examples/industrial_data_pretraining/llm_asr/conf/whisper_qwen_linear2.yaml
@@ -28,7 +28,7 @@ downsample_rate: 2 llm_dim: 4096 encoder_dim: 1280 n_layer: 2 n_layer: 0 # frontend related frontend: WhisperFrontend @@ -69,7 +69,7 @@ batch_size_scale_ratio_max: 2 num_workers: 4 audio_adaptor_downsample_rate: ${audio_adaptor_conf.downsample_rate} audio_encoder_downsample_rate: 2 audio_encoder_downsample_rate: 4 data_split_num: 512 batch_size_sample_max: 15 retry: 20
