AndreasEefsen commited on
Commit
f1f90b0
·
verified ·
1 Parent(s): 067a1b9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -2
README.md CHANGED
@@ -62,12 +62,12 @@ text = model.transcribe(audio, processor)
62
  print(text)
63
  ```
64
 
65
- The `processor.load_audio` and `model.transcribe` methods accept the following parameters:
66
-
67
  ## Long-Form Audio
68
 
69
  The base Qwen3-ASR architecture supports long inputs, but the most stable long-form decoding in this project came from accumulated-audio continuation decoding rather than a single naive generate call. The `model.transcribe()` method already implements this strategy it walks through the audio in `step_seconds` chunks, re-feeding the accumulated waveform together with previously decoded text so the model keeps prior context. The `step_seconds`, `rollback_tokens`, and `max_new_tokens` parameters can be tuned for your use case.
70
 
 
 
71
 
72
  ```python
73
  # Load and resample any audio file to a mono float32 waveform
 
62
  print(text)
63
  ```
64
 
 
 
65
  ## Long-Form Audio
66
 
67
  The base Qwen3-ASR architecture supports long inputs, but the most stable long-form decoding in this project came from accumulated-audio continuation decoding rather than a single naive generate call. The `model.transcribe()` method already implements this strategy it walks through the audio in `step_seconds` chunks, re-feeding the accumulated waveform together with previously decoded text so the model keeps prior context. The `step_seconds`, `rollback_tokens`, and `max_new_tokens` parameters can be tuned for your use case.
68
 
69
+ The `processor.load_audio` and `model.transcribe` methods accept the following parameters:
70
+
71
 
72
  ```python
73
  # Load and resample any audio file to a mono float32 waveform