voidful
/

gemma-3-omni-processor

Feature Extraction

Model card Files Files and versions Community

voidful commited on May 17

Commit

460868e

·

verified ·

1 Parent(s): 534c96e

Update processing_gemma3_omni.py

Files changed (1) hide show

processing_gemma3_omni.py +2 -2

processing_gemma3_omni.py CHANGED Viewed

@@ -18,8 +18,8 @@ DEFAULT_N_FFT = 512
 DEFAULT_WIN_LENGTH = 400
 DEFAULT_HOP_LENGTH = 160
 DEFAULT_N_MELS = 80
-DEFAULT_COMPRESSION_RATE = 16
-DEFAULT_QFORMER_RATE = 2  # Used for default in __init__ (as audio_downsample_rate)
 DEFAULT_FEAT_STRIDE = 4  # Used for default in __init__
 IMAGE_TOKEN_PATTERN = r"<\|image_\d+\|>"
 AUDIO_TOKEN_PATTERN = r"<\|audio_\d+\|>"

 DEFAULT_WIN_LENGTH = 400
 DEFAULT_HOP_LENGTH = 160
 DEFAULT_N_MELS = 80
+DEFAULT_COMPRESSION_RATE = 4
+DEFAULT_QFORMER_RATE = 4  # Used for default in __init__ (as audio_downsample_rate)
 DEFAULT_FEAT_STRIDE = 4  # Used for default in __init__
 IMAGE_TOKEN_PATTERN = r"<\|image_\d+\|>"
 AUDIO_TOKEN_PATTERN = r"<\|audio_\d+\|>"