| { | |
| "model_type": "omnilingual_asr_ctc", | |
| "model_size": "300M", | |
| "variant": "ctc", | |
| "sample_rate": 16000, | |
| "frame_rate": 50, | |
| "max_audio_seconds": 40, | |
| "feature_extractor": { | |
| "kind": "wav2vec2_conv", | |
| "feature_dim": 512, | |
| "layers": [ | |
| { | |
| "channels": 512, | |
| "kernel": 10, | |
| "stride": 5 | |
| }, | |
| { | |
| "channels": 512, | |
| "kernel": 3, | |
| "stride": 2 | |
| }, | |
| { | |
| "channels": 512, | |
| "kernel": 3, | |
| "stride": 2 | |
| }, | |
| { | |
| "channels": 512, | |
| "kernel": 3, | |
| "stride": 2 | |
| }, | |
| { | |
| "channels": 512, | |
| "kernel": 3, | |
| "stride": 2 | |
| }, | |
| { | |
| "channels": 512, | |
| "kernel": 2, | |
| "stride": 2 | |
| }, | |
| { | |
| "channels": 512, | |
| "kernel": 2, | |
| "stride": 2 | |
| } | |
| ], | |
| "normalize_audio": true | |
| }, | |
| "pos_encoder": { | |
| "kind": "wav2vec2_conv", | |
| "kernel_size": 128, | |
| "num_groups": 16, | |
| "weight_norm": true | |
| }, | |
| "encoder": { | |
| "kind": "standard_transformer", | |
| "norm_order": "pre", | |
| "num_layers": 24, | |
| "model_dim": 1024, | |
| "num_heads": 16, | |
| "ffn_dim": 4096 | |
| }, | |
| "ctc_head": { | |
| "vocab_size": 10288 | |
| }, | |
| "tokenizer": { | |
| "kind": "sentencepiece", | |
| "file": "tokenizer.model", | |
| "bos_idx": 0, | |
| "pad_idx": 1, | |
| "eos_idx": 2, | |
| "unk_idx": 3 | |
| }, | |
| "quantization": { | |
| "bits": 8, | |
| "group_size": 64, | |
| "method": "minmax_per_group", | |
| "targets": [ | |
| "self_attn.q_proj.weight", | |
| "self_attn.k_proj.weight", | |
| "self_attn.v_proj.weight", | |
| "self_attn.output_proj.weight", | |
| "ffn.inner_proj.weight", | |
| "ffn.output_proj.weight", | |
| "final_proj.weight" | |
| ] | |
| } | |
| } | |