Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Complete pytorch transformers interface, deprecate old GPT implement (#…
…881) * Rename namespaces to suppress warnings. * Revert "Rename namespaces to suppress warnings." This reverts commit 0cf7b23. * Initial working-ish attempt. * Intermediate check-in... * More partial progress. * Another pass... * Fix sep/cls handling, cleanup. * Further cleanup. * Keyword name fix. * Another flag fix. * Pull debug print. * Line length cleanup. * WiC fix. * Two task setup bugs. * BoolQ typo * Improved segment handling. * Delete unused is_pair_task, other cleanup/fixes. * Fix deleted path from merge. * Fix cache path. * relocate tasks from seminar * add linguistic phenomena benchmark tasks * Address (spurious?) tokenization warning. * Select pool_type automatically to match model. h/t Haokun Liu * Config updates. * Path fix * add two prefix method and simple LM * Fix XLNet UNK handling. * Internal temporary MNLI alternate. * Revert "Internal temporary MNLI alternate." This reverts commit 455792a. * refacor tags in data loader * Add helper fn tests * Finish merge * Remove unused argument. * update task init * Possible ReCoRD bug fix * Cleanup * Fix merge issues. * Revert "Remove unused argument." This reverts commit 96a7c37. * Assorted responses to Alex's commenst. * Further ReCoRD fix. * @iftenney's comments. * Fix/simplify segment logic. * @W4ngatang's comments * Cleanup. * add forward functinos * bugfix * merge pytorch transformer * update old process split * add gpt2 * add get_pretrained_lm_head for transformers * update filename * add config * debug * update config * allow evaluate with raw parameter * debug * Cleanup * Fix issues with alternative embeddings_mode settings, max_layer. * More mix cleanup. * Masking fix. * cleanup * simplify get_seg_ids * debug * related adjustments to add pytorch transformers * pytorch transformer refactor * formatting * formatting * debug * TransformerXL fix * update test script * formatting again * add note to transfo-xl * debug * update test script * update test script * tokenized_name change * cleanup * pool type fix * config update * Update defaults.conf * rename use_pytorch_transformer * cleanup * Update test_preprocess.py * Update test_checkpointing.py * Update test_write_preds.py * clean up * debug * name changes * name changes * update message * name changes * tokenizer name fix * docstring changes * name changes * restore asserts * add pair embedding for pytorch_transformers * add max position embedding assert * deal with gpt-like boundary fn * roberta tokenizer support * roberta model support * roberta embedder * fix roberta seg_id * change unused_task_name message * more test cases for pytorch_tranformers_interface * gpt-style mirrored pair forward func for similarity tasks * Update environment.yml * adjust import location * black * move import location * update test script * add comments to test script * update test script * pool type fix * tokenizer fix * debug * special tokens fix * roberta vocab fix * roberta tokenizer fix * clean up * Update test_pytorch_transformers_interface.py * add_special_token fix * black * fix roberta message logic * fix embedding extend bug * black * clean up * simplify add_special_token fix * add assert for lm task & pytorch_transformers * black * relocate task_modulator initialization * minor changes * rename task_modulator -> model_preprocessing_interface * change lm_parsing process_split docstring * black * add gpt2-large * update dependency * update dependency for real * clean up * add a forgotten similarity task for gpt * update setup * update setup
- Loading branch information
1 parent
815beea
commit 6921e4d
Showing
25 changed files
with
1,117 additions
and
700 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Empty file.
Submodule pytorch_huggingface
deleted from
bfd8e0
Oops, something went wrong.