SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

doc and conversion

Browse Source
This commit is contained in:
thomwolf
2019-10-09 17:06:30 +02:00
parent c19b8e4ae0
commit 48b438ff2a
3 changed files with 18 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
transformers/__init__.py
View File

@@ -155,6 +155,11 @@ if is_tf_available():
load_distilbert_pt_weights_in_tf2,
TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP)
from .modeling_tf_ctrl import (TFCTRLPreTrainedModel, TFCTRLModel,
TFCTRLLMHeadModel,
load_ctrl_pt_weights_in_tf2,
TF_CTRL_PRETRAINED_MODEL_ARCHIVE_MAP)
# TF 2.0 <=> PyTorch conversion utilities
if is_tf_available() and is_torch_available():
from .modeling_tf_pytorch_utils import (convert_tf_weight_name_to_pt_weight_name,

14
transformers/convert_pytorch_checkpoint_to_tf2.py
View File

@@ -31,7 +31,8 @@ from transformers import (BertConfig, TFBertForPreTraining, TFBertForQuestionAns
TransfoXLConfig, TFTransfoXLLMHeadModel, load_transfo_xl_pt_weights_in_tf2, TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP,
OpenAIGPTConfig, TFOpenAIGPTLMHeadModel, load_openai_gpt_pt_weights_in_tf2, OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP,
RobertaConfig, TFRobertaForMaskedLM, TFRobertaForSequenceClassification, load_roberta_pt_weights_in_tf2, ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
DistilBertConfig, TFDistilBertForMaskedLM, TFDistilBertForQuestionAnswering, load_distilbert_pt_weights_in_tf2, DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP)
DistilBertConfig, TFDistilBertForMaskedLM, TFDistilBertForQuestionAnswering, load_distilbert_pt_weights_in_tf2, DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
CTRLConfig, TFCTRLLMHeadModel, load_ctrl_pt_weights_in_tf2, CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP)
if is_torch_available():
import torch
@@ -43,7 +44,8 @@ if is_torch_available():
TransfoXLLMHeadModel, TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_MAP,
OpenAIGPTLMHeadModel, OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP,
RobertaForMaskedLM, RobertaForSequenceClassification, ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
DistilBertForMaskedLM, DistilBertForQuestionAnswering, DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP)
DistilBertForMaskedLM, DistilBertForQuestionAnswering, DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
CTRLLMHeadModel, CTRL_PRETRAINED_MODEL_ARCHIVE_MAP)
else:
(BertForPreTraining, BertForQuestionAnswering, BertForSequenceClassification, BERT_PRETRAINED_MODEL_ARCHIVE_MAP,
GPT2LMHeadModel, GPT2_PRETRAINED_MODEL_ARCHIVE_MAP,
@@ -52,7 +54,8 @@ else:
TransfoXLLMHeadModel, TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_MAP,
OpenAIGPTLMHeadModel, OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP,
RobertaForMaskedLM, RobertaForSequenceClassification, ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
DistilBertForMaskedLM, DistilBertForQuestionAnswering, DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP,) = (
DistilBertForMaskedLM, DistilBertForQuestionAnswering, DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
CTRLLMHeadModel, CTRL_PRETRAINED_MODEL_ARCHIVE_MAP) = (
None, None, None, None,
None, None,
None, None,
@@ -60,7 +63,8 @@ else:
None, None,
None, None,
None, None, None,
None, None, None,)
None, None, None,
None, None)
import logging
@@ -80,6 +84,7 @@ MODEL_CLASSES = {
'roberta-large-mnli': (RobertaConfig, TFRobertaForSequenceClassification, load_roberta_pt_weights_in_tf2, RobertaForSequenceClassification, ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP, ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP),
'distilbert': (DistilBertConfig, TFDistilBertForMaskedLM, load_distilbert_pt_weights_in_tf2, DistilBertForMaskedLM, DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP, DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP),
'distilbert-base-uncased-distilled-squad': (DistilBertConfig, TFDistilBertForQuestionAnswering, load_distilbert_pt_weights_in_tf2, DistilBertForQuestionAnswering, DISTILBERT_PRETRAINED_MODEL_ARCHIVE_MAP, DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP),
'ctrl': (CTRLConfig, TFCTRLLMHeadModel, load_ctrl_pt_weights_in_tf2, CTRLLMHeadModel, CTRL_PRETRAINED_MODEL_ARCHIVE_MAP, CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP)
}
def convert_pt_checkpoint_to_tf(model_type, pytorch_checkpoint_path, config_file, tf_dump_path, compare_with_pt_model=False, use_cached_models=True):
@@ -228,6 +233,7 @@ if __name__ == "__main__":
convert_all_pt_checkpoints_to_tf(args.model_type.lower() if args.model_type is not None else None,
args.tf_dump_path,
model_shortcut_names_or_path=[args.pytorch_checkpoint_path] if args.pytorch_checkpoint_path is not None else None,
config_shortcut_names_or_path=[args.config_file] if args.config_file is not None else None,
compare_with_pt_model=args.compare_with_pt_model,
use_cached_models=args.use_cached_models,
only_convert_finetuned_models=args.only_convert_finetuned_models)