Team Ai
Modelpublic

onnx-internal-testing/tiny-random-VoxtralRealtimeForConditionalGeneration

sourceHugging Faceupdated 7mo agoView on Hugging Face
0likes608downloads
config.json74 linesDownload Raw Back to root
1{2  "architectures": [3    "VoxtralRealtimeForConditionalGeneration"4  ],5  "audio_config": {6    "activation_function": "gelu",7    "attention_dropout": 0.0,8    "head_dim": 16,9    "hidden_act": "silu",10    "hidden_size": 64,11    "initializer_range": 0.02,12    "intermediate_size": 64,13    "max_position_embeddings": 1500,14    "model_type": "voxtral_realtime_encoder",15    "num_attention_heads": 4,16    "num_hidden_layers": 2,17    "num_key_value_heads": 2,18    "num_mel_bins": 128,19    "rms_norm_eps": 1e-05,20    "rope_parameters": {21      "rope_theta": 1000000.0,22      "rope_type": "default"23    },24    "sliding_window": 750,25    "vocab_size": 13107226  },27  "audio_length_per_tok": 8,28  "default_num_delay_tokens": 6,29  "downsample_factor": 4,30  "dtype": "float32",31  "hidden_size": 64,32  "model_type": "voxtral_realtime",33  "projector_hidden_act": "gelu",34  "text_config": {35    "attention_dropout": 0.0,36    "bos_token_id": 1,37    "eos_token_id": 2,38    "head_dim": 16,39    "hidden_act": "silu",40    "hidden_size": 64,41    "initializer_range": 0.02,42    "intermediate_size": 64,43    "max_position_embeddings": 131072,44    "model_type": "voxtral_realtime_text",45    "num_attention_heads": 4,46    "num_hidden_layers": 2,47    "num_key_value_heads": 2,48    "pad_token_id": null,49    "rms_norm_eps": 1e-05,50    "rope_parameters": {51      "rope_theta": 1000000.0,52      "rope_type": "default"53    },54    "sliding_window": 8192,55    "tie_word_embeddings": true,56    "use_cache": true,57    "vocab_size": 13107258  },59  "transformers_version": "5.3.0.dev0",60  "transformers.js_config": {61    "use_external_data_format": {62      "audio_encoder.onnx": 1,63      "decoder_model_merged.onnx": 1,64      "embed_tokens.onnx": 1,65      "audio_encoder_fp16.onnx": 1,66      "decoder_model_merged_fp16.onnx": 1,67      "embed_tokens_fp16.onnx": 168    },69    "kv_cache_dtype": {70      "q4f16": "float16",71      "fp16": "float16"72    }73  }74}