update

.github/workflows/pr_tests_gpu.yml

@@ -11,8 +11,6 @@ on:
       - "src/diffusers/loaders/lora_base.py"
       - "src/diffusers/loaders/lora_pipeline.py"
       - "src/diffusers/loaders/peft.py"
-      - "tests/pipelines/test_pipelines_common.py"
-      - "tests/models/test_modeling_common.py"
   workflow_dispatch:
 
 concurrency:
@@ -106,11 +104,18 @@ jobs:
           # https://pytorch.org/docs/stable/notes/randomness.html#avoiding-nondeterministic-algorithms
           CUBLAS_WORKSPACE_CONFIG: :16:8
         run: |
-          pattern=$(cat ${{ steps.extract_tests.outputs.pattern_file }})
-          python -m pytest -n 1 --max-worker-restart=0 --dist=loadfile \
-            -s -v -k "not Flax and not Onnx and $pattern" \
-            --make-reports=tests_pipeline_${{ matrix.module }}_cuda \
-            tests/pipelines/${{ matrix.module }}
+          if [ "${{ matrix.module }}" = "ip_adapters" ]; then 
+              python -m pytest -n 1 --max-worker-restart=0 --dist=loadfile \
+              -s -v -k "not Flax and not Onnx" \
+              --make-reports=tests_pipeline_${{ matrix.module }}_cuda \
+              tests/pipelines/${{ matrix.module }}
+          else 
+              pattern=$(cat ${{ steps.extract_tests.outputs.pattern_file }})
+              python -m pytest -n 1 --max-worker-restart=0 --dist=loadfile \
+              -s -v -k "not Flax and not Onnx and $pattern" \
+              --make-reports=tests_pipeline_${{ matrix.module }}_cuda \
+              tests/pipelines/${{ matrix.module }}
+          fi 
 
       - name: Failure short reports
         if: ${{ failure() }}

tests/models/test_modeling_common.py

@@ -1169,16 +1169,17 @@ class ModelTesterMixin:
         base_output = model(**inputs_dict)
 
         model_size = compute_module_sizes(model)[""]
-        max_size = int(self.model_split_percents[0] * model_size)
-        # Force disk offload by setting very small CPU memory
-        max_memory = {0: max_size, "cpu": int(0.1 * max_size)}
-
         with tempfile.TemporaryDirectory() as tmp_dir:
             model.cpu().save_pretrained(tmp_dir, safe_serialization=False)
+
             with self.assertRaises(ValueError):
+                max_size = int(self.model_split_percents[0] * model_size)
+                max_memory = {0: max_size, "cpu": max_size}
                 # This errors out because it's missing an offload folder
                 new_model = self.model_class.from_pretrained(tmp_dir, device_map="auto", max_memory=max_memory)
 
+            max_size = int(self.model_split_percents[0] * model_size)
+            max_memory = {0: max_size, "cpu": max_size}
             new_model = self.model_class.from_pretrained(
                 tmp_dir, device_map="auto", max_memory=max_memory, offload_folder=tmp_dir
             )

tests/models/transformers/test_models_transformer_omnigen.py

@@ -30,7 +30,6 @@ class OmniGenTransformerTests(ModelTesterMixin, unittest.TestCase):
     model_class = OmniGenTransformer2DModel
     main_input_name = "hidden_states"
     uses_custom_attn_processor = True
-    model_split_percents = [0.1, 0.1, 0.1]
 
     @property
     def dummy_input(self):
@@ -74,9 +73,9 @@ class OmniGenTransformerTests(ModelTesterMixin, unittest.TestCase):
             "num_attention_heads": 4,
             "num_key_value_heads": 4,
             "intermediate_size": 32,
-            "num_layers": 20,
+            "num_layers": 1,
             "pad_token_id": 0,
-            "vocab_size": 1000,
+            "vocab_size": 100,
             "in_channels": 4,
             "time_step_dim": 4,
             "rope_scaling": {"long_factor": list(range(1, 3)), "short_factor": list(range(1, 3))},

tests/models/transformers/test_models_transformer_sd3.py

@@ -33,7 +33,6 @@ enable_full_determinism()
 class SD3TransformerTests(ModelTesterMixin, unittest.TestCase):
     model_class = SD3Transformer2DModel
     main_input_name = "hidden_states"
-    model_split_percents = [0.8, 0.8, 0.9]
 
     @property
     def dummy_input(self):
@@ -68,7 +67,7 @@ class SD3TransformerTests(ModelTesterMixin, unittest.TestCase):
             "sample_size": 32,
             "patch_size": 1,
             "in_channels": 4,
-            "num_layers": 4,
+            "num_layers": 1,
             "attention_head_dim": 8,
             "num_attention_heads": 4,
             "caption_projection_dim": 32,
@@ -108,7 +107,6 @@ class SD3TransformerTests(ModelTesterMixin, unittest.TestCase):
 class SD35TransformerTests(ModelTesterMixin, unittest.TestCase):
     model_class = SD3Transformer2DModel
     main_input_name = "hidden_states"
-    model_split_percents = [0.8, 0.8, 0.9]
 
     @property
     def dummy_input(self):
@@ -143,7 +141,7 @@ class SD35TransformerTests(ModelTesterMixin, unittest.TestCase):
             "sample_size": 32,
             "patch_size": 1,
             "in_channels": 4,
-            "num_layers": 4,
+            "num_layers": 2,
             "attention_head_dim": 8,
             "num_attention_heads": 4,
             "caption_projection_dim": 32,