TorchAO compile + offloading tests

tests/quantization/bnb/test_4bit.py

@@ -880,5 +880,7 @@ def test_torch_compile(self):
     def test_torch_compile_with_cpu_offload(self):
         super()._test_torch_compile_with_cpu_offload(quantization_config=self.quantization_config)
 
-    def test_torch_compile_with_group_offload(self):
-        super()._test_torch_compile_with_group_offload(quantization_config=self.quantization_config)
+    def test_torch_compile_with_group_offload_leaf(self):
+        super()._test_torch_compile_with_group_offload_leaf(
+            quantization_config=self.quantization_config, use_stream=True
+        )

tests/quantization/bnb/test_mixed_int8.py

@@ -844,7 +844,7 @@ def test_torch_compile_with_cpu_offload(self):
         )
 
     @pytest.mark.xfail(reason="Test fails because of an offloading problem from Accelerate with confusion in hooks.")
-    def test_torch_compile_with_group_offload(self):
-        super()._test_torch_compile_with_group_offload(
-            quantization_config=self.quantization_config, torch_dtype=torch.float16
+    def test_torch_compile_with_group_offload_leaf(self):
+        super()._test_torch_compile_with_group_offload_leaf(
+            quantization_config=self.quantization_config, torch_dtype=torch.float16, use_stream=True
         )

tests/quantization/test_torch_compile_utils.py

@@ -64,16 +64,17 @@ def _test_torch_compile_with_cpu_offload(self, quantization_config, torch_dtype=
             # small resolutions to ensure speedy execution.
             pipe("a dog", num_inference_steps=3, max_sequence_length=16, height=256, width=256)
 
-    def _test_torch_compile_with_group_offload(self, quantization_config, torch_dtype=torch.bfloat16):
+    def _test_torch_compile_with_group_offload_leaf(
+        self, quantization_config, torch_dtype=torch.bfloat16, *, use_stream: bool = False
+    ):
         torch._dynamo.config.cache_size_limit = 10000
 
         pipe = self._init_pipeline(quantization_config, torch_dtype)
         group_offload_kwargs = {
             "onload_device": torch.device("cuda"),
             "offload_device": torch.device("cpu"),
             "offload_type": "leaf_level",
-            "use_stream": True,
-            "non_blocking": True,
+            "use_stream": use_stream,
         }
         pipe.transformer.enable_group_offload(**group_offload_kwargs)
         pipe.transformer.compile()

tests/quantization/torchao/test_torchao.py

@@ -19,6 +19,7 @@
 from typing import List
 
 import numpy as np
+from parameterized import parameterized
 from transformers import AutoTokenizer, CLIPTextModel, CLIPTokenizer, T5EncoderModel
 
 from diffusers import (
@@ -29,6 +30,7 @@
     TorchAoConfig,
 )
 from diffusers.models.attention_processor import Attention
+from diffusers.quantizers import PipelineQuantizationConfig
 from diffusers.utils.testing_utils import (
     backend_empty_cache,
     backend_synchronize,
@@ -44,6 +46,8 @@
     torch_device,
 )
 
+from ..test_torch_compile_utils import QuantCompileTests
+
 
 enable_full_determinism()
 
@@ -625,6 +629,51 @@ def test_int_a16w8_cpu(self):
         self._check_serialization_expected_slice(quant_method, quant_method_kwargs, expected_slice, device)
 
 
+@require_torchao_version_greater_or_equal("0.7.0")
+class TorchAoCompileTest(QuantCompileTests):
+    quantization_config = PipelineQuantizationConfig(
+        quant_mapping={
+            "transformer": TorchAoConfig(quant_type="int8_weight_only"),
+        },
+    )
+
+    def test_torch_compile(self):
+        super()._test_torch_compile(quantization_config=self.quantization_config)
+
+    @unittest.skip(
+        "Changing the device of AQT tensor with module._apply (called from doing module.to() in accelerate) does not work "
+        "when compiling."
+    )
+    def test_torch_compile_with_cpu_offload(self):
+        # RuntimeError: _apply(): Couldn't swap Linear.weight
+        super()._test_torch_compile_with_cpu_offload(quantization_config=self.quantization_config)
+
+    @unittest.skip(
+        """
+        For `use_stream=False`:
+            - Changing the device of AQT tensor, with `param.data = param.data.to(device)` as done in group offloading implementation
+            is unsupported in TorchAO. When compiling, FakeTensor device mismatch causes failure.
+        For `use_stream=True`:
+            Using non-default stream requires ability to pin tensors. AQT does not seem to support this yet in TorchAO.
+        """
+    )
+    @parameterized.expand([False, True])
+    def test_torch_compile_with_group_offload_leaf(self):
+        # For use_stream=False:
+        # If we run group offloading without compilation, we will see:
+        #   RuntimeError: Attempted to set the storage of a tensor on device "cpu" to a storage on different device "cuda:0".  This is no longer allowed; the devices must match.
+        # When running with compilation, the error ends up being different:
+        #   Dynamo failed to run FX node with fake tensors: call_function <built-in function linear>(*(FakeTensor(..., device='cuda:0', size=(s0, 256), dtype=torch.bfloat16), AffineQuantizedTensor(tensor_impl=PlainAQTTensorImpl(data=FakeTensor(..., size=(1536, 256), dtype=torch.int8)... , scale=FakeTensor(..., size=(1536,), dtype=torch.bfloat16)... , zero_point=FakeTensor(..., size=(1536,), dtype=torch.int64)... , _layout=PlainLayout()), block_size=(1, 256), shape=torch.Size([1536, 256]), device=cpu, dtype=torch.bfloat16, requires_grad=False), Parameter(FakeTensor(..., device='cuda:0', size=(1536,), dtype=torch.bfloat16,
+        #   requires_grad=True))), **{}): got RuntimeError('Unhandled FakeTensor Device Propagation for aten.mm.default, found two different devices cuda:0, cpu')
+        # Looks like something that will have to be looked into upstream.
+        # for linear layers, weight.tensor_impl shows cuda... but:
+        # weight.tensor_impl.{data,scale,zero_point}.device will be cpu
+
+        # For use_stream=True:
+        # NotImplementedError: AffineQuantizedTensor dispatch: attempting to run unimplemented operator/function: func=<OpOverload(op='aten.is_pinned', overload='default')>, types=(<class 'torchao.dtypes.affine_quantized_tensor.AffineQuantizedTensor'>,), arg_types=(<class 'torchao.dtypes.affine_quantized_tensor.AffineQuantizedTensor'>,), kwarg_types={}
+        super()._test_torch_compile_with_group_offload_leaf(quantization_config=self.quantization_config)
+
+
 # Slices for these tests have been obtained on our aws-g6e-xlarge-plus runners
 @require_torch
 @require_torch_accelerator