Debug converter python code with debugpy (#7283)

Add support for debugging tfjs-converter Bazel tests using debugpy in vscode. Support is not perfectly ergonomic yet, but it works and is much better than nothing.

Co-authored-by: Linchenn <40653845+Linchenn@users.noreply.github.com>

.bazelrc

@@ -69,15 +69,21 @@ build --action_env=BROWSERSTACK_USERNAME --action_env=BROWSERSTACK_KEY
 run --action_env=BROWSERSTACK_USERNAME --action_env=BROWSERSTACK_KEY
 test --action_env=BROWSERSTACK_USERNAME --action_env=BROWSERSTACK_KEY
 
+# Make python debugging refer to the real files instead of symlinks
+run --action_env=PYDEVD_RESOLVE_SYMLINKS=true
+
 # Platform specific DISPLAY environment variable for webgl and headless setting
 # for browser tests.
 test:linux --test_env=DISPLAY
 test:macos --define DISPLAY=true
 test:windows --define DISPLAY=true --//:headless=false
 
 # Enable debugging tests with --config=debug
-run:debug --test_arg=--node_options=--inspect-brk --test_output=streamed --test_strategy=exclusive --test_timeout=9999 --nocache_test_results
-test:debug --test_arg=--node_options=--inspect-brk --test_output=streamed --test_strategy=exclusive --test_timeout=9999 --nocache_test_results
+run:debug --test_arg=--node_options=--inspect-brk --test_output=streamed --test_strategy=exclusive --test_timeout=999999 --nocache_test_results
+test:debug --test_arg=--node_options=--inspect-brk --test_output=streamed --test_strategy=exclusive --test_timeout=999999 --nocache_test_results
+
+test:debugpy --test_output=streamed --test_strategy=exclusive --test_timeout=999999 --nocache_test_results
+run:debugpy --test_output=streamed --test_strategy=exclusive --test_timeout=999999 --nocache_test_results
 
 # Load any settings specific to the current user.
 # .bazelrc.user should appear in .gitignore so that settings are not shared with

tfjs-converter/.vscode/launch.json

@@ -0,0 +1,13 @@
+{
+  "version": "0.2.0",
+  "configurations": [
+    {
+      "name": "Python: Attach",
+      "type": "python",
+      "request": "attach",
+      "port": 5724,
+      "host": "localhost"
+    }
+  ],
+  "compounds": []
+}

tfjs-converter/python/README.md

@@ -7,3 +7,57 @@ Use following command to install the library with support of interactive CLI:
 ```bash
 pip install tensorflowjs[wizard]
 ```
+
+Then, run the following to see a list of CLI options
+
+```bash
+tensorflowjs_converter --help
+```
+
+or, use the wizard
+
+```bash
+tensorflowjs_wizard
+```
+
+Alternatively, run the converter via its Bazel target. This must be run from withing the tfjs repo:
+
+```bash
+yarn bazel run //tfjs-converter/python/tensorflowjs/converters:converter -- --help
+```
+
+## Development
+
+The python tests are run with Bazel.
+
+```bash
+yarn bazel test //tfjs-converter/python/...
+```
+
+Alternatively, run `yarn run-python-tests` to run the above command.
+
+To debug a specific test case, use the `--test_filter` option. For example,
+
+```bash
+yarn bazel test //tfjs-converter/python/tensorflowjs/converters:tf_saved_model_conversion_v2_test --test_filter=ConvertTest.test_convert_saved_model_v1
+```
+
+Interactive debugging with breakpoints is supported by `debugpy` in VSCode.
+To enable debugging, put this code at the top of the test file you want to
+debug.
+
+```python
+import debugpy
+debugpy.listen(('localhost', 5724))
+print("Waiting for debugger to connect. See tfjs-converter python README")
+debugpy.wait_for_client()
+```
+
+You may also need to add the following dependency to the test target in the
+Bazel `BUILD` file if it's not already present.
+```starlark
+"//tfjs-converter/python/tensorflowjs:expect_debugpy_installed"
+```
+
+Then, run the test with `bazel run --config=debugpy` and connect
+the VSCode debugger by selecting the `Python: Attach (Converter)` option.

tfjs-converter/python/requirements-dev.txt

@@ -2,3 +2,4 @@
 PyInquirer==1.0.3
 pylint==2.5.0; python_version > '3.0'
 setuptools==65.6.3
+debugpy==1.6.5

tfjs-converter/python/requirements-dev_lock.txt

@@ -116,6 +116,26 @@ cycler==0.11.0 \
     --hash=sha256:3a27e95f763a428a739d2add979fa7494c912a32c17c4c38c4d5f082cad165a3 \
     --hash=sha256:9c87405839a19696e837b3b818fed3f5f69f16f1eec1a1ad77e043dcea9c772f
     # via matplotlib
+debugpy==1.6.5 \
+    --hash=sha256:048368f121c08b00bbded161e8583817af5055982d2722450a69efe2051621c2 \
+    --hash=sha256:0f9afcc8cad6424695f3356dc9a7406d5b18e37ee2e73f34792881a44b02cc50 \
+    --hash=sha256:15bc5febe0edc79726517b1f8d57d7ac7c784567b5ba804aab8b1c9d07a57018 \
+    --hash=sha256:17039e392d6f38388a68bd02c5f823b32a92142a851e96ba3ec52aeb1ce9d900 \
+    --hash=sha256:286ae0c2def18ee0dc8a61fa76d51039ca8c11485b6ed3ef83e3efe8a23926ae \
+    --hash=sha256:377391341c4b86f403d93e467da8e2d05c22b683f08f9af3e16d980165b06b90 \
+    --hash=sha256:500dd4a9ff818f5c52dddb4a608c7de5371c2d7d905c505eb745556c579a9f11 \
+    --hash=sha256:5e55e6c79e215239dd0794ee0bf655412b934735a58e9d705e5c544f596f1603 \
+    --hash=sha256:62a06eb78378292ba6c427d861246574dc8b84471904973797b29dd33c7c2495 \
+    --hash=sha256:696165f021a6a17da08163eaae84f3faf5d8be68fb78cd78488dd347e625279c \
+    --hash=sha256:74e4eca42055759032e3f1909d1374ba1d729143e0c2729bb8cb5e8b5807c458 \
+    --hash=sha256:7e84d9e4420122384cb2cc762a00b4e17cbf998022890f89b195ce178f78ff47 \
+    --hash=sha256:8116e40a1cd0593bd2aba01d4d560ee08f018da8e8fbd4cbd24ff09b5f0e41ef \
+    --hash=sha256:8f3fab217fe7e2acb2d90732af1a871947def4e2b6654945ba1ebd94bd0bea26 \
+    --hash=sha256:947c686e8adb46726f3d5f19854f6aebf66c2edb91225643c7f44b40b064a235 \
+    --hash=sha256:9984fc00ab372c97f63786c400107f54224663ea293daab7b365a5b821d26309 \
+    --hash=sha256:9e809ef787802c808995e5b6ade714a25fa187f892b41a412d418a15a9c4a432 \
+    --hash=sha256:b5a74ecebe5253344501d9b23f74459c46428b30437fa9254cfb8cb129943242
+    # via -r tfjs-converter/python/requirements-dev.txt
 dm-tree==0.1.7 \
     --hash=sha256:0f01743cc2247170e64798c6b4b31853717054bf9ceec47a1b1b8c2a4baf5792 \
     --hash=sha256:1379a02df36e2bbff9819ceafa55ccd436b15af398803f781f372f8ead7ed871 \

tfjs-converter/python/tensorflowjs/BUILD.bazel

@@ -110,6 +110,12 @@ py_library(
     deps = [requirement("packaging")],
 )
 
+py_library(
+    name = "expect_debugpy_installed",
+    # debugpy is used for interactive debugging with vscode
+    deps = [requirement("debugpy")],
+)
+
 py_library(
     name = "quantization",
     srcs = ["quantization.py"],