request with your suggestion, I would happy to add it.\r\n\r\n@BramVanroy do you see any drawbacks of having parallelized pre-processing by default?\r\n\r\nI tried to integrate your few lines and had this error:\r\n```\r\n File \"/usr/lib/python3.6/multiprocessing/reduction.py\", line 51, in dumps\r\n cls(buf, protocol).dump(obj)\r\nAttributeError: Can't pickle local object 'main.<locals>.process_data'\r\n```\r\n\r\nIt seems like `process_data` should be outside of the `main`, that shouldn't be too complicated.\r\n\r\n(Also, how many parallel processes/cpus do you have on your server for this order of magnitude in reduction?)\r\n\r\nVictor\r\n", "@VictorSanh At first reading I thought the suggestion was to implement a default multiprocessing encoding for tokenizers. That would seem like a large change that needs a lot of testing across multiple platforms (note the different between fork and spawn) as well as a possible reproducibility issue when retrieving results from different threads, and thus different batch orders. Of course these problems could be mitigated but it seemed like a lot of work to suddenly overhaul all tokenizers in this way.\r\n\r\nNow that it's clear that it's only for the distillation script, I'm sure there's no big issue here even though I would like to see this implemented in a deterministic way, i.e. order of return values should always be identical. ", "Hi! Yeah I will create a pull request for this code! On my machine there are 80 CPU threads available!", "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" ]

[]

[]

[ "Oh yes indeed. Let me correct it. Thank you for pointing that out @jianwolf!" ]

[]

[]

[ "cc @keskarnitish :)", "Sure, I'll get on this soon. I'll push it to https://github.com/salesforce/ctrl and link here once I'm done. ", "Thanks @keskarnitish! That would be great!", "Added in https://github.com/salesforce/ctrl/commit/a0d0b4d2f38ae55a1396dfad4d6bff7cc9435c2d , see updated `README.md` for usage. ", "That is awesome @keskarnitish! We'll add it to our doc here as well." ]

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

[]

[ "To answer question 2:\r\nYou can see the assumed indices for special characters by loading a tokenizer and inspecting the `vocab` attribute.\r\n\r\nFor example:\r\n```\r\nfrom transformers import BertTokenizer\r\ntokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\r\nprint(tokenizer.vocab)\r\n```\r\nThis shows that for the `bert-base-uncased` model it is assumed that:\r\n\r\n| Special Token | Index |\r\n| --- | --- |\r\n| [PAD] | 0 |\r\n| [UNK] | 100 |\r\n| [CLS] | 101 |\r\n| [SEP] | 102 |\r\n| [MASK] | 103 |", "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 } ]

[]

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

[]

[]

[ "The easiest way is to use the torch launch script. It will automatically set the local rank correctly. It would look something like this (can't test, am on phone) :\r\n\r\n```bash\r\npython -m torch.distributed.launch --nproc_per_node 8 run_squad.py <your arguments>\r\n```", "Hi,\r\n\r\nThanks for the fast answer!\r\n\r\nYes I saw this solution in the examples, but I am interested in the case when I am using PyTorch container and I have to set up an entry point for the training (= run_squad.py) and its parameters . And so in that case how should I set it? Or just let it to be -1? \r\n\r\n(Or you recommend in that case to create a bash file as entry where I start this torch lunch.)\r\n\r\nThanks again! ", "If you want to run it manually, you'll have to run the script once for each GPU, and set the local rank to the GPU ID for each process. It might help to look at the contents of the launch script that I mentioned before. It shows you how to set the local rank automatically for multiple processes, which I think is what you want. ", "Ok, thanks for the response! I will try that!", "If your problem is fixed, please do close this issue. ", "@tothniki Did you have to modify the script very much to run with SM? Attempting to do so now, as well. ", "@petulla No, at the end i didn't modify anything regarding to the multiple GPU problem. ( of course I had to modify the read-in and the save to a S3 Bucket).I tried with SageMaker as it was, and it seemed to me that the distribution between GPUs worked.", "> The easiest way is to use the torch launch script. It will automatically set the local rank correctly. It would look something like this (can't test, am on phone) :\r\n> \r\n> ```shell\r\n> python -m torch.distributed.launch --nproc_per_node 8 run_squad.py <your arguments>\r\n> ```\r\n\r\nHi @ugent\r\n\r\nwhat about ( run_language_modeling.py ) ? \r\nDoes passing local_rank = 0 to it means it will automatically do the task on 4 GPUs (for ex.) which we have available ? and our speed will be 4 times faster ? (by distributed training)\r\n\r\nor we have to run script by ( python -m torch.distributed.launch .....)\r\n \r\n\r\n", "@mahdirezaey\r\n\r\nPlease use the correct tag when tagging...\r\n\r\nNo, it will not do this automatically, you have to use the launch utility." ]

[ { "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" ]

[]

[]

[ "Merging with #1370 " ]

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

[]

[ "This is quite a general question. Perhaps it's more useful to put this on Stack Overflow. ", "We plan to have a forum associated to the repo to discuss these types of general questions.\r\nIn the meantime, we are still happy to welcome them in the PR but the visibility is limited indeed.", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "Yes, we should push a new pip release this coming week. In the meantime please use master." ]

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

[]

[ "Yes thanks would be happy to welcome a PR.\r\nThanks for ca(t😂)ching that", "This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.\n" ]

[]

[]

[ "You have opened an issue for the transformers repository but executed code from fairseq. Don't you think you should create an issue there [1]?\r\n\r\n[1] https://github.com/pytorch/fairseq/blob/master/examples/roberta/README.md", "Right. Sorry! I made a mistake.", "I have error when i run this code : please how to fix it?????\r\n\r\n\r\n# Load the model in fairseq\r\nfrom fairseq.models.roberta import RobertaModel\r\nroberta = RobertaModel.from_pretrained('/path/to/roberta.large', checkpoint_file='model.pt')\r\nroberta.eval() # disable dropout (or leave in train mode to finetune)\r\n\r\n---------------------------------------------------------------------------\r\nTypeError Traceback (most recent call last)\r\n[<ipython-input-23-cd858fcec71b>](https://localhost:8080/#) in <module>\r\n 1 # Load the model in fairseq\r\n 2 from fairseq.models.roberta import RobertaModel\r\n----> 3 roberta = RobertaModel.from_pretrained('/path/to/roberta.large', checkpoint_file='model.pt')\r\n 4 roberta.eval() # disable dropout (or leave in train mode to finetune)\r\n\r\n2 frames\r\n[/usr/lib/python3.8/posixpath.py](https://localhost:8080/#) in join(a, *p)\r\n 74 will be discarded. An empty last part will result in a path that\r\n 75 ends with a separator.\"\"\"\r\n---> 76 a = os.fspath(a)\r\n 77 sep = _get_sep(a)\r\n 78 path = a\r\n\r\nTypeError: expected str, bytes or os.PathLike object, not NoneType", "Same comment as above, please open your issue in the correct repository.", "i did not understand this.. can you write the code for this, please?\r\nnote i run the previous code as following:\r\n\r\n\r\n\r\n", "can you write the correct code please for that?", "I fixed it ... thanks \r\n" ]