mindstudio-probe 1.0.3__py3-none-any.whl → 1.1.0__py3-none-any.whl

msprobe/pytorch/online_dispatch/torch_ops_config.yaml

@@ -1,50 +1,58 @@
-aten_ops_blacklist:
-  - _cudnn_rnn
-  - _local_scalar_dense
-  - _pin_memory
-  - _to_copy
-  - _unsafe_view
-  - clone
-  - contiguous
-  - copy_
-  - cudnn_batch_norm
-  - cudnn_batch_norm_backward
-  - detach
-  - empty
-  - index_put_
-  - lift_fresh
-  - max_pool2d_with_indices_backward # shape unmatch
-  - native_batch_norm_backward
-  - new_empty
-  - new_empty_strided
-  - new_full
-  - new_ones
-  - new_zeros
-  - ones
-  - ones_like
-  - permute
-  - rand
-  - rand_like
-  - randint
-  - randint_like
-  - randn
-  - randn_like
-  - randperm
-  - scalar_tensor
-  - select
-  - to
-  - transpose
-  - unbind
-  - view
-  - zero
-  - zero_
-  - zeros
-  - zeros_like
-
-npu_adjust_autogard:
-  - adaptive_avg_pool2d
-  - batch_norm
-  - log_softmax
-  - nll_loss
-  - to
+aten_ops_blacklist:
+  - max_pool2d_with_indices
+  - detach
+  - allreduce_
+  - max
+  - npu_rotary_mul
+  - split_with_sizes
+  - npu_dtype_cast  
+  - add_
+  - _cudnn_rnn
+  - _local_scalar_dense
+  - _pin_memory
+  - _to_copy
+  - _unsafe_view
+  - clone
+  - contiguous
+  - copy_
+  - cudnn_batch_norm
+  - cudnn_batch_norm_backward
+  - detach
+  - empty
+  - index_put_
+  - lift_fresh
+  - max_pool2d_with_indices_backward # shape unmatch
+  - native_batch_norm_backward
+  - new_empty
+  - new_empty_strided
+  - new_full
+  - new_ones
+  - new_zeros
+  - ones
+  - ones_like
+  - permute
+  - rand
+  - rand_like
+  - randint
+  - randint_like
+  - randn
+  - randn_like
+  - randperm
+  - scalar_tensor
+  - select
+  - to
+  - transpose
+  - unbind
+  - view
+  - zero
+  - zero_
+  - zeros
+  - zeros_like
+
+npu_adjust_autogard:
+  - adaptive_avg_pool2d
+  - batch_norm
+  - log_softmax
+  - nll_loss
+  - to
   

msprobe/pytorch/online_dispatch/utils.py

@@ -1,146 +1,127 @@
-import os
-import inspect
-import psutil
-import torch
-import numpy as np
-
-try:
-    import torch_npu
-except ImportError:
-    pta_cpu_device = None
-else:
-    pta_cpu_device = torch.device("cpu")
-
-from msprobe.core.common.const import CompareConst, FileCheckConst
-from msprobe.core.common.file_check import change_mode
-from msprobe.core.common.log import logger
-
-cpu_device = torch._C.device("cpu")
-COLOR_RED = '\033[31m'
-COLOR_GREEN = '\033[32m'
-COLOR_YELLOW = '\033[33m'
-COLOR_BLUE = '\033[34m'
-COLOR_PURPLE = '\033[35m'
-COLOR_CYAN = '\033[36m'
-COLOR_GRAY = '\033[37m'
-COLOR_RESET = '\033[0m'
-
-COMPARE_LOGO = '''
-            _ _                                                       
-  ___  _ __ | (_)_ __   ___    ___ ___  _ __ ___  _ __   __ _ _ __ ___ 
- / _ \\| '_ \\| | | '_ \\ / _ \\  / __/ _ \\| '_ ` _ \\| '_ \\ / _` | '__/ _ \\ 
-| (_) | | | | | | | | |  __/ | (_| (_) | | | | | | |_) | (_| | | |  __/
- \\___/|_| |_|_|_|_| |_|\\___|  \\___\\___/|_| |_| |_| .__/ \\__,_|_|  \\___|
-                                                 |_|    
-'''
-
-CSV_COLUMN_NAME = [CompareConst.NPU_NAME,
-                   CompareConst.BENCH_NAME,
-                   CompareConst.NPU_DTYPE,
-                   CompareConst.BENCH_DTYPE,
-                   CompareConst.NPU_SHAPE,
-                   CompareConst.BENCH_SHAPE,
-                   CompareConst.NPU_MAX,
-                   CompareConst.NPU_MIN,
-                   CompareConst.NPU_MEAN,
-                   CompareConst.BENCH_MAX,
-                   CompareConst.BENCH_MIN,
-                   CompareConst.BENCH_MEAN,
-                   CompareConst.COSINE,
-                   CompareConst.MAX_ABS_ERR,
-                   CompareConst.MAX_RELATIVE_ERR,
-                   CompareConst.ACCURACY,
-                   CompareConst.STACK,
-                   CompareConst.ERROR_MESSAGE]
-
-FLOAT_TYPE = [np.half, np.single, float, np.double, np.float64, np.longdouble, np.float32, np.float16]
-BOOL_TYPE = [bool, np.uint8]
-INT_TYPE = [np.int32, np.int64]
-
-
-def get_callstack():
-    callstack = []
-    for (_, path, line, func, code, _) in inspect.stack()[2:]:
-        if code:
-            stack_line = [path, str(line), func, code[0].strip() if code else code]
-        else:
-            stack_line = [path, str(line), func, code]
-        callstack.append(stack_line)
-    return callstack
-
-
-def np_save_data(data, file_name, data_path):
-    try:
-        if hasattr(data, "numpy"):
-            data = data.numpy()
-        dump_path = os.path.join(data_path, f'{file_name}.npy')
-        np.save(dump_path, data)
-        change_mode(dump_path, FileCheckConst.DATA_FILE_AUTHORITY)
-    except Exception as e:
-        logger.error("save numpy failed, error: {}".format(e))
-    finally:
-        pass
-
-
-def data_to_cpu(data, deep, data_cpu):
-    global cpu_device
-    list_cpu = []
-    if isinstance(data, torch.Tensor):
-        if data.device == cpu_device or data.device == pta_cpu_device:
-            tensor_copy = data.clone().detach()
-        else:
-            tensor_copy = data.cpu().detach()
-        if tensor_copy.dtype in [torch.float16, torch.half, torch.bfloat16]:
-            tensor_copy = tensor_copy.float()
-
-        if deep == 0:
-            data_cpu.append(tensor_copy)
-        return tensor_copy
-    elif isinstance(data, list):
-        for v in data:
-            list_cpu.append(data_to_cpu(v, deep + 1, data_cpu))
-        if deep == 0:
-            data_cpu.append(list_cpu)
-        return list_cpu
-    elif isinstance(data, tuple):
-        for v in data:
-            list_cpu.append(data_to_cpu(v, deep + 1, data_cpu))
-        tuple_cpu = tuple(list_cpu)
-        if deep == 0:
-            data_cpu.append(tuple_cpu)
-        return tuple_cpu
-    elif isinstance(data, dict):
-        dict_cpu = {}
-        for k, v in data.items():
-            dict_cpu[k] = data_to_cpu(v, deep + 1, data_cpu)
-        if deep == 0:
-            data_cpu.append(dict_cpu)
-        return dict_cpu
-    elif isinstance(data, torch._C.device):
-        return cpu_device
-    else:
-        if deep == 0:
-            data_cpu.append(data)
-        return data
-
-
-def get_sys_info():
-    mem = psutil.virtual_memory()
-    cpu_percent = psutil.cpu_percent(interval=1)
-    sys_info = f'Total: {mem.total / 1024 / 1024:.2f}MB ' \
-               f'Free: {mem.available / 1024 / 1024:.2f} MB ' \
-               f'Used: {mem.used / 1024 / 1024:.2f} MB ' \
-               f'CPU: {cpu_percent}% '
-    return sys_info
-
-
-class DispatchException(Exception):
-    INVALID_PARAMETER = 0
-
-    def __init__(self, err_code, err_msg=""):
-        super(DispatchException, self).__init__()
-        self.err_code = err_code
-        self.err_msg = err_msg
-
-    def __str__(self):
-        return self.err_msg
+import inspect
+import psutil
+import torch
+import numpy as np
+
+try:
+    import torch_npu
+except ImportError:
+    pta_cpu_device = None
+else:
+    pta_cpu_device = torch.device("cpu")
+
+from msprobe.core.common.const import CompareConst
+
+cpu_device = torch._C.device("cpu")
+COLOR_RED = '\033[31m'
+COLOR_GREEN = '\033[32m'
+COLOR_YELLOW = '\033[33m'
+COLOR_BLUE = '\033[34m'
+COLOR_PURPLE = '\033[35m'
+COLOR_CYAN = '\033[36m'
+COLOR_GRAY = '\033[37m'
+COLOR_RESET = '\033[0m'
+
+COMPARE_LOGO = '''
+            _ _                                                       
+  ___  _ __ | (_)_ __   ___    ___ ___  _ __ ___  _ __   __ _ _ __ ___ 
+ / _ \\| '_ \\| | | '_ \\ / _ \\  / __/ _ \\| '_ ` _ \\| '_ \\ / _` | '__/ _ \\ 
+| (_) | | | | | | | | |  __/ | (_| (_) | | | | | | |_) | (_| | | |  __/
+ \\___/|_| |_|_|_|_| |_|\\___|  \\___\\___/|_| |_| |_| .__/ \\__,_|_|  \\___|
+                                                 |_|    
+'''
+
+CSV_COLUMN_NAME = [CompareConst.NPU_NAME,
+                   CompareConst.BENCH_NAME,
+                   CompareConst.NPU_DTYPE,
+                   CompareConst.BENCH_DTYPE,
+                   CompareConst.NPU_SHAPE,
+                   CompareConst.BENCH_SHAPE,
+                   CompareConst.NPU_MAX,
+                   CompareConst.NPU_MIN,
+                   CompareConst.NPU_MEAN,
+                   CompareConst.BENCH_MAX,
+                   CompareConst.BENCH_MIN,
+                   CompareConst.BENCH_MEAN,
+                   CompareConst.COSINE,
+                   CompareConst.MAX_ABS_ERR,
+                   CompareConst.MAX_RELATIVE_ERR,
+                   CompareConst.ACCURACY,
+                   CompareConst.STACK,
+                   CompareConst.ERROR_MESSAGE]
+
+FLOAT_TYPE = [np.half, np.single, float, np.double, np.float64, np.longdouble, np.float32, np.float16]
+BOOL_TYPE = [bool, np.uint8]
+INT_TYPE = [np.int32, np.int64]
+
+
+def get_callstack():
+    callstack = []
+    for (_, path, line, func, code, _) in inspect.stack()[2:]:
+        stack_line = [path, str(line), func, code[0].strip() if code else code]
+        callstack.append(stack_line)
+    return callstack
+
+
+def data_to_cpu(data, deep, data_cpu):
+    global cpu_device
+    list_cpu = []
+    if isinstance(data, torch.Tensor):
+        if data.device == cpu_device or data.device == pta_cpu_device:
+            tensor_copy = data.clone().detach()
+        else:
+            tensor_copy = data.cpu().detach()
+        if tensor_copy.dtype in [torch.float16, torch.half, torch.bfloat16]:
+            tensor_copy = tensor_copy.float()
+
+        if deep == 0:
+            data_cpu.append(tensor_copy)
+        return tensor_copy
+    elif isinstance(data, list):
+        for v in data:
+            list_cpu.append(data_to_cpu(v, deep + 1, data_cpu))
+        if deep == 0:
+            data_cpu.append(list_cpu)
+        return list_cpu
+    elif isinstance(data, tuple):
+        for v in data:
+            list_cpu.append(data_to_cpu(v, deep + 1, data_cpu))
+        tuple_cpu = tuple(list_cpu)
+        if deep == 0:
+            data_cpu.append(tuple_cpu)
+        return tuple_cpu
+    elif isinstance(data, dict):
+        dict_cpu = {}
+        for k, v in data.items():
+            dict_cpu[k] = data_to_cpu(v, deep + 1, data_cpu)
+        if deep == 0:
+            data_cpu.append(dict_cpu)
+        return dict_cpu
+    elif isinstance(data, torch._C.device):
+        return cpu_device
+    else:
+        if deep == 0:
+            data_cpu.append(data)
+        return data
+
+
+def get_sys_info():
+    mem = psutil.virtual_memory()
+    cpu_percent = psutil.cpu_percent(interval=1)
+    sys_info = f'Total: {mem.total / 1024 / 1024:.2f}MB ' \
+               f'Free: {mem.available / 1024 / 1024:.2f} MB ' \
+               f'Used: {mem.used / 1024 / 1024:.2f} MB ' \
+               f'CPU: {cpu_percent}% '
+    return sys_info
+
+
+class DispatchException(Exception):
+    INVALID_PARAMETER = 0
+
+    def __init__(self, err_code, err_msg=""):
+        super(DispatchException, self).__init__()
+        self.err_code = err_code
+        self.err_msg = err_msg
+
+    def __str__(self):
+        return self.err_msg

msprobe/pytorch/parse.py

@@ -1,4 +1,19 @@
-from msprobe.pytorch.parse_tool import cli
-
-if __name__ == '__main__':
-    cli.parse()
+# Copyright (c) 2024-2024, Huawei Technologies Co., Ltd.
+# 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.
+
+from msprobe.pytorch.parse_tool import cli
+
+if __name__ == '__main__':
+    cli.parse()

msprobe/pytorch/parse_tool/cli.py

@@ -1,32 +1,31 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-"""
-# Copyright (C) 2022-2024. Huawei Technologies Co., Ltd. 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.
-"""
-from msprobe.pytorch.parse_tool.lib.interactive_cli import InteractiveCli
-from msprobe.pytorch.common.log import logger
-
-
-def _run_interactive_cli(cli=None):
-    logger.info("Interactive command mode")
-    if not cli:
-        cli = InteractiveCli()
-    try:
-        cli.cmdloop(intro="Start Parsing........")
-    except KeyboardInterrupt:
-        logger.info("Exit parsing.......")
-
-
-def parse():
-    _run_interactive_cli()
+# Copyright (c) 2024-2024, Huawei Technologies Co., Ltd.
+# 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.
+
+from msprobe.pytorch.parse_tool.lib.interactive_cli import InteractiveCli
+from msprobe.pytorch.common.log import logger
+
+
+def _run_interactive_cli(cli=None):
+    logger.info("Interactive command mode")
+    if not cli:
+        cli = InteractiveCli()
+    try:
+        cli.cmdloop(intro="Start Parsing........")
+    except KeyboardInterrupt:
+        logger.info("Exit parsing.......")
+
+
+def parse():
+    _run_interactive_cli()