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

Support segformer fx (#19924)

Browse Source 
* Support segformer fx

* Add fx_compatible attribute to test_modeling_segformer.py

* Update glpn model (fx support)

glpn model was copied from segformer.

* Update utils/fx.py | add semantic-segmentation

for SegformerForSemanticSegmentation model

* Fix minor import order(isort)

* Add random input generation for segformer fx

Co-authored-by: noelbird <lduldu00228@gmail.com>
This commit is contained in:
donguk.lim
2022-10-28 21:44:38 +09:00
committed by GitHub
parent dcca71be61
commit 347ba38cb4
4 changed files with 9 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/transformers/models/glpn/modeling_glpn.py
View File

@@ -149,7 +149,7 @@ class GLPNEfficientSelfAttention(nn.Module):
def transpose_for_scores(self, hidden_states): def transpose_for_scores(self, hidden_states):
new_shape = hidden_states.size()[:-1] + (self.num_attention_heads, self.attention_head_size) new_shape = hidden_states.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
hidden_states = hidden_states.view(*new_shape) hidden_states = hidden_states.view(new_shape)
return hidden_states.permute(0, 2, 1, 3) return hidden_states.permute(0, 2, 1, 3)
def forward( def forward(
@@ -190,7 +190,7 @@ class GLPNEfficientSelfAttention(nn.Module):
context_layer = context_layer.permute(0, 2, 1, 3).contiguous() context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,) new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
context_layer = context_layer.view(*new_context_layer_shape) context_layer = context_layer.view(new_context_layer_shape)
outputs = (context_layer, attention_probs) if output_attentions else (context_layer,) outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)

4
src/transformers/models/segformer/modeling_segformer.py
View File

@@ -179,7 +179,7 @@ class SegformerEfficientSelfAttention(nn.Module):
def transpose_for_scores(self, hidden_states): def transpose_for_scores(self, hidden_states):
new_shape = hidden_states.size()[:-1] + (self.num_attention_heads, self.attention_head_size) new_shape = hidden_states.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
hidden_states = hidden_states.view(*new_shape) hidden_states = hidden_states.view(new_shape)
return hidden_states.permute(0, 2, 1, 3) return hidden_states.permute(0, 2, 1, 3)
def forward( def forward(
@@ -220,7 +220,7 @@ class SegformerEfficientSelfAttention(nn.Module):
context_layer = context_layer.permute(0, 2, 1, 3).contiguous() context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,) new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
context_layer = context_layer.view(*new_context_layer_shape) context_layer = context_layer.view(new_context_layer_shape)
outputs = (context_layer, attention_probs) if output_attentions else (context_layer,) outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)

4
src/transformers/utils/fx.py
View File

@@ -44,6 +44,7 @@ from ..models.auto.modeling_auto import (
MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING_NAMES, MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING_NAMES,
MODEL_FOR_PRETRAINING_MAPPING_NAMES, MODEL_FOR_PRETRAINING_MAPPING_NAMES,
MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES, MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES,
MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING_NAMES,
MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES, MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES,
MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES, MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES,
MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES, MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES,
@@ -80,6 +81,7 @@ def _generate_supported_model_class_names(
"image-classification": MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES, "image-classification": MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES,
"ctc": MODEL_FOR_CTC_MAPPING_NAMES, "ctc": MODEL_FOR_CTC_MAPPING_NAMES,
"audio-classification": MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMES, "audio-classification": MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMES,
"semantic-segmentation": MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING_NAMES,
} }
if supported_tasks is None: if supported_tasks is None:
@@ -128,6 +130,7 @@ _REGULAR_SUPPORTED_MODEL_NAMES_AND_TASKS = [
"plbart", "plbart",
"resnet", "resnet",
"roberta", "roberta",
"segformer",
"speech_to_text", "speech_to_text",
"speech_to_text_2", "speech_to_text_2",
"swin", "swin",
@@ -730,6 +733,7 @@ class HFTracer(Tracer):
*get_values(MODEL_FOR_CAUSAL_LM_MAPPING_NAMES), *get_values(MODEL_FOR_CAUSAL_LM_MAPPING_NAMES),
*get_values(MODEL_FOR_MASKED_LM_MAPPING_NAMES), *get_values(MODEL_FOR_MASKED_LM_MAPPING_NAMES),
*get_values(MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES), *get_values(MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES),
*get_values(MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING_NAMES),
"GPT2DoubleHeadsModel", "GPT2DoubleHeadsModel",
]: ]:
inputs_dict["labels"] = torch.zeros(shape, dtype=torch.long, device=device) inputs_dict["labels"] = torch.zeros(shape, dtype=torch.long, device=device)

1
tests/models/segformer/test_modeling_segformer.py
View File

@@ -161,6 +161,7 @@ class SegformerModelTest(ModelTesterMixin, unittest.TestCase):
else () else ()
) )
fx_compatible = True
test_head_masking = False test_head_masking = False
test_pruning = False test_pruning = False
test_resize_embeddings = False test_resize_embeddings = False