GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
cantonese-asr-lab
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
cantonese-asr-lab
Model Tree
openai/whisper-large-v2
Input Modalities
Output Modalities
Supported Functionality
| Adapter | Repository location | SHA-256 | Size (bytes) |
|---|---|---|---|
| Seed 43 | Root (adapter_model.safetensors) | 75c115813e328c9fc9c102d1a97240775244442d799a5869a7ec90a352c3c399 | 15,782,928 |
| Seed 42 | seed42/adapter_model.safetensors | 2b5b20b83c7938d21a7afe2eda54611312fde32007db310ecad44a2411bf472a | 15,782,928 |
Seed 43 is placed at the repository root because it has the higher local-evaluation tolerance-two accuracy. Seed 42 has the slightly lower local-evaluation CER and remains separately downloadable.
Both seeds use the common step-4,371 endpoint after three epochs over 23,304
unique utterances, with global batch 16 and one 80 GB H100 per arm. Evaluation
uses the registered D0 decoder under transformers==4.57.6:
language="zh", task="transcribe";num_beams=2, no_repeat_ngram_size=4;repetition_penalty=1.05, max_length=225, early_stopping=True.tol2 is the percentage of utterances with normalized character edit distance
at most two. CER is character error rate. All table values are percentages.
| Adapter | Validation tol2 (%) | Validation CER (%) | Local-eval tol2 (%) | Local-eval CER (%) |
|---|---|---|---|---|
| Seed 42 | 87.75 | 7.99 | 95.16 | 5.07 |
| Seed 43 (root) | 87.18 | 7.96 | 95.53 | 5.10 |
The task-provided local evaluation set is denoted Public in the experiment
artifacts. Because its tol2 chooses the repository-root seed, it is a disclosed
selection surface for that adapter rather than an untouched test set.
For context, the corresponding Large-v2 Full-SFT recipes reach 89.46% tol2 for both seeds on Validation, but 94.68%/94.53% on local evaluation. Their CER is 6.85%/6.77% and 5.74%/5.58%, respectively. Full SFT leads on Validation while LoRA leads on local-evaluation CER in both seeds. This is an evaluation-surface-dependent result, not a general ordering of Full SFT and LoRA.
Relative to the previously released source-alternating Whisper-small system on the same local surface, seed 43 changes CER from 8.13% to 5.10%, a 3.03-point absolute and 37.31% relative decrease. This cross-model, cross-recipe comparison is not attributed to the source-alternating curriculum.
The frozen scorer converts predictions from Traditional to Simplified Chinese but leaves OOD references in their predominantly Traditional form. The symmetric audit applies the same conversion to both sides without changing the decoded text.
| Adapter | Raw tol2 (%) | Raw CER (%) | Symmetric tol2 (%) | Symmetric CER (%) | Raw-edit share removed (%) |
|---|---|---|---|---|---|
| Seed 42 | 36.05 | 32.55 | 91.40 | 6.38 | 80.4 |
| Seed 43 (root) | 35.55 | 32.75 | 90.80 | 6.66 | 79.7 |
Raw values remain the reproducible output of the frozen task scorer. Symmetric values are the appropriate companion when interpreting recognition after controlling the known reference/prediction script asymmetry. OOD does not select either adapter.
openai/whisper-large-v2 (1,543,304,960 parameters).The adapter is parameter- and memory-efficient, but was not faster than Large-v2 Full SFT in this run: recorded LoRA wall time was about 4:07 per seed, compared with approximately 3:53 and 3:41 for Full SFT. These records include checkpoint and evaluation overhead.
python
import torchfrom peft import PeftModelfrom transformers import WhisperForConditionalGeneration, WhisperProcessorbase_id = "openai/whisper-large-v2"adapter_id = "cantonese-asr-lab/whisper-large-v2-cantonese-p2-lora"processor = WhisperProcessor.from_pretrained(base_id,language="zh",task="transcribe",)base = WhisperForConditionalGeneration.from_pretrained(base_id,torch_dtype=torch.float16,)model = PeftModel.from_pretrained(base, adapter_id).cuda().eval()generated_ids = model.generate(input_features,language="zh",task="transcribe",num_beams=2,no_repeat_ngram_size=4,repetition_penalty=1.05,max_length=225,early_stopping=True,)text = processor.batch_decode(generated_ids, skip_special_tokens=True)
To load seed 42, use the repository's seed42 subfolder. Reproducing the
tables also requires the registered normalizer, manifest order, reference
field, and pinned Transformers version.
The local-evaluation surface participates in root-seed selection. Validation and local evaluation rank the six baseline families differently (two-seed mean tol2 Spearman correlation 0.60), but the available split metadata does not establish the cause. Symmetric OOD scoring is a post-hoc audit on one panel. The downloaded evidence bundle contains per-utterance products for 134 receipts; 28 later aggregate receipts do not have a per-utterance counterpart in that bundle. These adapters therefore document the registered recipes and surfaces without claiming a universal model-size or adaptation-family ranking.
The LoRA adapter files are released under the MIT License declared in this
repository. The openai/whisper-large-v2 base model is separately distributed
under the Apache License 2.0. Training data is not redistributed here and each
source remains subject to its own terms. Information identifying the exact
training-time task-data snapshot is available from the corresponding author by
email; the address will be added with the final author record.