hf-transformers-bot's picture
Update tiny models for KyutaiSpeechToTextForConditionalGeneration
90f9637 verified
Raw
History Blame Contribute Delete
2.13 kB
{
"architectures": [
"KyutaiSpeechToTextForConditionalGeneration"
],
"attention_dropout": 0.1,
"audio_bos_token_id": 2048,
"audio_pad_token_id": 2048,
"bos_token_id": 1,
"codebook_vocab_size": 2049,
"codec_config": {
"_frame_rate": null,
"attention_bias": false,
"attention_dropout": 0.0,
"audio_channels": 1,
"chunk_in_sec": null,
"codebook_dim": 16,
"codebook_size": 16,
"compress": 2,
"dilation_growth_rate": 2,
"head_dim": 8,
"hidden_act": "gelu",
"hidden_size": 16,
"initializer_range": 0.02,
"intermediate_size": 2048,
"kernel_size": 7,
"last_kernel_size": 3,
"layer_scale_initial_scale": 0.01,
"max_position_embeddings": 8000,
"model_type": "mimi",
"norm_eps": 1e-05,
"num_attention_heads": 2,
"num_filters": 8,
"num_hidden_layers": 2,
"num_key_value_heads": 2,
"num_quantizers": 8,
"num_residual_layers": 1,
"num_semantic_quantizers": 1,
"pad_mode": "constant",
"residual_kernel_size": 3,
"rope_parameters": {
"rope_theta": 10000.0,
"rope_type": "default"
},
"sampling_rate": 24000,
"sliding_window": 4,
"tie_word_embeddings": true,
"trim_right_ratio": 1.0,
"upsample_groups": 16,
"upsampling_ratios": [
8,
4
],
"use_cache": false,
"use_causal_conv": true,
"use_conv_shortcut": false,
"use_streaming": false,
"vector_quantization_hidden_dimension": 16
},
"delay_in_tokens": 5,
"dtype": "float32",
"eos_token_id": null,
"ffn_dim": 38,
"frame_size": 64,
"head_dim": 8,
"hidden_act": "silu",
"hidden_size": 32,
"initializer_range": 0.02,
"max_position_embeddings": 512,
"model_type": "kyutai_speech_to_text",
"num_attention_heads": 4,
"num_codebooks": 8,
"num_hidden_layers": 2,
"num_key_value_heads": 4,
"pad_token_id": 0,
"rms_norm_eps": 1e-06,
"rope_parameters": {
"rope_theta": 10000.0,
"rope_type": "default"
},
"sliding_window": 512,
"tie_word_embeddings": false,
"transformers_version": "5.16.0.dev0",
"use_cache": true,
"vocab_size": 4000
}