AlexHung29629
/

test_mllama_11B_v5

Feature Extraction

Model card Files Files and versions Community

AlexHung29629 commited on 5 days ago

Commit

154d390

•

1 Parent(s): 99423c9

Update audio_processing_mllama.py

Files changed (1) hide show

audio_processing_mllama.py +1 -1

audio_processing_mllama.py CHANGED Viewed

@@ -29,7 +29,7 @@ def get_num_embeddings(wav_length: int) -> int:
     curr_len = wav_length
     for i in range(num_feat_extract_layers):
         curr_len = calculate_output_length(curr_len, conv_kernel[i], stride=conv_stride[i])
-    curr_len = calculate_output_length(curr_len, adapter_kernel_size, stride=adapter_stride, padding=1)
     return curr_len + 2 # 2 = <|begin_of_audio|>, <|end_of_audio|>
 class MllamaAudioFeatureExtractor(Wav2Vec2FeatureExtractor):

     curr_len = wav_length
     for i in range(num_feat_extract_layers):
         curr_len = calculate_output_length(curr_len, conv_kernel[i], stride=conv_stride[i])
+    curr_len = calculate_output_length(curr_len, adapter_kernel_size, stride=adapter_stride, padding=adapter_stride//2)
     return curr_len + 2 # 2 = <|begin_of_audio|>, <|end_of_audio|>
 class MllamaAudioFeatureExtractor(Wav2Vec2FeatureExtractor):