[test] added transformers models to test model zoo

FrankLeeeee · FrankLeeeee · commit 298d2e83dbc4 · 2023-03-15T09:55:01.000+08:00
diff --git a/tests/kit/model_zoo/__init__.py b/tests/kit/model_zoo/__init__.py
@@ -1,4 +1,4 @@
-from . import diffusers, timm
+from . import diffusers, timm, transformers
 from .registry import model_zoo
 
 __all__ = ['model_zoo']
diff --git a/tests/kit/model_zoo/transformers/__init__.py b/tests/kit/model_zoo/transformers/__init__.py
@@ -0,0 +1,5 @@
+from .albert import *
+from .bert import *
+from .gpt import *
+from .opt import *
+from .t5 import *
diff --git a/tests/kit/model_zoo/transformers/albert.py b/tests/kit/model_zoo/transformers/albert.py
@@ -0,0 +1,85 @@
+import torch
+import transformers
+
+from ..registry import ModelAttribute, model_zoo
+
+# ===============================
+# Register single-sentence ALBERT
+# ===============================
+BATCH_SIZE = 2
+SEQ_LENGTH = 16
+
+
+def data_gen_fn():
+    input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    token_type_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    return dict(input_ids=input_ids, token_type_ids=token_type_ids, attention_mask=attention_mask)
+
+
+output_transform_fn = lambda x: x
+
+config = transformers.AlbertConfig(embedding_size=128,
+                                   hidden_size=128,
+                                   num_hidden_layers=2,
+                                   num_attention_heads=4,
+                                   intermediate_size=256)
+
+model_zoo.register(name='transformers_albert',
+                   model_fn=lambda: transformers.AlbertModel(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_albert_for_pretraining',
+                   model_fn=lambda: transformers.AlbertForPreTraining(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_albert_for_masked_lm',
+                   model_fn=lambda: transformers.AlbertForMaskedLM(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_albert_for_sequence_classification',
+                   model_fn=lambda: transformers.AlbertForSequenceClassification(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_albert_for_token_classification',
+                   model_fn=lambda: transformers.AlbertForTokenClassification(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+
+# ===============================
+# Register multi-sentence ALBERT
+# ===============================
+
+
+def data_gen_for_qa():
+    question, text = "Who was Jim Henson?", "Jim Henson was a nice puppet"
+    tokenizer = transformers.BertTokenizer.from_pretrained("bert-base-uncased")
+    inputs = tokenizer(question, text, return_tensors="pt")
+    return inputs
+
+
+def data_gen_for_mcq():
+    prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
+    choice0 = "It is eaten with a fork and a knife."
+    choice1 = "It is eaten while held in the hand."
+    tokenizer = transformers.BertTokenizer.from_pretrained("bert-base-uncased")
+    encoding = tokenizer([prompt, prompt], [choice0, choice1], return_tensors="pt", padding=True)
+    encoding = {k: v.unsqueeze(0) for k, v in encoding.items()}
+    return encoding
+
+
+model_zoo.register(name='transformers_albert_for_question_answering',
+                   model_fn=lambda: transformers.AlbertForQuestionAnswering(config),
+                   data_gen_fn=data_gen_for_qa,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_albert_for_multiple_choice',
+                   model_fn=lambda: transformers.AlbertForMultipleChoice(config),
+                   data_gen_fn=data_gen_for_mcq,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
diff --git a/tests/kit/model_zoo/transformers/bert.py b/tests/kit/model_zoo/transformers/bert.py
@@ -0,0 +1,88 @@
+import torch
+import transformers
+
+from ..registry import ModelAttribute, model_zoo
+
+# ===============================
+# Register single-sentence BERT
+# ===============================
+BATCH_SIZE = 2
+SEQ_LENGTH = 16
+
+
+def data_gen_fn():
+    input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    token_type_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    return dict(input_ids=input_ids, token_type_ids=token_type_ids, attention_mask=attention_mask)
+
+
+output_transform_fn = lambda x: x
+
+config = transformers.BertConfig(hidden_size=128, num_hidden_layers=2, num_attention_heads=4, intermediate_size=256)
+
+# register the BERT variants
+model_zoo.register(name='transformers_bert',
+                   model_fn=lambda: transformers.BertModel(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_bert_for_pretraining',
+                   model_fn=lambda: transformers.BertForPreTraining(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_bert_lm_head_model',
+                   model_fn=lambda: transformers.BertLMHeadModel(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_bert_for_masked_lm',
+                   model_fn=lambda: transformers.BertForMaskedLM(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_bert_for_sequence_classification',
+                   model_fn=lambda: transformers.BertForSequenceClassification(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_bert_for_token_classification',
+                   model_fn=lambda: transformers.BertForTokenClassification(config),
+                   data_gen_fn=data_gen_fn,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+
+
+# ===============================
+# Register multi-sentence BERT
+# ===============================
+def data_gen_for_next_sentence():
+    tokenizer = transformers.BertTokenizer.from_pretrained("bert-base-uncased")
+    prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
+    next_sentence = "The sky is blue due to the shorter wavelength of blue light."
+    encoding = tokenizer(prompt, next_sentence, return_tensors="pt")
+    return encoding
+
+
+def data_gen_for_mcq():
+    tokenizer = transformers.BertTokenizer.from_pretrained("bert-base-uncased")
+    prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
+    choice0 = "It is eaten with a fork and a knife."
+    choice1 = "It is eaten while held in the hand."
+    encoding = tokenizer([prompt, prompt], [choice0, choice1], return_tensors="pt", padding=True)
+    encoding = {k: v.unsqueeze(0) for k, v in encoding.items()}
+    return encoding
+
+
+# register the following models
+model_zoo.register(name='transformers_bert_for_next_sentence',
+                   model_fn=lambda: transformers.BertForNextSentencePrediction(config),
+                   data_gen_fn=data_gen_for_next_sentence,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_bert_for_mcq',
+                   model_fn=lambda: transformers.BertForMultipleChoice(config),
+                   data_gen_fn=data_gen_for_mcq,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
diff --git a/tests/kit/model_zoo/transformers/gpt.py b/tests/kit/model_zoo/transformers/gpt.py
@@ -0,0 +1,49 @@
+import torch
+import transformers
+
+from ..registry import ModelAttribute, model_zoo
+
+# ===============================
+# Register single-sentence GPT
+# ===============================
+BATCH_SIZE = 2
+SEQ_LENGTH = 16
+
+
+def data_gen():
+    input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    token_type_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    return dict(input_ids=input_ids, token_type_ids=token_type_ids, attention_mask=attention_mask)
+
+
+output_transform_fn = lambda x: x
+
+config = transformers.GPT2Config(n_position=64, n_layer=2, n_head=4)
+
+# register the following models
+model_zoo.register(name='transformers_gpt',
+                   model_fn=lambda: transformers.GPT2Model(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_gpt_lm',
+                   model_fn=lambda: transformers.GPT2LMHeadModel(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_gpt_double_heads',
+                   model_fn=lambda: transformers.GPT2DoubleHeadsModel(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_gpt_for_token_classification',
+                   model_fn=lambda: transformers.GPT2ForTokenClassification(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_gpt_for_sequence_classification',
+                   model_fn=lambda: transformers.GPT2ForSequenceClassification(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
diff --git a/tests/kit/model_zoo/transformers/opt.py b/tests/kit/model_zoo/transformers/opt.py
@@ -0,0 +1,35 @@
+import torch
+import transformers
+
+from ..registry import ModelAttribute, model_zoo
+
+# ===============================
+# Register single-sentence OPT
+# ===============================
+BATCH_SIZE = 2
+SEQ_LENGTH = 16
+
+
+def data_gen():
+    input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    return dict(input_ids=input_ids, attention_mask=attention_mask)
+
+
+output_transform_fn = lambda x: x
+
+config = transformers.OPTConfig(hidden_size=128, num_hidden_layers=2, num_attention_heads=4)
+
+# register the following models
+# transformers.OPTModel,
+# transformers.OPTForCausalLM,
+model_zoo.register(name='transformers_opt',
+                   model_fn=lambda: transformers.OPTModel(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_opt_for_causal_lm',
+                   model_fn=lambda: transformers.OPTForCausalLM(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
diff --git a/tests/kit/model_zoo/transformers/t5.py b/tests/kit/model_zoo/transformers/t5.py
@@ -0,0 +1,46 @@
+import torch
+import transformers
+
+from ..registry import ModelAttribute, model_zoo
+
+# ===============================
+# Register single-sentence T5
+# ===============================
+BATCH_SIZE = 2
+SEQ_LENGTH = 16
+
+
+def data_gen():
+    input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    decoder_input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    return dict(input_ids=input_ids, decoder_input_ids=decoder_input_ids)
+
+
+def data_gen_for_encoder_only():
+    input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+    return dict(input_ids=input_ids)
+
+
+output_transform_fn = lambda x: x
+
+config = transformers.T5Config(d_model=128, num_layers=2)
+
+# register the following models
+# transformers.T5Model,
+# transformers.T5ForConditionalGeneration,
+# transformers.T5EncoderModel,
+model_zoo.register(name='transformers_t5',
+                   model_fn=lambda: transformers.T5Model(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_t5_for_conditional_generation',
+                   model_fn=lambda: transformers.T5ForConditionalGeneration(config),
+                   data_gen_fn=data_gen,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
+model_zoo.register(name='transformers_t5_encoder_model',
+                   model_fn=lambda: transformers.T5EncoderModel(config),
+                   data_gen_fn=data_gen_for_encoder_only,
+                   output_transform_fn=output_transform_fn,
+                   model_attribute=ModelAttribute(has_control_flow=True))
diff --git a/tests/test_fx/test_tracer/test_hf_model/test_hf_albert.py b/tests/test_fx/test_tracer/test_hf_model/test_hf_albert.py
@@ -1,66 +1,18 @@
-import pytest
-import torch
-import transformers
 from hf_tracer_utils import trace_model_and_compare_output
 
+from tests.kit.model_zoo import model_zoo
+
 BATCH_SIZE = 2
 SEQ_LENGTH = 16
 
 
-def test_single_sentence_albert():
-    MODEL_LIST = [
-        transformers.AlbertModel,
-        transformers.AlbertForPreTraining,
-        transformers.AlbertForMaskedLM,
-        transformers.AlbertForSequenceClassification,
-        transformers.AlbertForTokenClassification,
-    ]
-
-    config = transformers.AlbertConfig(embedding_size=128,
-                                       hidden_size=128,
-                                       num_hidden_layers=2,
-                                       num_attention_heads=4,
-                                       intermediate_size=256)
-
-    def data_gen():
-        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
-        token_type_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
-        attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
-        meta_args = dict(input_ids=input_ids, token_type_ids=token_type_ids, attention_mask=attention_mask)
-        return meta_args
-
-    for model_cls in MODEL_LIST:
-        model = model_cls(config=config)
-        trace_model_and_compare_output(model, data_gen)
-
-
-def test_multi_sentence_albert():
-    config = transformers.AlbertConfig(hidden_size=128,
-                                       num_hidden_layers=2,
-                                       num_attention_heads=4,
-                                       intermediate_size=256)
-    tokenizer = transformers.BertTokenizer.from_pretrained("bert-base-uncased")
-
-    def data_gen_for_qa():
-        question, text = "Who was Jim Henson?", "Jim Henson was a nice puppet"
-        inputs = tokenizer(question, text, return_tensors="pt")
-        return inputs
-
-    model = transformers.AlbertForQuestionAnswering(config)
-    trace_model_and_compare_output(model, data_gen_for_qa)
-
-    def data_gen_for_mcq():
-        prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
-        choice0 = "It is eaten with a fork and a knife."
-        choice1 = "It is eaten while held in the hand."
-        encoding = tokenizer([prompt, prompt], [choice0, choice1], return_tensors="pt", padding=True)
-        encoding = {k: v.unsqueeze(0) for k, v in encoding.items()}
-        return encoding
+def test_albert():
+    sub_registry = model_zoo.get_sub_registry('transformers_albert')
 
-    model = transformers.AlbertForMultipleChoice(config)
-    trace_model_and_compare_output(model, data_gen_for_mcq)
+    for name, (model_fn, data_gen_fn, _, _) in sub_registry.items():
+        model = model_fn()
+        trace_model_and_compare_output(model, data_gen_fn)
 
 
 if __name__ == '__main__':
-    test_single_sentence_albert()
-    test_multi_sentence_albert()
+    test_albert()
diff --git a/tests/test_fx/test_tracer/test_hf_model/test_hf_bert.py b/tests/test_fx/test_tracer/test_hf_model/test_hf_bert.py
diff --git a/tests/test_fx/test_tracer/test_hf_model/test_hf_gpt.py b/tests/test_fx/test_tracer/test_hf_model/test_hf_gpt.py
diff --git a/tests/test_fx/test_tracer/test_hf_model/test_hf_opt.py b/tests/test_fx/test_tracer/test_hf_model/test_hf_opt.py
diff --git a/tests/test_fx/test_tracer/test_hf_model/test_hf_t5.py b/tests/test_fx/test_tracer/test_hf_model/test_hf_t5.py
