[WIP] Add Tapas (bis)#8988

Closed

NielsRogge wants to merge 68 commits intohuggingface:masterfrom

NielsRogge:tapas_v4

Commits on Dec 11, 2020

First commit: adding all files from tapas_v3
Rogge Niels
committed
Fix multiple bugs including soft dependency and new structure of the library
Rogge Niels
committed
Improve testing by adding torch_device to inputs and adding dependency on scatter
Rogge Niels
committed
Use Python 3 inheritance rather than Python 2
Rogge Niels
committed
First draft model cards of base sized models
Rogge Niels
committed
Remove model cards as they are already on the hub
Rogge Niels
committed
Fix multiple bugs with integration tests
Rogge Niels
committed
All model integration tests pass
Rogge Niels
committed
Remove print statement
Rogge Niels
committed
Add test for convert_logits_to_predictions method of TapasTokenizer
Rogge Niels
committed
Incorporate suggestions by Google authors
Rogge Niels
committed
Fix remaining tests

LysandreJik
authored and
Rogge Niels
committed
Change position embeddings sizes to 512 instead of 1024
Rogge Niels
committed
Comment out positional embedding sizes
Rogge Niels
committed
Update PRETRAINED_VOCAB_FILES_MAP and PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
Rogge Niels
committed
Added more model names
Rogge Niels
committed
Fix truncation when no max length is specified

LysandreJik
authored and
Rogge Niels
committed
Disable torchscript test

LysandreJik
authored and
Rogge Niels
committed
Make style & make quality
Rogge Niels
committed
Quality

LysandreJik
authored and
Rogge Niels
committed
Address CI needs

LysandreJik
authored and
Rogge Niels
committed
Test the Masked LM model

LysandreJik
authored and
Rogge Niels
committed
Fix the masked LM model

LysandreJik
authored and
Rogge Niels
committed
Truncate when overflowing

LysandreJik
authored and
Rogge Niels
committed
More much needed docs improvements
Rogge Niels
committed
Fix some URLs
Rogge Niels
committed
Some more docs improvements
Rogge Niels
committed
Test PyTorch scatter
LysandreJik
committed
Set to slow + minify
LysandreJik
committed
Merge pull request #8 from huggingface/tapas_v4_final
NielsRogge
authored
Calm flake8 down
LysandreJik
committed
Merge pull request #9 from huggingface/tapas_v4_final
NielsRogge
authored
Fix min_null_pred in the run_qa script (#9067)
sgugger
authored
[model_cards] Migrate cards from this repo to model repos on huggingface.co (#9013)

authored

Commits on Dec 14, 2020

Fix embeddings resizing in TF models (#8657)
jplu
authored
Patch *ForCausalLM model (#9092)
LysandreJik
authored
First commit: adding all files from tapas_v3
Rogge Niels
committed
Fix multiple bugs including soft dependency and new structure of the library
Rogge Niels
committed
Improve testing by adding torch_device to inputs and adding dependency on scatter
Rogge Niels
committed
Use Python 3 inheritance rather than Python 2
Rogge Niels
committed
First draft model cards of base sized models
Rogge Niels
committed
Remove model cards as they are already on the hub
Rogge Niels
committed
Fix multiple bugs with integration tests
Rogge Niels
committed
All model integration tests pass
Rogge Niels
committed
Remove print statement
Rogge Niels
committed
Add test for convert_logits_to_predictions method of TapasTokenizer
Rogge Niels
committed
Incorporate suggestions by Google authors
Rogge Niels
committed
Fix remaining tests

LysandreJik
authored and
Rogge Niels
committed
Change position embeddings sizes to 512 instead of 1024
Rogge Niels
committed
Comment out positional embedding sizes
Rogge Niels
committed
Update PRETRAINED_VOCAB_FILES_MAP and PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
Rogge Niels
committed
Added more model names
Rogge Niels
committed
Fix truncation when no max length is specified

LysandreJik
authored and
Rogge Niels
committed
Disable torchscript test

LysandreJik
authored and
Rogge Niels
committed
Make style & make quality
Rogge Niels
committed
Quality

LysandreJik
authored and
Rogge Niels
committed
Address CI needs

LysandreJik
authored and
Rogge Niels
committed
Test the Masked LM model

LysandreJik
authored and
Rogge Niels
committed
Fix the masked LM model

LysandreJik
authored and
Rogge Niels
committed
Truncate when overflowing

LysandreJik
authored and
Rogge Niels
committed
More much needed docs improvements
Rogge Niels
committed
Fix some URLs
Rogge Niels
committed
Some more docs improvements
Rogge Niels
committed
Add add_pooling_layer argument to TapasModel
Rogge Niels
committed
Fix merge conflicts
Rogge Niels
committed
Fix issue in docs + fix style and quality
Rogge Niels
committed
Clean up conversion script and add task parameter to TapasConfig
Rogge Niels
committed
Revert the task parameter of TapasConfig
Rogge Niels
committed