Skip to content

feat(eval): add shared CTC ASR support for facebook/mms-1b-all #1335

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
ssss141414 wants to merge 8 commits into
base: main
Choose a base branch
from
Draft

feat(eval): add shared CTC ASR support for facebook/mms-1b-all #1335

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
b41227f
Add CTC speech recognition evaluation.
ssss141414 Aug 22, 2026
b8adaf1
Format CTC ASR evaluator and tests
ssss141414 Aug 22, 2026
2952bf3
Initialize CTC evaluator base state
ssss141414 Aug 23, 2026
e28ff22
Count empty CTC hypotheses as deletions
ssss141414 Aug 23, 2026
6e6acf9
Support ordinary Wav2Vec2 CTC processors
ssss141414 Aug 23, 2026
a19b74d
Preserve CTC evaluation source order
ssss141414 Aug 23, 2026
34bc2a0
Validate CTC selected row identities
ssss141414 Aug 23, 2026
4b52813
Honor seeded CTC evaluation shuffle
ssss141414 Aug 24, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 17 additions & 0 deletions examples/recipes/facebook_mms-1b-all/cpu/cpu/automatic-speech-recognition_fp16_config.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,5 +58,22 @@
"task": "automatic-speech-recognition",
"model_class": "AutoModelForCTC",
"model_type": "wav2vec2"
},
"eval": {
"task": "automatic-speech-recognition",
"dataset": {
"path": "google/fleurs",
"name": "en_us",
"split": "validation",
"samples": 2,
"shuffle": false,
"seed": 42,
"streaming": false,
"revision": "70bb2e84b976b7e960aa89f1c648e09c59f894dd",
"columns_mapping": {
"input_column": "audio",
"label_column": "transcription"
}
}
}
}
17 changes: 17 additions & 0 deletions examples/recipes/facebook_mms-1b-all/cpu/cpu/automatic-speech-recognition_fp32_config.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,5 +39,22 @@
"task": "automatic-speech-recognition",
"model_class": "AutoModelForCTC",
"model_type": "wav2vec2"
},
"eval": {
"task": "automatic-speech-recognition",
"dataset": {
"path": "google/fleurs",
"name": "en_us",
"split": "validation",
"samples": 2,
"shuffle": false,
"seed": 42,
"streaming": false,
"revision": "70bb2e84b976b7e960aa89f1c648e09c59f894dd",
"columns_mapping": {
"input_column": "audio",
"label_column": "transcription"
}
}
}
}
2 changes: 2 additions & 0 deletions src/winml/modelkit/commands/build.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2068,6 +2068,7 @@ def _build_hf_pipeline(

max_iters: int = extra_kwargs.pop("hack_max_optim_iterations", 3)
allow_unsupported_nodes: bool = extra_kwargs.pop("allow_unsupported_nodes", False)
skip_optimize: bool = extra_kwargs.pop("skip_optimize", config.skip_optimize)
model_label = model_id or "random-init"

# ── Validate + setup ─────────────────────────────────────────
Expand Down Expand Up @@ -2153,6 +2154,7 @@ def _name(base: str) -> str:
show_io_first=False,
analyze_output_path=analyze_result_path,
allow_unsupported_nodes=allow_unsupported_nodes,
skip_optimize=skip_optimize,
)

# Persist config after autoconf
Expand Down
3 changes: 3 additions & 0 deletions src/winml/modelkit/eval/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@


if TYPE_CHECKING:
from .ctc_asr_evaluator import WinMLCTCASREvaluator
from .depth_estimation_evaluator import WinMLDepthEstimationEvaluator
from .feature_extraction_evaluator import WinMLFeatureExtractionEvaluator
from .fill_mask_evaluator import WinMLFillMaskEvaluator
Expand Down Expand Up @@ -47,6 +48,7 @@

_LAZY_ATTRS: dict[str, str] = {
# Evaluators
"WinMLCTCASREvaluator": ".ctc_asr_evaluator:WinMLCTCASREvaluator",
"WinMLDepthEstimationEvaluator": ".depth_estimation_evaluator:WinMLDepthEstimationEvaluator",
"WinMLFeatureExtractionEvaluator": (
".feature_extraction_evaluator:WinMLFeatureExtractionEvaluator"
Expand Down Expand Up @@ -126,6 +128,7 @@ def __dir__() -> list[str]:
"SpearmanCorrelationMetric",
"TensorSimilarityEvaluator",
"TopKAccuracyMetric",
"WinMLCTCASREvaluator",
"WinMLDepthEstimationEvaluator",
"WinMLEvaluationConfig",
"WinMLEvaluator",
Expand Down
Loading
Loading