Add Weekly job for model analysis

.github/workflows/model-analysis-weekly.yml

@@ -0,0 +1,59 @@
+name: Model Analysis Weekly
+
+on:
+  workflow_dispatch:
+  # schedule:
+  #   - cron: '0 23 * * 5' # 11:00 PM UTC Friday (12:00 AM Saturday Serbia)
+  push:
+    branches: ["pchandrasekaran/model_analysis_weekly_job"]
+
+jobs:
+  build-and-push-test:
+    uses: ./.github/workflows/build-and-test.yml
+    secrets: inherit
+    with:
+      test_mark: 'push'
+
+  model-analysis:
+    needs: build-and-push-test
+    runs-on: in-service
+    steps:
+      - name: Run Model Analysis Script
+        shell: bash
+        run: |
+          source env/activate
+          python scripts/model_analysis.py \
+            --test_directory_or_file_path forge/test/models/pytorch \
+            --dump_failure_logs \
+            --markdown_directory_path ./model_analysis_docs \
+            --unique_ops_output_directory_path ./models_unique_ops_output \
+            2>&1 | tee model_analysis.log
+
+      - name: Upload Model Analysis Script Logs
+        uses: actions/upload-artifact@v4
+        if: success() || failure()
+        with:
+          name: model-analysis-outputs
+          path: model_analysis.log
+
+      - name: Upload Models Unique Ops test Failure Logs
+        uses: actions/upload-artifact@v4
+        if: success() || failure()
+        with:
+          name: unique-ops-logs
+          path: ./models_unique_ops_output
+
+      - name: Create Pull Request
+        uses: peter-evans/create-pull-request@v7
+        with:
+          branch: model_analysis
+          committer: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
+          author: ${{ github.actor }} <${{ github.actor }}@users.noreply.github.com>
+          base: main
+          commit-message: "Update model analysis docs"
+          title: "Update model analysis docs"
+          body: "This PR will update model analysis docs"
+          labels: model_analysis
+          delete-branch: true
+          files: ./model_analysis_docs
+          token: ${{ secrets.GH_TOKEN }}

forge/test/conftest.py

@@ -448,3 +448,24 @@ def pytest_runtest_logreport(report):
         for key, default_value in environ_before_test.items():
             if os.environ.get(key, "") != default_value:
                 os.environ[key] = default_value
+
+
+def pytest_collection_modifyitems(config, items):
+
+    marker = config.getoption("-m")  # Get the marker from the -m option
+
+    if marker and marker == "automatic_model_analysis":  # If a marker is specified
+        filtered_items = [item for item in items if marker in item.keywords]
+        print("Automatic Model Analysis Collected tests: ")
+        test_count = 0
+        for item in items:
+            if marker in item.keywords:
+                test_file_path = item.location[0]
+                test_name = item.location[2]
+                print(f"{test_file_path}::{test_name}")
+                test_count += 1
+        print(f"Automatic Model Analysis Collected test count: {test_count}")
+        if not filtered_items:  # Warn if no tests match the marker
+            print(f"Warning: No tests found with marker '{marker}'.")
+    else:
+        print(items)

forge/test/models/pytorch/text/bart/test_bart.py

@@ -23,6 +23,7 @@ def forward(self, input_ids, attention_mask, decoder_input_ids):
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 def test_pt_bart_classifier(test_device):
     compiler_cfg = _get_global_compiler_config()
     compiler_cfg.compile_depth = CompileDepth.SPLIT_GRAPH

forge/test/models/pytorch/text/distilbert/test_distilbert.py

@@ -16,6 +16,7 @@
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 @pytest.mark.parametrize("variant", variants, ids=variants)
 def test_distilbert_masked_lm_pytorch(variant, test_device):
     # Load DistilBert tokenizer and model from HuggingFace
@@ -46,6 +47,7 @@ def test_distilbert_masked_lm_pytorch(variant, test_device):
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 def test_distilbert_question_answering_pytorch(test_device):
     # Load Bert tokenizer and model from HuggingFace
     model_ckpt = "distilbert-base-cased-distilled-squad"
@@ -82,6 +84,7 @@ def test_distilbert_question_answering_pytorch(test_device):
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 def test_distilbert_sequence_classification_pytorch(test_device):
 
     # Load DistilBert tokenizer and model from HuggingFace
@@ -109,6 +112,7 @@ def test_distilbert_sequence_classification_pytorch(test_device):
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 def test_distilbert_token_classification_pytorch(test_device):
     # Load DistilBERT tokenizer and model from HuggingFace
     model_ckpt = "Davlan/distilbert-base-multilingual-cased-ner-hrl"

forge/test/models/pytorch/vision/autoencoder/test_autoencoder.py

@@ -13,6 +13,7 @@
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 def test_conv_ae_pytorch(test_device):
     # Set Forge configuration parameters
     compiler_cfg = forge.config._get_global_compiler_config()
@@ -40,6 +41,7 @@ def test_conv_ae_pytorch(test_device):
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 def test_linear_ae_pytorch(test_device):
     # Set Forge configuration parameters
     compiler_cfg = forge.config._get_global_compiler_config()

forge/test/models/pytorch/vision/fpn/test_fpn.py

@@ -8,6 +8,7 @@
 
 
 @pytest.mark.nightly
+@pytest.mark.automatic_model_analysis
 def test_fpn_pytorch(test_device):
     compiler_cfg = forge.config._get_global_compiler_config()
     compiler_cfg.compile_depth = forge.CompileDepth.SPLIT_GRAPH

scripts/model_analysis.py

@@ -542,53 +542,56 @@ def dump_logs(log_file_dir_path: str, log_file_name: str, content: str):
         logger.info(f"Dumped test logs in {log_file}")
 
 
-def collect_all_pytests(root_dir_path):
+def collect_all_model_analysis_test(directory_or_file_path, output_directory_path):
 
-    assert check_path(root_dir_path), f"The directory path for collecting pytest {root_dir_path} doesn't exists"
+    assert check_path(
+        directory_or_file_path
+    ), f"The directory path for collecting test {directory_or_file_path} doesn't exists"
 
-    logger.info(f"Collecting all pytests in {root_dir_path}")
+    logger.info(f"Collecting all test that has automatic_model_analysis marker in {directory_or_file_path}")
 
+    collected_test_outputs = ""
     try:
-        res = subprocess.check_output(["pytest", root_dir_path, "--setup-plan"], stderr=subprocess.STDOUT).decode(
-            "utf-8"
+        result = subprocess.run(
+            ["pytest", directory_or_file_path, "-m", "automatic_model_analysis", "--collect-only"],
+            capture_output=True,
+            text=True,
+            check=True,
         )
-    except subprocess.CalledProcessError as e:
-        output = e.output.decode("utf-8")
-        logger.error(f"[Error!] output = {output}")
-        return []
 
-    test_list = []
-    lines = res.split("\n")
-    for line in lines:
-        if "warnings summary" in line or "slowest durations" in line:
-            break
+        collected_test_outputs += "STDOUT:\n"
+        collected_test_outputs += result.stdout
+        collected_test_outputs += "STDERR:\n"
+        collected_test_outputs += result.stderr
 
-        if line and line.startswith("        " + root_dir_path) and "::" in line and "training" not in line:
-            line = line.strip()
-            line = line.split(" (fixtures used:")[0] if " (fixtures used:" in line else line
-            if "Grayskull" not in line and "Wormhole_B0" not in line:
-                test_list.append(line)
+    except subprocess.CalledProcessError as e:
+        collected_test_outputs += e.output
 
-    return test_list
+    dump_logs(output_directory_path, "collected_tests.txt", collected_test_outputs)
 
+    test_list = []
+    with open(os.path.join(output_directory_path, "collected_tests.txt"), "r") as collected_test_file:
+        lines = collected_test_file.readlines()
+        test_lines = False
+        for line in lines:
+            if "Automatic Model Analysis Collected tests:" in line:
+                test_lines = True
+            elif "Automatic Model Analysis Collected test count:" in line:
+                test_lines = False
+                break
+            elif test_lines:
+                test_list.append(str(line).replace("\n", ""))
 
-def generate_and_export_unique_ops_tests(pytest_directory_path, model_file_path, unique_ops_output_directory_path):
+    return test_list
 
-    # If model_file_path is specified, collect all the tests in the model_file_path parent directory path
-    # and in the test_list will only include the tests matching with model_file_path,
-    # otherwise collect all the tests in the pytest_directory_path specified by the user
-    if model_file_path:
-        model_file_path_list = model_file_path.split("/")[:-1]
-        tests_directory_path = "/".join(model_file_path_list)
-    else:
-        tests_directory_path = pytest_directory_path
 
-    test_list = collect_all_pytests(tests_directory_path)
+def generate_and_export_unique_ops_tests(test_directory_or_file_path, unique_ops_output_directory_path):
 
-    if model_file_path:
-        test_list = [test for test in test_list if test.split("::")[0] == model_file_path]
+    test_list = collect_all_model_analysis_test(test_directory_or_file_path, unique_ops_output_directory_path)
 
-    assert test_list != [], f"No tests found in the {tests_directory_path} path"
+    assert (
+        test_list != []
+    ), f"No tests found in the {test_directory_or_file_path} path with automatic_model_analysis pytest marker"
 
     # Create a dictonary contains model_name as key and model tests(i.e include variant, task) as values
     model_name_to_tests = {}
@@ -599,6 +602,11 @@ def generate_and_export_unique_ops_tests(pytest_directory_path, model_file_path,
         else:
             model_name_to_tests[model_name].append(test)
 
+    for model_name, tests in model_name_to_tests.items():
+        print(f"{model_name}:")
+        for test in tests:
+            print(f"\t\t\t{test}")
+
     # Generate unique op test for the all collected test and save the models unique ops test information in the unique_ops_output_directory_path
     model_output_dir_paths = []
     for model_name, tests in model_name_to_tests.items():
@@ -941,21 +949,16 @@ def run_model_unique_op_tests_and_generate_markdowns(
 
 def main():
     parser = argparse.ArgumentParser(
-        description="""Generate unique ops test for the models present in the  pytest_directory_path or model_file_path
+        description="""Generate unique ops test for the models present in the test_directory_or_file_path
         specified by the user and run the unique ops test and generate markdown files, the root markdown file contains model name,
         variant name, framework and compiler components supported rate and sub-markdown file contains  model variant unique op tests info"""
     )
 
-    group = parser.add_mutually_exclusive_group(required=True)
-    group.add_argument(
-        "--pytest_directory_path",
-        type=str,
-        help="Specify the directory path containing models to test.",
-    )
-    group.add_argument(
-        "--model_file_path",
+    parser.add_argument(
+        "--test_directory_or_file_path",
         type=str,
-        help="Specify the model file path to generate unique op tests and markdown file.",
+        default=os.path.join(os.getcwd(), "forge/test"),
+        help="Specify the directory or file path containing models test with automatic_model_analysis pytest marker",
     )
     parser.add_argument(
         "--dump_failure_logs",
@@ -964,29 +967,21 @@ def main():
     )
     parser.add_argument(
         "--markdown_directory_path",
-        default=os.path.join(os.getcwd(), "new_models_analysis_docs"),
+        default=os.path.join(os.getcwd(), "models_analysis_docs"),
         required=False,
         help="Specify the directory path for saving models information as markdowns file",
     )
     parser.add_argument(
         "--unique_ops_output_directory_path",
-        default=os.path.join(os.getcwd(), "new_unique_ops"),
+        default=os.path.join(os.getcwd(), "unique_ops"),
         required=False,
         help="Specify the output directory path for saving models unique op tests outputs(i.e failure logs, xlsx file)",
     )
 
     args = parser.parse_args()
 
-    if args.pytest_directory_path is not None:
-        assert check_path(
-            args.pytest_directory_path
-        ), f"Specified pytest directory path {args.pytest_directory_path} doesn't exists"
-    else:
-        assert check_path(args.model_file_path), f"Specified model file path {args.model_file_path} doesn't exists"
-
     model_output_dir_paths = generate_and_export_unique_ops_tests(
-        pytest_directory_path=args.pytest_directory_path,
-        model_file_path=args.model_file_path,
+        test_directory_or_file_path=args.test_directory_or_file_path,
         unique_ops_output_directory_path=args.unique_ops_output_directory_path,
     )