[fx] test tracer on diffuser modules. (#1750)

* [fx] test tracer on diffuser modules.

* [fx] shorter seq_len.

* Update requirements-test.txt

requirements/requirements-test.txt

@@ -1,3 +1,4 @@
+diffusers
 pytest
 torchvision
 transformers

tests/test_fx/test_tracer/test_hf_model/test_hf_albert.py

@@ -1,10 +1,10 @@
-import transformers
-import torch
 import pytest
+import torch
+import transformers
 from utils import trace_model_and_compare_output
 
 BATCH_SIZE = 2
-SEQ_LENGHT = 16
+SEQ_LENGTH = 16
 
 
 def test_single_sentence_albert():
@@ -23,9 +23,9 @@ def test_single_sentence_albert():
                                        intermediate_size=256)
 
     def data_gen():
-        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        token_type_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
+        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
 

tests/test_fx/test_tracer/test_hf_model/test_hf_bert.py

@@ -1,10 +1,10 @@
-import transformers
-import torch
 import pytest
+import torch
+import transformers
 from utils import trace_model_and_compare_output
 
 BATCH_SIZE = 2
-SEQ_LENGHT = 16
+SEQ_LENGTH = 16
 
 
 def test_single_sentence_bert():
@@ -20,9 +20,9 @@ def test_single_sentence_bert():
     config = transformers.BertConfig(hidden_size=128, num_hidden_layers=2, num_attention_heads=4, intermediate_size=256)
 
     def data_gen():
-        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        token_type_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
+        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
 

tests/test_fx/test_tracer/test_hf_model/test_hf_diffuser.py

@@ -0,0 +1,116 @@
+import diffusers
+import pytest
+import torch
+import transformers
+from torch.fx import GraphModule
+from utils import trace_model_and_compare_output
+
+from colossalai.fx import ColoTracer
+
+BATCH_SIZE = 2
+SEQ_LENGTH = 5
+HEIGHT = 224
+WIDTH = 224
+IN_CHANNELS = 3
+LATENTS_SHAPE = (BATCH_SIZE, IN_CHANNELS, HEIGHT // 8, WIDTH // 8)
+TIME_STEP = 2
+
+
+def test_vae():
+    MODEL_LIST = [
+        diffusers.AutoencoderKL,
+        diffusers.VQModel,
+    ]
+
+    for model_cls in MODEL_LIST:
+        model = model_cls()
+        sample = torch.zeros(LATENTS_SHAPE)
+
+        tracer = ColoTracer()
+        graph = tracer.trace(root=model)
+
+        gm = GraphModule(model, graph, model.__class__.__name__)
+        gm.recompile()
+
+        model.eval()
+        gm.eval()
+
+        with torch.no_grad():
+            fx_out = gm(sample)
+            non_fx_out = model(sample)
+        assert torch.allclose(
+            fx_out['sample'],
+            non_fx_out['sample']), f'{model.__class__.__name__} has inconsistent outputs, {fx_out} vs {non_fx_out}'
+
+
+def test_clip():
+    MODEL_LIST = [
+        transformers.CLIPModel,
+        transformers.CLIPTextModel,
+        transformers.CLIPVisionModel,
+    ]
+
+    CONFIG_LIST = [
+        transformers.CLIPConfig,
+        transformers.CLIPTextConfig,
+        transformers.CLIPVisionConfig,
+    ]
+
+    def data_gen():
+        if isinstance(model, transformers.CLIPModel):
+            input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+            attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+            position_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+            pixel_values = torch.zeros((BATCH_SIZE, IN_CHANNELS, HEIGHT, WIDTH), dtype=torch.float32)
+            kwargs = dict(input_ids=input_ids,
+                          attention_mask=attention_mask,
+                          position_ids=position_ids,
+                          pixel_values=pixel_values)
+        elif isinstance(model, transformers.CLIPTextModel):
+            input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+            attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+            kwargs = dict(input_ids=input_ids, attention_mask=attention_mask)
+        elif isinstance(model, transformers.CLIPVisionModel):
+            pixel_values = torch.zeros((BATCH_SIZE, IN_CHANNELS, HEIGHT, WIDTH), dtype=torch.float32)
+            kwargs = dict(pixel_values=pixel_values)
+        return kwargs
+
+    for model_cls, config in zip(MODEL_LIST, CONFIG_LIST):
+        model = model_cls(config=config())
+        trace_model_and_compare_output(model, data_gen)
+
+
+@pytest.mark.skip(reason='cannot pass the test yet')
+def test_unet():
+    MODEL_LIST = [
+        diffusers.UNet2DModel,
+        diffusers.UNet2DConditionModel,
+    ]
+
+    for model_cls in MODEL_LIST:
+        model = model_cls()
+        sample = torch.zeros(LATENTS_SHAPE)
+
+        tracer = ColoTracer()
+        graph = tracer.trace(root=model)
+
+        gm = GraphModule(model, graph, model.__class__.__name__)
+        gm.recompile()
+
+        model.eval()
+        gm.eval()
+
+        with torch.no_grad():
+            fx_out = gm(sample, TIME_STEP)
+            non_fx_out = model(sample, TIME_STEP)
+        assert torch.allclose(
+            fx_out['sample'],
+            non_fx_out['sample']), f'{model.__class__.__name__} has inconsistent outputs, {fx_out} vs {non_fx_out}'
+
+
+if __name__ == "__main__":
+    test_vae()
+    test_clip()
+
+    # skip because of failure
+    # test_unet()

tests/test_fx/test_tracer/test_hf_model/test_hf_gpt.py

@@ -1,10 +1,10 @@
-import transformers
-import torch
 import pytest
+import torch
+import transformers
 from utils import trace_model_and_compare_output
 
 BATCH_SIZE = 1
-SEQ_LENGHT = 16
+SEQ_LENGTH = 16
 
 
 def test_gpt():
@@ -19,9 +19,9 @@ def test_gpt():
     config = transformers.GPT2Config(n_position=64, n_layer=2, n_head=4)
 
     def data_gen():
-        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        token_type_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
+        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)
         kwargs = dict(input_ids=input_ids, token_type_ids=token_type_ids, attention_mask=attention_mask)
         return kwargs
 

tests/test_fx/test_tracer/test_hf_model/test_hf_opt.py

@@ -1,10 +1,10 @@
 import pytest
-import transformers
 import torch
+import transformers
 from utils import trace_model_and_compare_output
 
 BATCH_SIZE = 1
-SEQ_LENGHT = 16
+SEQ_LENGTH = 16
 
 
 def test_opt():
@@ -16,8 +16,8 @@ def test_opt():
     config = transformers.OPTConfig(hidden_size=128, num_hidden_layers=2, num_attention_heads=4)
 
     def data_gen():
-        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
+        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+        attention_mask = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
         kwargs = dict(input_ids=input_ids, attention_mask=attention_mask)
         return kwargs
 

tests/test_fx/test_tracer/test_hf_model/test_hf_t5.py

@@ -1,10 +1,10 @@
 import pytest
-import transformers
 import torch
+import transformers
 from utils import trace_model_and_compare_output
 
 BATCH_SIZE = 1
-SEQ_LENGHT = 16
+SEQ_LENGTH = 16
 
 
 def test_t5():
@@ -17,13 +17,13 @@ def test_t5():
     config = transformers.T5Config(d_model=128, num_layers=2)
 
     def data_gen():
-        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
-        decoder_input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
+        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
+        decoder_input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
         kwargs = dict(input_ids=input_ids, decoder_input_ids=decoder_input_ids)
         return kwargs
 
     def data_gen_for_encoder_only():
-        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGHT), dtype=torch.int64)
+        input_ids = torch.zeros((BATCH_SIZE, SEQ_LENGTH), dtype=torch.int64)
         kwargs = dict(input_ids=input_ids)
         return kwargs
 

tests/test_fx/test_tracer/test_hf_model/utils.py

@@ -1,9 +1,10 @@
-from numpy import isin
 import torch
-from colossalai.fx import ColoTracer
+from numpy import isin
 from torch.fx import GraphModule
 from torch.utils._pytree import tree_flatten
 
+from colossalai.fx import ColoTracer
+
 
 def trace_model_and_compare_output(model, data_gen):
     # must turn on eval mode to ensure the output is consistent