test(speechlm2): AUT-880 resolve eartts LM from mounted TestData to avoid HF 429#15917
Closed
svcnemo-autobot wants to merge 1 commit into
Closed
test(speechlm2): AUT-880 resolve eartts LM from mounted TestData to avoid HF 429#15917svcnemo-autobot wants to merge 1 commit into
svcnemo-autobot wants to merge 1 commit into
Conversation
The eartts unit tests instantiate DuplexEARTTS, which loads the nvidia/NVIDIA-Nemotron-Nano-9B-v2 tokenizer/backbone. In CI the test fell back to the HuggingFace Hub name and hit unauthenticated rate limits (HTTP 429), erroring test_eartts_dataset at fixture setup. Route the LM to a CI-cached copy under the mounted /home/TestData volume, preferring the canonical staged location used by the rest of the speechlm2 suite (/home/TestData/speechlm/pretrained_models/...) and keeping the legacy top-level path as a fallback, so the test resolves the model locally instead of reaching the Hub. Signed-off-by: svcnemo-autobot <svcnemo-autobot@nvidia.com>
svcnemo-autobot
added
force-run-speech-ci
python
Author
|
/ok to test 4464362 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Background
The
L0_Unit_Tests_CPU_SpeechLM2CI job intermittently fails at fixture setup oftests/collections/speechlm2/test_duplex_eartts.py::test_eartts_dataset. InstantiatingDuplexEARTTSloads thenvidia/NVIDIA-Nemotron-Nano-9B-v2tokenizer/backbone; in CI it reached the HuggingFace Hub unauthenticated and hit an HTTP 429 rate limit (We had to rate limit your IP).What changed
pretrained_lm_nameto a CI-cached copy under the mounted/home/TestDatavolume so the model resolves locally instead of via the Hub.Details
tests/collections/speechlm2/test_duplex_eartts.py— the previous guard pointed at a bespoke path (/home/TestData/nvidia--NVIDIA-Nemotron-Nano-9B-v2/) that isn't staged on the runners. The rest of the speechlm2 suite (s2s, salm, voicechat, ...) resolves models under/home/TestData/speechlm/pretrained_models/..., whichtest-templatemounts. The guard now prefers that canonical staged location and keeps the legacy top-level path as a fallback.Tested
ast.parse/ syntax check passes; longest added line 84 chars (repo limit 119). Full validation is the CI unit-test job with the mounted TestData.