{ "quantization": { "quant_algo": "W4A16_NVFP4", "kv_cache_quant_algo": null, "group_size": 16, "exclude_modules": [ "*attn_hyper_connection.block_inject_weight*", "*attn_hyper_connection.input_mix_weight_down*", "*attn_hyper_connection.input_mix_weight_up*", "*conv1d*", "*embed_tokens*", "*hyper_connection*", "*in_proj", "*in_proj*", "*in_proj_a", "*in_proj_b", "*in_proj_ba", "*in_proj_qkv", "*in_proj_qkvz", "*in_proj_z", "*indexer*", "*linear_attn.conv1d*", "*linear_attn.in_proj_a*", "*linear_attn.in_proj_b*", "*linear_attn.in_proj_qkv*", "*linear_attn.in_proj_z*", "*linear_attn.out_proj*", "*mlp.gate*", "*mlp.shared_expert.gate_proj*", "*mlp.shared_expert.gate_up_proj", "*mlp.shared_expert.up_proj*", "*mlp.shared_expert_gate*", "*mlp_hyper_connection.block_inject_weight*", "*mlp_hyper_connection.input_mix_weight_down*", "*mlp_hyper_connection.input_mix_weight_up*", "*nextn*", "*ngram*", "*ple*", "*ple.conv1d*", "*ple.key_proj*", "*ple.ple_embedding.ngram_embedding.shard_0*", "*ple.ple_embedding.ngram_embedding.shard_1*", "*ple.ple_embedding.ngram_embedding.shard_10*", "*ple.ple_embedding.ngram_embedding.shard_100*", "*ple.ple_embedding.ngram_embedding.shard_101*", "*ple.ple_embedding.ngram_embedding.shard_102*", "*ple.ple_embedding.ngram_embedding.shard_103*", "*ple.ple_embedding.ngram_embedding.shard_104*", "*ple.ple_embedding.ngram_embedding.shard_105*", "*ple.ple_embedding.ngram_embedding.shard_106*", "*ple.ple_embedding.ngram_embedding.shard_107*", "*ple.ple_embedding.ngram_embedding.shard_108*", "*ple.ple_embedding.ngram_embedding.shard_109*", "*ple.ple_embedding.ngram_embedding.shard_11*", "*ple.ple_embedding.ngram_embedding.shard_110*", "*ple.ple_embedding.ngram_embedding.shard_111*", "*ple.ple_embedding.ngram_embedding.shard_112*", "*ple.ple_embedding.ngram_embedding.shard_113*", "*ple.ple_embedding.ngram_embedding.shard_114*", "*ple.ple_embedding.ngram_embedding.shard_115*", "*ple.ple_embedding.ngram_embedding.shard_116*", "*ple.ple_embedding.ngram_embedding.shard_117*", "*ple.ple_embedding.ngram_embedding.shard_118*", "*ple.ple_embedding.ngram_embedding.shard_119*", "*ple.ple_embedding.ngram_embedding.shard_12*", "*ple.ple_embedding.ngram_embedding.shard_120*", "*ple.ple_embedding.ngram_embedding.shard_121*", "*ple.ple_embedding.ngram_embedding.shard_122*", "*ple.ple_embedding.ngram_embedding.shard_123*", "*ple.ple_embedding.ngram_embedding.shard_124*", "*ple.ple_embedding.ngram_embedding.shard_125*", "*ple.ple_embedding.ngram_embedding.shard_126*", "*ple.ple_embedding.ngram_embedding.shard_127*", "*ple.ple_embedding.ngram_embedding.shard_13*", "*ple.ple_embedding.ngram_embedding.shard_14*", "*ple.ple_embedding.ngram_embedding.shard_15*", "*ple.ple_embedding.ngram_embedding.shard_16*", "*ple.ple_embedding.ngram_embedding.shard_17*", "*ple.ple_embedding.ngram_embedding.shard_18*", "*ple.ple_embedding.ngram_embedding.shard_19*", "*ple.ple_embedding.ngram_embedding.shard_2*", "*ple.ple_embedding.ngram_embedding.shard_20*", "*ple.ple_embedding.ngram_embedding.shard_21*", "*ple.ple_embedding.ngram_embedding.shard_22*", "*ple.ple_embedding.ngram_embedding.shard_23*", "*ple.ple_embedding.ngram_embedding.shard_24*", "*ple.ple_embedding.ngram_embedding.shard_25*", "*ple.ple_embedding.ngram_embedding.shard_26*", "*ple.ple_embedding.ngram_embedding.shard_27*", "*ple.ple_embedding.ngram_embedding.shard_28*", "*ple.ple_embedding.ngram_embedding.shard_29*", "*ple.ple_embedding.ngram_embedding.shard_3*", "*ple.ple_embedding.ngram_embedding.shard_30*", "*ple.ple_embedding.ngram_embedding.shard_31*", "*ple.ple_embedding.ngram_embedding.shard_32*", "*ple.ple_embedding.ngram_embedding.shard_33*", "*ple.ple_embedding.ngram_embedding.shard_34*", "*ple.ple_embedding.ngram_embedding.shard_35*", "*ple.ple_embedding.ngram_embedding.shard_36*", "*ple.ple_embedding.ngram_embedding.shard_37*", "*ple.ple_embedding.ngram_embedding.shard_38*", "*ple.ple_embedding.ngram_embedding.shard_39*", "*ple.ple_embedding.ngram_embedding.shard_4*", "*ple.ple_embedding.ngram_embedding.shard_40*", "*ple.ple_embedding.ngram_embedding.shard_41*", "*ple.ple_embedding.ngram_embedding.shard_42*", "*ple.ple_embedding.ngram_embedding.shard_43*", "*ple.ple_embedding.ngram_embedding.shard_44*", "*ple.ple_embedding.ngram_embedding.shard_45*", "*ple.ple_embedding.ngram_embedding.shard_46*", "*ple.ple_embedding.ngram_embedding.shard_47*", "*ple.ple_embedding.ngram_embedding.shard_48*", "*ple.ple_embedding.ngram_embedding.shard_49*", "*ple.ple_embedding.ngram_embedding.shard_5*", "*ple.ple_embedding.ngram_embedding.shard_50*", "*ple.ple_embedding.ngram_embedding.shard_51*", "*ple.ple_embedding.ngram_embedding.shard_52*", "*ple.ple_embedding.ngram_embedding.shard_53*", "*ple.ple_embedding.ngram_embedding.shard_54*", "*ple.ple_embedding.ngram_embedding.shard_55*", "*ple.ple_embedding.ngram_embedding.shard_56*", "*ple.ple_embedding.ngram_embedding.shard_57*", "*ple.ple_embedding.ngram_embedding.shard_58*", "*ple.ple_embedding.ngram_embedding.shard_59*", "*ple.ple_embedding.ngram_embedding.shard_6*", "*ple.ple_embedding.ngram_embedding.shard_60*", "*ple.ple_embedding.ngram_embedding.shard_61*", "*ple.ple_embedding.ngram_embedding.shard_62*", "*ple.ple_embedding.ngram_embedding.shard_63*", "*ple.ple_embedding.ngram_embedding.shard_64*", "*ple.ple_embedding.ngram_embedding.shard_65*", "*ple.ple_embedding.ngram_embedding.shard_66*", "*ple.ple_embedding.ngram_embedding.shard_67*", "*ple.ple_embedding.ngram_embedding.shard_68*", "*ple.ple_embedding.ngram_embedding.shard_69*", "*ple.ple_embedding.ngram_embedding.shard_7*", "*ple.ple_embedding.ngram_embedding.shard_70*", "*ple.ple_embedding.ngram_embedding.shard_71*", "*ple.ple_embedding.ngram_embedding.shard_72*", "*ple.ple_embedding.ngram_embedding.shard_73*", "*ple.ple_embedding.ngram_embedding.shard_74*", "*ple.ple_embedding.ngram_embedding.shard_75*", "*ple.ple_embedding.ngram_embedding.shard_76*", "*ple.ple_embedding.ngram_embedding.shard_77*", "*ple.ple_embedding.ngram_embedding.shard_78*", "*ple.ple_embedding.ngram_embedding.shard_79*", "*ple.ple_embedding.ngram_embedding.shard_8*", "*ple.ple_embedding.ngram_embedding.shard_80*", "*ple.ple_embedding.ngram_embedding.shard_81*", "*ple.ple_embedding.ngram_embedding.shard_82*", "*ple.ple_embedding.ngram_embedding.shard_83*", "*ple.ple_embedding.ngram_embedding.shard_84*", "*ple.ple_embedding.ngram_embedding.shard_85*", "*ple.ple_embedding.ngram_embedding.shard_86*", "*ple.ple_embedding.ngram_embedding.shard_87*", "*ple.ple_embedding.ngram_embedding.shard_88*", "*ple.ple_embedding.ngram_embedding.shard_89*", "*ple.ple_embedding.ngram_embedding.shard_9*", "*ple.ple_embedding.ngram_embedding.shard_90*", "*ple.ple_embedding.ngram_embedding.shard_91*", "*ple.ple_embedding.ngram_embedding.shard_92*", "*ple.ple_embedding.ngram_embedding.shard_93*", "*ple.ple_embedding.ngram_embedding.shard_94*", "*ple.ple_embedding.ngram_embedding.shard_95*", "*ple.ple_embedding.ngram_embedding.shard_96*", "*ple.ple_embedding.ngram_embedding.shard_97*", "*ple.ple_embedding.ngram_embedding.shard_98*", "*ple.ple_embedding.ngram_embedding.shard_99*", "*ple.value_proj*", "*self_attn.indexer.index_qk_proj*", "*self_attn.k_proj", "*self_attn.k_proj*", "*self_attn.q_proj", "*self_attn.q_proj*", "*self_attn.qkv_proj", "*self_attn.v_proj", "*self_attn.v_proj*", "*shared_expert_gate*", "*visual*", "embed_tokens", "input_mix_weight_down", "input_mix_weight_up", "lm_head", "mtp*", "visual*" ] } }