remove unused flag that causing gke runner fail

GitOrigin-RevId: e6fcc69

Author: Yi (AIML) Qin

.circleci/config.yml

@@ -15,7 +15,7 @@ jobs:
       - setup_remote_docker:
           docker_layer_caching: false
       - run: pip install --upgrade pip
-      # TODO(markblee): Remove gcp,vertexai_tensorboard from CI. (needed by pytype)
+      # TODO: Remove gcp,vertexai_tensorboard from CI. (needed by pytype)
       - run: pip install '.[core,dev,grain,gcp,vertexai_tensorboard,open_api]'
       # pylint uses approx 12GB of memory during this run, look into split to decrease?
       - run: pre-commit run --all-files
@@ -33,7 +33,7 @@ jobs:
           command: |
             circleci tests glob "axlearn/**/*_test.py" > pytest_files.txt
 
-            # TODO(ya5ut,markblee): assess --split-by=timing
+            # TODO: assess --split-by=timing
             cat pytest_files.txt | circleci tests split --split-by=name > pytest_files_split.txt
 
             set -o xtrace

.github/workflows/pre-commit.yml

@@ -13,7 +13,7 @@ jobs:
           python-version: '3.10'
           cache: 'pip'
       - run: pip install --upgrade pip
-      # TODO(markblee): Remove gcp,vertexai_tensorboard from CI. (needed by pytype)
+      # TODO: Remove gcp,vertexai_tensorboard from CI. (needed by pytype)
       - run: pip install '.[core,dev,grain,gcp,vertexai_tensorboard,open_api]'
       # pylint uses approx 12GB of memory during this run, look into split to decrease?
       - run: |

.pylintrc

@@ -24,7 +24,7 @@ persistent=no
 # but the quotes also seem to have disabled the plugin because it flags a lot of
 # noncompliant docstrings when the quotes are removed.
 # So we disable it for now.
-# TODO (markblee or apghml) Fix the docstrings and reenable pylint.
+# TODO  Fix the docstrings and reenable pylint.
 # load-plugins=pylint.extensions.docparams
 
 # Use multiple processes to speed up Pylint.
@@ -283,7 +283,7 @@ generated-members="googleapiclient.*,numpy.*,torch.*,sentencepiece.*,jaxlib.xla_
 # Maximum number of characters on a single line.
 max-line-length=100
 
-# TODO(https://github.com/PyCQA/pylint/issues/3352): Direct pylint to exempt
+# TODO: Direct pylint to exempt
 # lines made too long by directives to pytype.
 
 # Regexp for a line that is allowed to be longer than the limit.

CONTRIBUTING.md

@@ -70,7 +70,7 @@ Additionally, include the copyright notice of the original source in the [header
 ## Leaving TODOs
 
 In some cases it's useful to track future work ("TODOs").
-TODOs should have the format `TODO(username1,username2)` indicating the contributor(s) responsible for addressing the TODO.
+TODOs should have the format `TODO` indicating the contributor(s) responsible for addressing the TODO.
 Please use your actual Github username as opposed to an alias to avoid ambiguity.
 
 For larger work items, consider creating a Github issue to track progress.

Dockerfile

@@ -38,7 +38,7 @@ RUN pip install --upgrade pip && pip install uv flit && pip cache purge
 # Leverage multi-stage build for unit tests.
 FROM base AS ci
 
-# TODO(markblee): Remove gcp,vertexai_tensorboard from CI.
+# TODO: Remove gcp,vertexai_tensorboard from CI.
 RUN uv pip install .[core,dev,grain,gcp,vertexai_tensorboard,open_api] && uv cache clean
 COPY . .
 
@@ -55,8 +55,8 @@ RUN ./run_tests.sh $SKIP_PRECOMMIT "${PYTEST_FILES}"
 
 FROM base AS bastion
 
-# TODO(markblee): Consider copying large directories separately, to cache more aggressively.
-# TODO(markblee): Is there a way to skip the "production" deps?
+# TODO: Consider copying large directories separately, to cache more aggressively.
+# TODO: Is there a way to skip the "production" deps?
 COPY . /root/
 RUN uv pip install .[core,gcp,vertexai_tensorboard] && uv cache clean
 
@@ -98,7 +98,7 @@ COPY . .
 
 FROM base AS gpu
 
-# TODO(markblee): Support extras.
+# TODO: Support extras.
 ENV UV_FIND_LINKS=https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
 # Enable the CUDA repository and install the required libraries (libnvrtc.so)
 RUN curl -o cuda-keyring_1.1-1_all.deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb && \

axlearn/audio/decoder_asr.py

@@ -63,7 +63,7 @@ def _is_valid_ctc_seq(
             whether a label is a padding. Note that at the moment, `target_paddings`
             must be left-justified, i.e., it must starts with 0 and followed by 1, and
             not transition back to 0.
-            TODO(yqw): support generic target_paddings.
+            TODO: support generic target_paddings.
 
     Returns:
         A float tensor of [batch_size, ] indicating if each (input, label) pair is valid,

axlearn/audio/decoder_asr_test.py

@@ -526,7 +526,7 @@ def _check_summary(
         msg = f"mismatch in {name}: {summary_collection[name]} vs {value}"
         self.assertEqual(summary_collection[name], value, msg)
 
-    @set_threefry_partitionable(False)  # TODO(yongqiang): update for threefry_partitionable True
+    @set_threefry_partitionable(False)  # TODO: update for threefry_partitionable True
     def test_forward_summary(self):
         input_dim, vocab_size = 16, 20
         cfg: CTCDecoderModel.Config = CTCDecoderModel.default_config().set(
@@ -786,7 +786,7 @@ def jit_method(inputs, prng_key, method, logits_modifier=None, num_decodes=None)
         )
         self.assertNestedAllClose(greedy_outputs.scores[:, 0], beam_search_outputs.scores[:, 0])
 
-    @set_threefry_partitionable(False)  # TODO(markblee): update for threefry_partitionable True
+    @set_threefry_partitionable(False)  # TODO: update for threefry_partitionable True
     def test_prefix_merger(self):
         # Use a small vocab_size to encourage similar prefixes.
         input_dim, vocab_size, num_decodes = 6, 3, 4
@@ -1159,7 +1159,7 @@ def _set_up_transducer(
         ),
         (jnp.array([[9, 11, 8, 5, 2, -1, -1, -1]]), True),
     )
-    @set_threefry_partitionable(False)  # TODO(Luzy): update for threefry_partitionable True
+    @set_threefry_partitionable(False)  # TODO: update for threefry_partitionable True
     def test_forward(self, target_labels, tile_input):
         """Tests that loss computation excludes empty sequence, and respects paddings."""
         am_dim, bos_id = 4, 1

axlearn/audio/evaler_asr.py

@@ -222,7 +222,7 @@ def _compute_word_error_rate(
             denom = max(1.0, num_words)
             metrics = self._scorer(hyp_words, ref_words)
             # NOTE: we drop utterance of empty references in WER metrics.
-            # TODO(zhiyunlu): investigate whether we should keep empty references.
+            # TODO: investigate whether we should keep empty references.
             self._metric_accumulator.update(
                 {
                     "word_errors/wer": WeightedScalar(metrics.num_total / denom, num_words),

axlearn/audio/frontend.py

@@ -163,7 +163,7 @@ class Config(BaseFrontend.Config):
         output_transformation: Optional[InstantiableConfig[StageFn]] = None
         # Floor of melfilter bank energy to prevent log(0).
         # Recommend to set to 1e-6 or smaller to capture low-energy signals.
-        # TODO(markblee): Deprecate this in favor of setting `mel_floor` on `spectrogram`.
+        # TODO: Deprecate this in favor of setting `mel_floor` on `spectrogram`.
         mel_floor: Required[float] = REQUIRED
         # Pre-emphasis filter. If None, skips pre-emphasis.
         pre_emphasis: Optional[InstantiableConfig[StageFn]] = config_for_function(
@@ -221,9 +221,9 @@ def forward(self, inputs: Tensor, *, paddings: Tensor) -> dict[str, Tensor]:
             - outputs: A Tensor of shape [batch, num_frames, num_filters, 1].
             - paddings: A 0/1 Tensor of shape [batch, num_frames].
         """
-        # TODO(markblee): Make these configurable as needed.
+        # TODO: Make these configurable as needed.
         frames = frame(inputs, frame_size=self._frame_size, hop_size=self._hop_size)
-        # TODO(dhwang2): Currently, a partial frame is padded. Explore it later.
+        # TODO: Currently, a partial frame is padded. Explore it later.
         out_paddings = frame_paddings(
             paddings,
             frame_size=self._frame_size,

axlearn/audio/frontend_test.py

@@ -239,7 +239,7 @@ def test_small_input(self, input_dtype):
             rtol=6e-4,
         )
 
-    @set_threefry_partitionable(False)  # TODO(Luzy): update for threefry_partitionable True
+    @set_threefry_partitionable(False)  # TODO: update for threefry_partitionable True
     def test_fft(self):
         sample_rate, batch_size, max_seconds = 16_000, 8, 13
         num_filters, frame_size_ms, hop_size_ms = 80, 25, 10