[dicp] add dicp ci (#648)

* add dicp ci

* fix cache dir not exist error

.github/workflows/main.yml

@@ -441,6 +441,17 @@ jobs:
           cd ${DEEPLINK_PATH}/${GITHUB_RUN_NUMBER}/Build-Ascend-910b/dipu
           source scripts/ci/ascend/ci_ascend_env.sh
           bash tests/run_ascend_tests.sh
+      
+      - name: Run dicp op test
+        run: |
+          set -ex
+          cd ${DEEPLINK_PATH}/${GITHUB_RUN_NUMBER}/Build-Ascend-910b/dipu
+          source scripts/ci/ascend/ci_ascend_env.sh
+          cd ${DEEPLINK_PATH}/${GITHUB_RUN_NUMBER}/Build-Ascend-910b/dicp
+          pip uninstall dicp -y
+          python setup.py clean && python setup.py install --user
+          export TEST_DIR=$(pwd)/test
+          bash ${TEST_DIR}/ascend_scripts/ops/run_test_ops.sh false
 
   Test-One-Iter-Ascend-910b:
     name: Test-one-iter-ascend-910b

dicp/dicp/dynamo_bridge/utils.py

@@ -1,4 +1,5 @@
 import copy
+from pathlib import Path
 from typing import Any, Dict, Tuple
 
 import torch.fx
@@ -7,6 +8,7 @@
 
 
 def save_cpu_gm(gm: torch.fx.GraphModule, folder: str):
+    Path(folder).mkdir(exist_ok=True)
     cpu_gm = copy_gm_to_cpu(gm)
     grap_code = cpu_gm.code
     graph_key = code_hash(grap_code)

dicp/dicp/vendor/AscendGraph/compile_job.py

@@ -3,7 +3,7 @@
 import time
 
 from dicp.dynamo_bridge.compile import DeviceCompileJob
-from torch._inductor.codecache import pick_vec_isa, cpp_compile_command, write, get_hash
+from torch._inductor.codecache import pick_vec_isa, cpp_compile_command, write, code_hash
 from torch._inductor import exc
 
 
@@ -24,7 +24,7 @@ def __init__(self, source_code) -> None:
             source_code.strip(),
             "json",
             extra=cpp_compile_command("i", "o", vec_isa=picked_vec_isa) +
-                  'local_rank' + str(self._local_rank) + get_hash(compile_file_code, 'code')
+                  'local_rank' + str(self._local_rank) + code_hash(compile_file_code)
         )
         self._output_graph_path = self._input_path[:-5] + '/graph'
         print('output_path: ', self._output_graph_path)

dicp/dicp/vendor/AscendGraph/opset_convert.py

@@ -1,12 +1,15 @@
 import torch
-from dicp.dynamo_bridge.op_transformer import BackendPatternMatcherTransformer
+from dicp.dynamo_bridge.compile_fx import is_torch_210
 from dicp.vendor.AscendGraph.ascend_op import MatMul, CastToCpu, IdentityInp
 from dicp.vendor.AscendGraph.conversion import AtenToAscendTransformer
-from dicp.vendor.AscendGraph.pattern_replacement import (
-    ascend_pattern_matcher,
-    aten_patterns_cls_list,
-    ascend_patterns_cls_list
-)
+
+if is_torch_210:
+    from dicp.dynamo_bridge.op_transformer import BackendPatternMatcherTransformer
+    from dicp.vendor.AscendGraph.pattern_replacement import (
+        ascend_pattern_matcher,
+        aten_patterns_cls_list,
+        ascend_patterns_cls_list
+    )
 
 
 # 该pass需要在FuseTransposeMatmul之后
@@ -74,13 +77,14 @@ def symint_in_inputs(nodes):
 def ascendgraph_opset_convert(
     gm: torch.fx.GraphModule,
 ):
-    gm = BackendPatternMatcherTransformer(
-        ascend_pattern_matcher, aten_patterns_cls_list).transform(gm)
+    if is_torch_210:
+        gm = BackendPatternMatcherTransformer(
+            ascend_pattern_matcher, aten_patterns_cls_list).transform(gm)
     gm = AtenToAscendTransformer(gm).transform()
 
     # For bug in pytorch
     # Avoid for dynamic shape
-    if not symint_in_inputs(list(gm.graph.nodes)):
+    if is_torch_210 and not symint_in_inputs(list(gm.graph.nodes)):
         gm = BackendPatternMatcherTransformer(
             ascend_pattern_matcher, ascend_patterns_cls_list).transform(gm)
     gm = OutputMarkPass().transform(gm)

dicp/test/common/utils.py

@@ -5,6 +5,7 @@
 import torch._dynamo as dynamo
 import torch_dipu
 from dicp.dynamo_bridge import pt_patch # noqa F401
+from dicp.dynamo_bridge.compile_fx import is_torch_210
 torch.manual_seed(1)
 random.seed(1)
 
@@ -22,12 +23,9 @@ def __init__(self, static_size, dynamic_size):
 
 
 def update_dynamo_config(dynamic=False):
-    if dynamic:
-        dynamo.config.dynamic_shapes = True
-        dynamo.config.assume_static_by_default = False
-    else:
-        dynamo.config.dynamic_shapes = False
-        dynamo.config.assume_static_by_default = True
+    dynamo.config.dynamic_shapes = dynamic
+    if is_torch_210:
+        dynamo.config.assume_static_by_default = not dynamic
 
 
 def get_device():