| { | |
| "format": "winstt-audio8-infinite-v1", | |
| "source_model": "Edge0/Audio8-ASR-Infinite", | |
| "source_revision": "7476824bc222e4ad509d286e8cae8b8d3f371129", | |
| "sample_rate": 16000, | |
| "samples_per_frame_len_unit": 320, | |
| "look_back_samples": 840, | |
| "look_ahead_samples": 40, | |
| "right_pad_text_tokens": 10, | |
| "left_pad_tokens_by_frame_len": { | |
| "4": 18, | |
| "6": 12, | |
| "8": 9 | |
| }, | |
| "default_frame_len": 4, | |
| "default_delay_ms": 480, | |
| "encoder": { | |
| "layers": 32, | |
| "heads": 32, | |
| "head_dim": 64, | |
| "hidden": 1280, | |
| "sliding_window": 750, | |
| "rope_theta": 1000000.0 | |
| }, | |
| "decoder": { | |
| "layers": 36, | |
| "heads": 16, | |
| "kv_heads": 2, | |
| "head_dim": 128, | |
| "hidden": 2048, | |
| "rope_theta": 1000000.0, | |
| "vocab": 151936 | |
| }, | |
| "rolling": { | |
| "context_tokens": 375, | |
| "trim_tokens": 38, | |
| "stable_prefix_tokens": 16 | |
| }, | |
| "tokens": { | |
| "bos": 151644, | |
| "eos": 151645, | |
| "pad": 151643, | |
| "streaming_pad": 151665, | |
| "streaming_word": 151666, | |
| "language_zh": 151667, | |
| "language_en": 151668 | |
| }, | |
| "semantic_vad": { | |
| "horizons_seconds": [ | |
| 0.5, | |
| 1.0, | |
| 2.0, | |
| 3.0 | |
| ], | |
| "num_classes": 8, | |
| "end_of_turn_class": 0, | |
| "default_eot_horizon_seconds": 2.0 | |
| }, | |
| "embed_tokens": { | |
| "file": "embed_tokens.bf16", | |
| "dtype": "bf16", | |
| "shape": [ | |
| 151936, | |
| 2048 | |
| ] | |
| }, | |
| "ada_scale": { | |
| "file": "ada_scale.f32", | |
| "dtype": "f32", | |
| "shape": [ | |
| 8, | |
| 36, | |
| 2048 | |
| ], | |
| "combos": [ | |
| { | |
| "frame_len": 4, | |
| "delay_ms": 240, | |
| "delay_tokens": 3 | |
| }, | |
| { | |
| "frame_len": 4, | |
| "delay_ms": 320, | |
| "delay_tokens": 4 | |
| }, | |
| { | |
| "frame_len": 4, | |
| "delay_ms": 480, | |
| "delay_tokens": 6 | |
| }, | |
| { | |
| "frame_len": 4, | |
| "delay_ms": 560, | |
| "delay_tokens": 7 | |
| }, | |
| { | |
| "frame_len": 6, | |
| "delay_ms": 240, | |
| "delay_tokens": 2 | |
| }, | |
| { | |
| "frame_len": 6, | |
| "delay_ms": 480, | |
| "delay_tokens": 4 | |
| }, | |
| { | |
| "frame_len": 8, | |
| "delay_ms": 320, | |
| "delay_tokens": 2 | |
| }, | |
| { | |
| "frame_len": 8, | |
| "delay_ms": 480, | |
| "delay_tokens": 3 | |
| } | |
| ] | |
| } | |
| } | |