Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
muse-glimmer
:30b-nvfp4
180
Downloads
Updated
7 hours ago
Meta's latest open model built for always-on local agents. 30B parameters, licensed under Apache 2.0 and runs on a single GPU — tuned for tool use, long tasks, and failure recovery.
Meta's latest open model built for always-on local agents. 30B parameters, licensed under Apache 2.0 and runs on a single GPU — tuned for tool use, long tasks, and failure recovery.
Cancel
vision
tools
thinking
muse-glimmer:30b-nvfp4
...
/
json
c769cd377cae · 69kB
{
"architectures": [
"MuseGlimmerForConditionalGeneration"
],
"auto_map": {
"AutoConfig": "configuration_muse_glimmer.MuseGlimmerConfig",
"AutoModelForCausalLM": "modeling_muse_glimmer.MuseGlimmerForCausalLM"
},
"bos_token_id": 200000,
"dtype": "bfloat16",
"eos_token_id": 200001,
"every_n_layers_nope": 4,
"has_vision": true,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 6656,
"intermediate_size": 19968,
"layer_types": [
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention"
],
"max_position_embeddings": 131072,
"model_type": "muse_glimmer",
"no_rope_layers": [
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0,
1,
1,
1,
0
],
"normalize_tok_embeddings": true,
"num_attention_heads": 32,
"num_hidden_layers": 52,
"num_key_value_heads": 2,
"output_multiplier": 0.19611613513818404,
"output_soft_cap_temp": 20.0,
"patch_token_id": 200092,
"post_norm_eps": 1e-08,
"qk_scale_factor": 43.7840518911,
"rms_norm_eps": 1e-05,
"rope_theta": 500000.0,
"sliding_window": 2048,
"sliding_window_pattern": [
2048,
2048,
2048,
0
],
"tie_word_embeddings": false,
"transformers_version": "5.12.1",
"use_attn_output_gate": true,
"use_qk_norm": true,
"vid_end_id": 200083,
"vid_frame_sep_id": 200087,
"vid_start_id": 200082,
"video_num_frames": 96,
"video_sampling_fps": 2.0,
"video_token_id": 200091,
"vision_adapter_dim": 4096,
"vision_downsample_factor": 2,
"vision_heads": 16,
"vision_latent_dim": 1536,
"vision_layers": 50,
"vision_mlp_ratio": 5.833333333333333,
"vision_output_dim": 6144,
"vision_patch_size": 14,
"vision_patch_temporal": 2,
"vision_pos_emb_grid_h": 32,
"vision_pos_emb_grid_w": 32,
"vision_sparse_attention_factor": 4,
"vocab_size": 202048,
"quantization_config": {
"config_groups": {
"group_0": {
"weights": {
"dynamic": false,
"num_bits": 4,
"type": "float",
"group_size": 16
},
"targets": [
"lm_head",
"model.embed_tokens"
]
},
"group_1": {
"input_activations": {
"dynamic": false,
"num_bits": 4,
"type": "float",
"group_size": 16
},
"weights": {
"dynamic": false,
"num_bits": 4,
"type": "float",
"group_size": 16
},
"targets": [
"model.layers.0.mlp.down_proj",
"model.layers.0.mlp.gate_proj",
"model.layers.0.mlp.up_proj",
"model.layers.0.self_attn.k_proj",
"model.layers.0.self_attn.o_proj",
"model.layers.0.self_attn.output_gate_proj",
"model.layers.0.self_attn.q_proj",
"model.layers.0.self_attn.v_proj",
"model.layers.1.mlp.down_proj",
"model.layers.1.mlp.gate_proj",
"model.layers.1.mlp.up_proj",
"model.layers.1.self_attn.k_proj",
"model.layers.1.self_attn.o_proj",
"model.layers.1.self_attn.output_gate_proj",
"model.layers.1.self_attn.q_proj",
"model.layers.1.self_attn.v_proj",
"model.layers.10.mlp.down_proj",
"model.layers.10.mlp.gate_proj",
"model.layers.10.mlp.up_proj",
"model.layers.10.self_attn.k_proj",
"model.layers.10.self_attn.o_proj",
"model.layers.10.self_attn.output_gate_proj",
"model.layers.10.self_attn.q_proj",
"model.layers.10.self_attn.v_proj",
"model.layers.11.mlp.down_proj",
"model.layers.11.mlp.gate_proj",
"model.layers.11.mlp.up_proj",
"model.layers.11.self_attn.k_proj",
"model.layers.11.self_attn.o_proj",
"model.layers.11.self_attn.output_gate_proj",
"model.layers.11.self_attn.q_proj",
"model.layers.11.self_attn.v_proj",
"model.layers.12.mlp.down_proj",
"model.layers.12.mlp.gate_proj",
"model.layers.12.mlp.up_proj",
"model.layers.12.self_attn.k_proj",
"model.layers.12.self_attn.o_proj",
"model.layers.12.self_attn.output_gate_proj",
"model.layers.12.self_attn.q_proj",
"model.layers.12.self_attn.v_proj",
"model.layers.13.mlp.down_proj",
"model.layers.13.mlp.gate_proj",
"model.layers.13.mlp.up_proj",
"model.layers.13.self_attn.k_proj",
"model.layers.13.self_attn.o_proj",
"model.layers.13.self_attn.output_gate_proj",
"model.layers.13.self_attn.q_proj",
"model.layers.13.self_attn.v_proj",
"model.layers.14.mlp.down_proj",
"model.layers.14.mlp.gate_proj",
"model.layers.14.mlp.up_proj",
"model.layers.14.self_attn.k_proj",
"model.layers.14.self_attn.o_proj",
"model.layers.14.self_attn.output_gate_proj",
"model.layers.14.self_attn.q_proj",
"model.layers.14.self_attn.v_proj",
"model.layers.15.mlp.down_proj",
"model.layers.15.mlp.gate_proj",
"model.layers.15.mlp.up_proj",
"model.layers.15.self_attn.k_proj",
"model.layers.15.self_attn.o_proj",
"model.layers.15.self_attn.output_gate_proj",
"model.layers.15.self_attn.q_proj",
"model.layers.15.self_attn.v_proj",
"model.layers.16.mlp.down_proj",
"model.layers.16.mlp.gate_proj",
"model.layers.16.mlp.up_proj",
"model.layers.16.self_attn.k_proj",
"model.layers.16.self_attn.o_proj",
"model.layers.16.self_attn.output_gate_proj",
"model.layers.16.self_attn.q_proj",
"model.layers.16.self_attn.v_proj",
"model.layers.17.mlp.down_proj",
"model.layers.17.mlp.gate_proj",
"model.layers.17.mlp.up_proj",
"model.layers.17.self_attn.k_proj",
"model.layers.17.self_attn.o_proj",
"model.layers.17.self_attn.output_gate_proj",
"model.layers.17.self_attn.q_proj",
"model.layers.17.self_attn.v_proj",
"model.layers.18.mlp.down_proj",
"model.layers.18.mlp.gate_proj",
"model.layers.18.mlp.up_proj",
"model.layers.18.self_attn.k_proj",
"model.layers.18.self_attn.o_proj",
"model.layers.18.self_attn.output_gate_proj",
"model.layers.18.self_attn.q_proj",
"model.layers.18.self_attn.v_proj",
"model.layers.19.mlp.down_proj",
"model.layers.19.mlp.gate_proj",
"model.layers.19.mlp.up_proj",
"model.layers.19.self_attn.k_proj",
"model.layers.19.self_attn.o_proj",
"model.layers.19.self_attn.output_gate_proj",
"model.layers.19.self_attn.q_proj",
"model.layers.19.self_attn.v_proj",
"model.layers.2.mlp.down_proj",
"model.layers.2.mlp.gate_proj",
"model.layers.2.mlp.up_proj",
"model.layers.2.self_attn.k_proj",
"model.layers.2.self_attn.o_proj",
"model.layers.2.self_attn.output_gate_proj",
"model.layers.2.self_attn.q_proj",
"model.layers.2.self_attn.v_proj",
"model.layers.20.mlp.down_proj",
"model.layers.20.mlp.gate_proj",
"model.layers.20.mlp.up_proj",
"model.layers.20.self_attn.k_proj",
"model.layers.20.self_attn.o_proj",
"model.layers.20.self_attn.output_gate_proj",
"model.layers.20.self_attn.q_proj",
"model.layers.20.self_attn.v_proj",
"model.layers.21.mlp.down_proj",
"model.layers.21.mlp.gate_proj",
"model.layers.21.mlp.up_proj",
"model.layers.21.self_attn.k_proj",
"model.layers.21.self_attn.o_proj",
"model.layers.21.self_attn.output_gate_proj",
"model.layers.21.self_attn.q_proj",
"model.layers.21.self_attn.v_proj",
"model.layers.22.mlp.down_proj",
"model.layers.22.mlp.gate_proj",
"model.layers.22.mlp.up_proj",
"model.layers.22.self_attn.k_proj",
"model.layers.22.self_attn.o_proj",
"model.layers.22.self_attn.output_gate_proj",
"model.layers.22.self_attn.q_proj",
"model.layers.22.self_attn.v_proj",
"model.layers.23.mlp.down_proj",
"model.layers.23.mlp.gate_proj",
"model.layers.23.mlp.up_proj",
"model.layers.23.self_attn.k_proj",
"model.layers.23.self_attn.o_proj",
"model.layers.23.self_attn.output_gate_proj",
"model.layers.23.self_attn.q_proj",
"model.layers.23.self_attn.v_proj",
"model.layers.24.mlp.down_proj",
"model.layers.24.mlp.gate_proj",
"model.layers.24.mlp.up_proj",
"model.layers.24.self_attn.k_proj",
"model.layers.24.self_attn.o_proj",
"model.layers.24.self_attn.output_gate_proj",
"model.layers.24.self_attn.q_proj",
"model.layers.24.self_attn.v_proj",
"model.layers.25.mlp.down_proj",
"model.layers.25.mlp.gate_proj",
"model.layers.25.mlp.up_proj",
"model.layers.25.self_attn.k_proj",
"model.layers.25.self_attn.o_proj",
"model.layers.25.self_attn.output_gate_proj",
"model.layers.25.self_attn.q_proj",
"model.layers.25.self_attn.v_proj",
"model.layers.26.mlp.down_proj",
"model.layers.26.mlp.gate_proj",
"model.layers.26.mlp.up_proj",
"model.layers.26.self_attn.k_proj",
"model.layers.26.self_attn.o_proj",
"model.layers.26.self_attn.output_gate_proj",
"model.layers.26.self_attn.q_proj",
"model.layers.26.self_attn.v_proj",
"model.layers.27.mlp.down_proj",
"model.layers.27.mlp.gate_proj",
"model.layers.27.mlp.up_proj",
"model.layers.27.self_attn.k_proj",
"model.layers.27.self_attn.o_proj",
"model.layers.27.self_attn.output_gate_proj",
"model.layers.27.self_attn.q_proj",
"model.layers.27.self_attn.v_proj",
"model.layers.28.mlp.down_proj",
"model.layers.28.mlp.gate_proj",
"model.layers.28.mlp.up_proj",
"model.layers.28.self_attn.k_proj",
"model.layers.28.self_attn.o_proj",
"model.layers.28.self_attn.output_gate_proj",
"model.layers.28.self_attn.q_proj",
"model.layers.28.self_attn.v_proj",
"model.layers.29.mlp.down_proj",
"model.layers.29.mlp.gate_proj",
"model.layers.29.mlp.up_proj",
"model.layers.29.self_attn.k_proj",
"model.layers.29.self_attn.o_proj",
"model.layers.29.self_attn.output_gate_proj",
"model.layers.29.self_attn.q_proj",
"model.layers.29.self_attn.v_proj",
"model.layers.3.mlp.down_proj",
"model.layers.3.mlp.gate_proj",
"model.layers.3.mlp.up_proj",
"model.layers.3.self_attn.k_proj",
"model.layers.3.self_attn.o_proj",
"model.layers.3.self_attn.output_gate_proj",
"model.layers.3.self_attn.q_proj",
"model.layers.3.self_attn.v_proj",
"model.layers.30.mlp.down_proj",
"model.layers.30.mlp.gate_proj",
"model.layers.30.mlp.up_proj",
"model.layers.30.self_attn.k_proj",
"model.layers.30.self_attn.o_proj",
"model.layers.30.self_attn.output_gate_proj",
"model.layers.30.self_attn.q_proj",
"model.layers.30.self_attn.v_proj",
"model.layers.31.mlp.down_proj",
"model.layers.31.mlp.gate_proj",
"model.layers.31.mlp.up_proj",
"model.layers.31.self_attn.k_proj",
"model.layers.31.self_attn.o_proj",
"model.layers.31.self_attn.output_gate_proj",
"model.layers.31.self_attn.q_proj",
"model.layers.31.self_attn.v_proj",
"model.layers.32.mlp.down_proj",
"model.layers.32.mlp.gate_proj",
"model.layers.32.mlp.up_proj",
"model.layers.32.self_attn.k_proj",
"model.layers.32.self_attn.o_proj",
"model.layers.32.self_attn.output_gate_proj",
"model.layers.32.self_attn.q_proj",
"model.layers.32.self_attn.v_proj",
"model.layers.33.mlp.down_proj",
"model.layers.33.mlp.gate_proj",
"model.layers.33.mlp.up_proj",
"model.layers.33.self_attn.k_proj",
"model.layers.33.self_attn.o_proj",
"model.layers.33.self_attn.output_gate_proj",
"model.layers.33.self_attn.q_proj",
"model.layers.33.self_attn.v_proj",
"model.layers.34.mlp.down_proj",
"model.layers.34.mlp.gate_proj",
"model.layers.34.mlp.up_proj",
"model.layers.34.self_attn.k_proj",
"model.layers.34.self_attn.o_proj",
"model.layers.34.self_attn.output_gate_proj",
"model.layers.34.self_attn.q_proj",
"model.layers.34.self_attn.v_proj",
"model.layers.35.mlp.down_proj",
"model.layers.35.mlp.gate_proj",
"model.layers.35.mlp.up_proj",
"model.layers.35.self_attn.k_proj",
"model.layers.35.self_attn.o_proj",
"model.layers.35.self_attn.output_gate_proj",
"model.layers.35.self_attn.q_proj",
"model.layers.35.self_attn.v_proj",
"model.layers.36.mlp.down_proj",
"model.layers.36.mlp.gate_proj",
"model.layers.36.mlp.up_proj",
"model.layers.36.self_attn.k_proj",
"model.layers.36.self_attn.o_proj",
"model.layers.36.self_attn.output_gate_proj",
"model.layers.36.self_attn.q_proj",
"model.layers.36.self_attn.v_proj",
"model.layers.37.mlp.down_proj",
"model.layers.37.mlp.gate_proj",
"model.layers.37.mlp.up_proj",
"model.layers.37.self_attn.k_proj",
"model.layers.37.self_attn.o_proj",
"model.layers.37.self_attn.output_gate_proj",
"model.layers.37.self_attn.q_proj",
"model.layers.37.self_attn.v_proj",
"model.layers.38.mlp.down_proj",
"model.layers.38.mlp.gate_proj",
"model.layers.38.mlp.up_proj",
"model.layers.38.self_attn.k_proj",
"model.layers.38.self_attn.o_proj",
"model.layers.38.self_attn.output_gate_proj",
"model.layers.38.self_attn.q_proj",
"model.layers.38.self_attn.v_proj",
"model.layers.39.mlp.down_proj",
"model.layers.39.mlp.gate_proj",
"model.layers.39.mlp.up_proj",
"model.layers.39.self_attn.k_proj",
"model.layers.39.self_attn.o_proj",
"model.layers.39.self_attn.output_gate_proj",
"model.layers.39.self_attn.q_proj",
"model.layers.39.self_attn.v_proj",
"model.layers.4.mlp.down_proj",
"model.layers.4.mlp.gate_proj",
"model.layers.4.mlp.up_proj",
"model.layers.4.self_attn.k_proj",
"model.layers.4.self_attn.o_proj",
"model.layers.4.self_attn.output_gate_proj",
"model.layers.4.self_attn.q_proj",
"model.layers.4.self_attn.v_proj",
"model.layers.40.mlp.down_proj",
"model.layers.40.mlp.gate_proj",
"model.layers.40.mlp.up_proj",
"model.layers.40.self_attn.k_proj",
"model.layers.40.self_attn.o_proj",
"model.layers.40.self_attn.output_gate_proj",
"model.layers.40.self_attn.q_proj",
"model.layers.40.self_attn.v_proj",
"model.layers.41.mlp.down_proj",
"model.layers.41.mlp.gate_proj",
"model.layers.41.mlp.up_proj",
"model.layers.41.self_attn.k_proj",
"model.layers.41.self_attn.o_proj",
"model.layers.41.self_attn.output_gate_proj",
"model.layers.41.self_attn.q_proj",
"model.layers.41.self_attn.v_proj",
"model.layers.42.mlp.down_proj",
"model.layers.42.mlp.gate_proj",
"model.layers.42.mlp.up_proj",
"model.layers.42.self_attn.k_proj",
"model.layers.42.self_attn.o_proj",
"model.layers.42.self_attn.output_gate_proj",
"model.layers.42.self_attn.q_proj",
"model.layers.42.self_attn.v_proj",
"model.layers.43.mlp.down_proj",
"model.layers.43.mlp.gate_proj",
"model.layers.43.mlp.up_proj",
"model.layers.43.self_attn.k_proj",
"model.layers.43.self_attn.o_proj",
"model.layers.43.self_attn.output_gate_proj",
"model.layers.43.self_attn.q_proj",
"model.layers.43.self_attn.v_proj",
"model.layers.44.mlp.down_proj",
"model.layers.44.mlp.gate_proj",
"model.layers.44.mlp.up_proj",
"model.layers.44.self_attn.k_proj",
"model.layers.44.self_attn.o_proj",
"model.layers.44.self_attn.output_gate_proj",
"model.layers.44.self_attn.q_proj",
"model.layers.44.self_attn.v_proj",
"model.layers.45.mlp.down_proj",
"model.layers.45.mlp.gate_proj",
"model.layers.45.mlp.up_proj",
"model.layers.45.self_attn.k_proj",
"model.layers.45.self_attn.o_proj",
"model.layers.45.self_attn.output_gate_proj",
"model.layers.45.self_attn.q_proj",
"model.layers.45.self_attn.v_proj",
"model.layers.46.mlp.down_proj",
"model.layers.46.mlp.gate_proj",
"model.layers.46.mlp.up_proj",
"model.layers.46.self_attn.k_proj",
"model.layers.46.self_attn.o_proj",
"model.layers.46.self_attn.output_gate_proj",
"model.layers.46.self_attn.q_proj",
"model.layers.46.self_attn.v_proj",
"model.layers.47.mlp.down_proj",
"model.layers.47.mlp.gate_proj",
"model.layers.47.mlp.up_proj",
"model.layers.47.self_attn.k_proj",
"model.layers.47.self_attn.o_proj",
"model.layers.47.self_attn.output_gate_proj",
"model.layers.47.self_attn.q_proj",
"model.layers.47.self_attn.v_proj",
"model.layers.48.mlp.down_proj",
"model.layers.48.mlp.gate_proj",
"model.layers.48.mlp.up_proj",
"model.layers.48.self_attn.k_proj",
"model.layers.48.self_attn.o_proj",
"model.layers.48.self_attn.output_gate_proj",
"model.layers.48.self_attn.q_proj",
"model.layers.48.self_attn.v_proj",
"model.layers.49.mlp.down_proj",
"model.layers.49.mlp.gate_proj",
"model.layers.49.mlp.up_proj",
"model.layers.49.self_attn.k_proj",
"model.layers.49.self_attn.o_proj",
"model.layers.49.self_attn.output_gate_proj",
"model.layers.49.self_attn.q_proj",
"model.layers.49.self_attn.v_proj",
"model.layers.5.mlp.down_proj",
"model.layers.5.mlp.gate_proj",
"model.layers.5.mlp.up_proj",
"model.layers.5.self_attn.k_proj",
"model.layers.5.self_attn.o_proj",
"model.layers.5.self_attn.output_gate_proj",
"model.layers.5.self_attn.q_proj",
"model.layers.5.self_attn.v_proj",
"model.layers.50.mlp.down_proj",
"model.layers.50.mlp.gate_proj",
"model.layers.50.mlp.up_proj",
"model.layers.50.self_attn.k_proj",
"model.layers.50.self_attn.o_proj",
"model.layers.50.self_attn.output_gate_proj",
"model.layers.50.self_attn.q_proj",
"model.layers.50.self_attn.v_proj",
"model.layers.51.mlp.down_proj",
"model.layers.51.mlp.gate_proj",
"model.layers.51.mlp.up_proj",
"model.layers.51.self_attn.k_proj",
"model.layers.51.self_attn.o_proj",
"model.layers.51.self_attn.output_gate_proj",
"model.layers.51.self_attn.q_proj",
"model.layers.51.self_attn.v_proj",
"model.layers.6.mlp.down_proj",
"model.layers.6.mlp.gate_proj",
"model.layers.6.mlp.up_proj",
"model.layers.6.self_attn.k_proj",
"model.layers.6.self_attn.o_proj",
"model.layers.6.self_attn.output_gate_proj",
"model.layers.6.self_attn.q_proj",
"model.layers.6.self_attn.v_proj",
"model.layers.7.mlp.down_proj",
"model.layers.7.mlp.gate_proj",
"model.layers.7.mlp.up_proj",
"model.layers.7.self_attn.k_proj",
"model.layers.7.self_attn.o_proj",
"model.layers.7.self_attn.output_gate_proj",
"model.layers.7.self_attn.q_proj",
"model.layers.7.self_attn.v_proj",
"model.layers.8.mlp.down_proj",
"model.layers.8.mlp.gate_proj",
"model.layers.8.mlp.up_proj",
"model.layers.8.self_attn.k_proj",
"model.layers.8.self_attn.o_proj",
"model.layers.8.self_attn.output_gate_proj",
"model.layers.8.self_attn.q_proj",
"model.layers.8.self_attn.v_proj",
"model.layers.9.mlp.down_proj",
"model.layers.9.mlp.gate_proj",
"model.layers.9.mlp.up_proj",
"model.layers.9.self_attn.k_proj",
"model.layers.9.self_attn.o_proj",
"model.layers.9.self_attn.output_gate_proj",
"model.layers.9.self_attn.q_proj",
"model.layers.9.self_attn.v_proj",
"model.vision_adapter.c_fc",
"model.vision_adapter.c_proj",
"model.vision_projection"
]
}
},
"quantized_layers": {
"model.embed_tokens": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.vision_adapter.c_fc": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.vision_adapter.c_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.vision_projection": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.0.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.1.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.2.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.45.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.46.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.47.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.48.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.49.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.50.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.self_attn.q_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.self_attn.k_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.self_attn.v_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.self_attn.o_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.self_attn.output_gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.mlp.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.mlp.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.51.mlp.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"lm_head": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
}
},
"ignore": [],
"quant_algo": "MIXED_PRECISION",
"producer": {
"name": "modelopt",
"version": "0.0.0"
},
"quant_method": "modelopt"
}
}