Merge pull request #14 from yuguo-Jack/rocm

merge master

README.md

@@ -9,8 +9,8 @@ OneFlow 是一个深度学习框架，旨在**易用，可扩展且高效**。
 
 ## Latest News
 
-- Version 0.9.0 is out!
-  - [Full changelog](https://github.com/Oneflow-Inc/oneflow/releases/tag/v0.9.0)
+- Version 1.0.0 is out!
+  - [Full changelog](https://github.com/Oneflow-Inc/oneflow/releases/tag/v1.0.0)
 
 ## 安装 OneFlow-DCU
 

python/oneflow/framework/infer_compiler/__init__.py

@@ -14,43 +14,14 @@
 limitations under the License.
 """
 
-import os
-
-import oneflow as flow
-from oneflow.framework.args_tree import ArgsTree
+try:
+    import torch
+except ImportError:
+    print("You should install torch also when use `oneflow.framework.infer_compiler`.")
 
 from .transform.custom_transform import register
 from .utils.patch_for_compiler import *
 from .with_fx_graph import fx_node_tranform
 from .with_fx_interpreter import OneFlowInterpreter
 from .with_oneflow_compile import compile_from_torch
-
-
-def oneflow_backend(gm, example_inputs, *args, **kwargs):
-    with_interp = os.getenv(
-        "ONEDIFF_INFER_COMPILER_USE_INTERPRETER", "False"
-    ).lower() in ("true", "1", "t",)
-    if not with_interp:
-        transformed_fn = fx_node_tranform(gm)
-
-    def wrapped_forward(*args, **kwargs):
-        def input_fn(value):
-            if isinstance(value, torch.Tensor):
-                return flow.utils.tensor.from_torch(value.contiguous())
-            else:
-                return value
-
-        args_tree = ArgsTree((args, kwargs), False, tensor_type=torch.Tensor)
-        out = args_tree.map_leaf(input_fn)
-        args = out[0]
-        if with_interp:
-            output = OneFlowInterpreter(gm, garbage_collect_values=False).run(
-                *args, **kwargs
-            )
-        else:
-            output = transformed_fn(*args, **kwargs)
-        if isinstance(output, tuple):
-            return tuple(flow.utils.tensor.to_torch(i) for i in output)
-        return flow.utils.tensor.to_torch(output)
-
-    return wrapped_forward
+from .with_oneflow_backend import oneflow_backend

python/oneflow/framework/infer_compiler/with_fx_graph.py

@@ -46,7 +46,7 @@ def fx_node_tranform(gm):
         os.environ.setdefault("ONEFLOW_MLIR_FUSE_FORWARD_OPS", "1")
         os.environ.setdefault("ONEFLOW_MLIR_FUSE_OPS_WITH_BACKWARD_IMPL", "1")
         os.environ.setdefault("ONEFLOW_MLIR_GROUP_MATMUL", "1")
-        os.environ.setdefault("ONEFLOW_MLIR_PREFER_NHWC", "1")
+        os.environ.setdefault("ONEFLOW_MLIR_PREFER_NHWC", "0")
         os.environ.setdefault("ONEFLOW_KERNEL_ENABLE_FUSED_CONV_BIAS", "1")
         os.environ.setdefault("ONEFLOW_KERNEL_ENABLE_FUSED_LINEAR", "1")
         os.environ.setdefault(

python/oneflow/framework/infer_compiler/with_oneflow_backend.py

@@ -0,0 +1,53 @@
+"""
+Copyright 2020 The OneFlow Authors. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+"""
+
+import os
+import torch
+
+import oneflow as flow
+from oneflow.framework.args_tree import ArgsTree
+from .with_fx_graph import fx_node_tranform
+from .with_fx_interpreter import OneFlowInterpreter
+
+
+def oneflow_backend(gm, example_inputs, *args, **kwargs):
+    with_interp = os.getenv(
+        "ONEDIFF_INFER_COMPILER_USE_INTERPRETER", "False"
+    ).lower() in ("true", "1", "t",)
+    if not with_interp:
+        transformed_fn = fx_node_tranform(gm)
+
+    def wrapped_forward(*args, **kwargs):
+        def input_fn(value):
+            if isinstance(value, torch.Tensor):
+                return flow.utils.tensor.from_torch(value.contiguous())
+            else:
+                return value
+
+        args_tree = ArgsTree((args, kwargs), False, tensor_type=torch.Tensor)
+        out = args_tree.map_leaf(input_fn)
+        args = out[0]
+        if with_interp:
+            output = OneFlowInterpreter(gm, garbage_collect_values=False).run(
+                *args, **kwargs
+            )
+        else:
+            output = transformed_fn(*args, **kwargs)
+        if isinstance(output, tuple):
+            return tuple(flow.utils.tensor.to_torch(i) for i in output)
+        return flow.utils.tensor.to_torch(output)
+
+    return wrapped_forward

python/oneflow/test/misc/test_autograd_functional.py

@@ -13,6 +13,7 @@
 See the License for the specific language governing permissions and
 limitations under the License.
 """
+from packaging import version
 import unittest
 import oneflow as flow
 import oneflow.unittest
@@ -110,44 +111,62 @@ def test_hvp(test_case):
         )
         result_tensors = torch.autograd.functional.hvp(_func_multi_scalar, inputs, v)
 
-    # TODO: The local test of test_jacobian and test_hessian passed, but the ci test failed
-    """
+    # TODO: "'jacobian' and 'hessian' has no strategy parameter in PyTorch before '1.11.0'"
     @autotest(n=1, check_graph=False)
     def test_jacobian(test_case):
         inputs = random_tensor(ndim=2, dim0=5, dim1=5)
-        result_tensor = torch.autograd.functional.jacobian(
-            _func_tensor, inputs, vectorize=False, strategy="reverse-mode"
-        )
+        if version.parse(torch.pytorch.__version__) < version.parse("1.11.0"):
+            result_tensor = torch.autograd.functional.jacobian(
+                _func_tensor, inputs, vectorize=False
+            )
+        else:
+            result_tensor = torch.autograd.functional.jacobian(
+                _func_tensor, inputs, vectorize=False, strategy="reverse-mode"
+            )
 
         inputs = (
             random_tensor(ndim=2, dim0=5, dim1=5),
             random_tensor(ndim=2, dim0=5, dim1=5),
         )
-        result_tensors = torch.autograd.functional.jacobian(
-            _func_multi_scalar, inputs, vectorize=False, strategy="reverse-mode"
-        )
+        if version.parse(torch.pytorch.__version__) < version.parse("1.11.0"):
+            result_tensors = torch.autograd.functional.jacobian(
+                _func_multi_scalar, inputs, vectorize=False
+            )
+        else:
+            result_tensors = torch.autograd.functional.jacobian(
+                _func_multi_scalar, inputs, vectorize=False, strategy="reverse-mode"
+            )
 
     @autotest(n=1, check_graph=False)
     def test_hessian(test_case):
         inputs = random_tensor(ndim=2, dim0=5, dim1=5)
-        result_tensor = torch.autograd.functional.hessian(
-            _func_scalar,
-            inputs,
-            vectorize=False,
-            outer_jacobian_strategy="reverse-mode",
-        )
+        if version.parse(torch.pytorch.__version__) < version.parse("1.11.0"):
+            result_tensor = torch.autograd.functional.hessian(
+                _func_scalar, inputs, vectorize=False,
+            )
+        else:
+            result_tensor = torch.autograd.functional.hessian(
+                _func_scalar,
+                inputs,
+                vectorize=False,
+                outer_jacobian_strategy="reverse-mode",
+            )
 
         inputs = (
             random_tensor(ndim=2, dim0=5, dim1=5),
             random_tensor(ndim=2, dim0=5, dim1=5),
         )
-        result_tensors = torch.autograd.functional.hessian(
-            _func_multi_scalar,
-            inputs,
-            vectorize=False,
-            outer_jacobian_strategy="reverse-mode",
-        )
-    """
+        if version.parse(torch.pytorch.__version__) < version.parse("1.11.0"):
+            result_tensors = torch.autograd.functional.hessian(
+                _func_multi_scalar, inputs, vectorize=False,
+            )
+        else:
+            result_tensors = torch.autograd.functional.hessian(
+                _func_multi_scalar,
+                inputs,
+                vectorize=False,
+                outer_jacobian_strategy="reverse-mode",
+            )
 
 
 if __name__ == "__main__":