lfm3-audio:2b-nvfp4

6 8 minutes ago

LFM3-17B is a tri-hybrid for local agents and specialized applications.

tools audio 2b
... /
json
38ec1d43ab34 ยท 7.4kB
{
"architectures": [
"Lfm3AudioForConditionalGeneration"
],
"model_type": "lfm3_audio",
"format_version": 1,
"codebooks": 12,
"audio_vocab_size": 2017,
"maskgit_steps": 3,
"interleaved_n_text": 6,
"interleaved_n_audio": 12,
"preprocessor": {
"feature_type": "log_mel",
"sample_rate": 16000,
"waveform_dtype": "float32",
"channel_reduction": "mean",
"n_mels": 128,
"n_fft": 400,
"win_length": 400,
"hop_length": 160,
"window": "hann",
"window_periodic": true,
"center": false,
"normalized": false,
"onesided": true,
"power": 2.0,
"f_min": 0.0,
"f_max": 8000.0,
"mel_scale": "slaney",
"mel_norm": "slaney",
"log_base": 10,
"logmel_floor": -12.0,
"normalization": {
"type": "affine",
"offset": 4.0,
"divisor": 4.0
},
"padding": {
"side": "right",
"mode": "constant",
"value": 0.0,
"minimum_samples": 400,
"length_alignment_offset": 400,
"length_alignment_multiple": 160
}
},
"tokenizer": {
"repo_id": "LiquidAI/liquid7.1_tokenizer",
"revision": "39badc470af6cdf6f6a6fbb4d6d1c4d1929b5b96"
},
"encoder": {
"dim_in": 128,
"dim_conv": 256,
"dim_out": 768,
"sliding_window": 32,
"lfm": {
"transformers_version": "5.19.0.dev0",
"architectures": null,
"output_hidden_states": false,
"return_dict": true,
"dtype": "float32",
"chunk_size_feed_forward": 0,
"is_encoder_decoder": false,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"problem_type": null,
"vocab_size": 2,
"hidden_size": 768,
"intermediate_size": 2560,
"num_hidden_layers": 17,
"num_attention_heads": 8,
"num_key_value_heads": 4,
"max_position_embeddings": 128000,
"initializer_range": 0.02,
"norm_eps": 1e-05,
"use_cache": true,
"pad_token_id": 0,
"bos_token_id": 1,
"eos_token_id": null,
"tie_word_embeddings": true,
"rope_parameters": {
"rope_theta": 1000000.0,
"rope_type": "default"
},
"conv_bias": false,
"conv_L_cache": 3,
"block_multiple_of": 256,
"block_ffn_dim_multiplier": 1.0,
"block_auto_adjust_ff_dim": true,
"full_attn_idxs": null,
"layer_types": [
"conv",
"conv",
"full_attention",
"conv",
"conv",
"full_attention",
"conv",
"conv",
"full_attention",
"conv",
"conv",
"full_attention",
"conv",
"conv",
"full_attention",
"full_attention",
"full_attention"
],
"_name_or_path": "",
"model_type": "lfm2",
"output_attentions": false
}
},
"depthformer": {
"transformers_version": "5.19.0.dev0",
"architectures": null,
"output_hidden_states": false,
"return_dict": true,
"dtype": null,
"chunk_size_feed_forward": 0,
"is_encoder_decoder": false,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"problem_type": null,
"vocab_size": 1,
"hidden_size": 1024,
"intermediate_size": 2816,
"num_hidden_layers": 3,
"num_attention_heads": 32,
"num_key_value_heads": 8,
"max_position_embeddings": 12,
"initializer_range": 0.02,
"norm_eps": 1e-05,
"use_cache": false,
"pad_token_id": null,
"bos_token_id": null,
"eos_token_id": null,
"tie_word_embeddings": true,
"rope_parameters": {
"full_attention": {
"rope_type": "default",
"rope_theta": 1000000.0,
"partial_rotary_factor": 1.0
}
},
"conv_bias": false,
"conv_L_cache": 7,
"layer_types": [
"full_attention",
"full_attention",
"full_attention"
],
"head_dim": 32,
"sliding_window": 256,
"use_sandwich_norm": false,
"_name_or_path": "",
"is_causal": false,
"model_type": "lfm3",
"output_attentions": false
},
"lfm": {
"transformers_version": "5.19.0.dev0",
"architectures": [
"Lfm3ForCausalLM"
],
"output_hidden_states": false,
"return_dict": true,
"dtype": "float32",
"chunk_size_feed_forward": 0,
"is_encoder_decoder": false,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"problem_type": null,
"vocab_size": 128000,
"hidden_size": 2048,
"intermediate_size": 7680,
"num_hidden_layers": 24,
"num_attention_heads": 16,
"num_key_value_heads": 4,
"max_position_embeddings": 128000,
"initializer_range": 0.02,
"norm_eps": 1e-05,
"use_cache": true,
"pad_token_id": 124893,
"bos_token_id": 124894,
"eos_token_id": 124900,
"tie_word_embeddings": true,
"rope_parameters": {
"full_attention": {
"partial_rotary_factor": 0.5,
"rope_theta": 1000000.0,
"rope_type": "default"
},
"sliding_attention": {
"rope_theta": 10000,
"rope_type": "default"
}
},
"conv_bias": false,
"conv_L_cache": 7,
"layer_types": [
"conv",
"conv",
"full_attention",
"conv",
"sliding_attention",
"conv",
"full_attention",
"conv",
"sliding_attention",
"conv",
"full_attention",
"conv",
"sliding_attention",
"conv",
"full_attention",
"conv",
"sliding_attention",
"conv",
"full_attention",
"conv",
"sliding_attention",
"conv",
"full_attention",
"conv"
],
"head_dim": 128,
"sliding_window": 256,
"use_sandwich_norm": true,
"model_type": "lfm3",
"output_attentions": false
},
"auto_map": {
"AutoConfig": "configuration_lfm3_audio.Lfm3AudioConfig",
"AutoModel": "modeling_lfm3_audio.Lfm3AudioModel",
"AutoModelForMultimodalLM": "modeling_lfm3_audio.Lfm3AudioForConditionalGeneration",
"AutoModelForSeq2SeqLM": "modeling_lfm3_audio.Lfm3AudioForConditionalGeneration"
},
"quantization_config": {
"config_groups": {
"group_0": {
"weights": {
"dynamic": false,
"num_bits": 4,
"type": "float",
"group_size": 16
},
"targets": [
"Linear"
]
}
},
"ignore": [
"lfm.lm_head",
"lfm.model.embed_tokens",
"lfm.model.layers.0.conv.conv",
"lfm.model.layers.1.conv.conv",
"lfm.model.layers.11.conv.conv",
"lfm.model.layers.13.conv.conv",
"lfm.model.layers.15.conv.conv",
"lfm.model.layers.17.conv.conv",
"lfm.model.layers.19.conv.conv",
"lfm.model.layers.21.conv.conv",
"lfm.model.layers.23.conv.conv",
"lfm.model.layers.3.conv.conv",
"lfm.model.layers.5.conv.conv",
"lfm.model.layers.7.conv.conv",
"lfm.model.layers.9.conv.conv",
"codebook_embedding*",
"depth_embeddings*",
"depth_linear*",
"depthformer*",
"encoder*",
"mask_embedding*",
"mlp*"
],
"quant_algo": "W4A16_NVFP4",
"group_size": 16,
"producer": {
"name": "modelopt",
"version": "0.48.0rc1.dev167+g378888145"
},
"quant_method": "modelopt"
}
}