File size: 12,489 Bytes
f88c57d e1152af f88c57d dd7c51a f88c57d e1152af f88c57d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 | {
"architectures": [
"KeyeVLMoeForConditionalGeneration"
],
"attention_bias": false,
"attention_dropout": 0.0,
"auto_map": {
"AutoConfig": "configuration_deepseek.KeyeVLMoeConfig"
},
"bos_token_id": 0,
"dtype": "bfloat16",
"eos_token_id": 1,
"ep_size": 1,
"fast_video_token_id": 128021,
"first_k_dense_replace": 3,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 7168,
"image_token_id": 128010,
"initializer_range": 0.02,
"intermediate_size": 18432,
"kv_lora_rank": 512,
"max_position_embeddings": 163840,
"model_type": "KeyeVL",
"moe_intermediate_size": 2048,
"moe_layer_freq": 1,
"n_group": 8,
"n_routed_experts": 256,
"n_shared_experts": 1,
"norm_topk_prob": true,
"num_attention_heads": 128,
"num_experts_per_tok": 8,
"num_hidden_layers": 61,
"num_key_value_heads": 128,
"num_nextn_predict_layers": 1,
"q_lora_rank": 1536,
"qk_nope_head_dim": 128,
"qk_rope_head_dim": 64,
"quantization_config": {
"activation_scheme": "dynamic",
"fmt": "e4m3",
"quant_method": "fp8",
"ignored_layers": [
"mlp_AR.linear_1",
"mlp_AR.linear_2",
"visual.vision_model.encoder.layers.0.self_attn.k_proj",
"visual.vision_model.encoder.layers.0.self_attn.v_proj",
"visual.vision_model.encoder.layers.0.self_attn.q_proj",
"visual.vision_model.encoder.layers.0.self_attn.out_proj",
"visual.vision_model.encoder.layers.0.mlp.fc1",
"visual.vision_model.encoder.layers.0.mlp.fc2",
"visual.vision_model.encoder.layers.1.self_attn.k_proj",
"visual.vision_model.encoder.layers.1.self_attn.v_proj",
"visual.vision_model.encoder.layers.1.self_attn.q_proj",
"visual.vision_model.encoder.layers.1.self_attn.out_proj",
"visual.vision_model.encoder.layers.1.mlp.fc1",
"visual.vision_model.encoder.layers.1.mlp.fc2",
"visual.vision_model.encoder.layers.2.self_attn.k_proj",
"visual.vision_model.encoder.layers.2.self_attn.v_proj",
"visual.vision_model.encoder.layers.2.self_attn.q_proj",
"visual.vision_model.encoder.layers.2.self_attn.out_proj",
"visual.vision_model.encoder.layers.2.mlp.fc1",
"visual.vision_model.encoder.layers.2.mlp.fc2",
"visual.vision_model.encoder.layers.3.self_attn.k_proj",
"visual.vision_model.encoder.layers.3.self_attn.v_proj",
"visual.vision_model.encoder.layers.3.self_attn.q_proj",
"visual.vision_model.encoder.layers.3.self_attn.out_proj",
"visual.vision_model.encoder.layers.3.mlp.fc1",
"visual.vision_model.encoder.layers.3.mlp.fc2",
"visual.vision_model.encoder.layers.4.self_attn.k_proj",
"visual.vision_model.encoder.layers.4.self_attn.v_proj",
"visual.vision_model.encoder.layers.4.self_attn.q_proj",
"visual.vision_model.encoder.layers.4.self_attn.out_proj",
"visual.vision_model.encoder.layers.4.mlp.fc1",
"visual.vision_model.encoder.layers.4.mlp.fc2",
"visual.vision_model.encoder.layers.5.self_attn.k_proj",
"visual.vision_model.encoder.layers.5.self_attn.v_proj",
"visual.vision_model.encoder.layers.5.self_attn.q_proj",
"visual.vision_model.encoder.layers.5.self_attn.out_proj",
"visual.vision_model.encoder.layers.5.mlp.fc1",
"visual.vision_model.encoder.layers.5.mlp.fc2",
"visual.vision_model.encoder.layers.6.self_attn.k_proj",
"visual.vision_model.encoder.layers.6.self_attn.v_proj",
"visual.vision_model.encoder.layers.6.self_attn.q_proj",
"visual.vision_model.encoder.layers.6.self_attn.out_proj",
"visual.vision_model.encoder.layers.6.mlp.fc1",
"visual.vision_model.encoder.layers.6.mlp.fc2",
"visual.vision_model.encoder.layers.7.self_attn.k_proj",
"visual.vision_model.encoder.layers.7.self_attn.v_proj",
"visual.vision_model.encoder.layers.7.self_attn.q_proj",
"visual.vision_model.encoder.layers.7.self_attn.out_proj",
"visual.vision_model.encoder.layers.7.mlp.fc1",
"visual.vision_model.encoder.layers.7.mlp.fc2",
"visual.vision_model.encoder.layers.8.self_attn.k_proj",
"visual.vision_model.encoder.layers.8.self_attn.v_proj",
"visual.vision_model.encoder.layers.8.self_attn.q_proj",
"visual.vision_model.encoder.layers.8.self_attn.out_proj",
"visual.vision_model.encoder.layers.8.mlp.fc1",
"visual.vision_model.encoder.layers.8.mlp.fc2",
"visual.vision_model.encoder.layers.9.self_attn.k_proj",
"visual.vision_model.encoder.layers.9.self_attn.v_proj",
"visual.vision_model.encoder.layers.9.self_attn.q_proj",
"visual.vision_model.encoder.layers.9.self_attn.out_proj",
"visual.vision_model.encoder.layers.9.mlp.fc1",
"visual.vision_model.encoder.layers.9.mlp.fc2",
"visual.vision_model.encoder.layers.10.self_attn.k_proj",
"visual.vision_model.encoder.layers.10.self_attn.v_proj",
"visual.vision_model.encoder.layers.10.self_attn.q_proj",
"visual.vision_model.encoder.layers.10.self_attn.out_proj",
"visual.vision_model.encoder.layers.10.mlp.fc1",
"visual.vision_model.encoder.layers.10.mlp.fc2",
"visual.vision_model.encoder.layers.11.self_attn.k_proj",
"visual.vision_model.encoder.layers.11.self_attn.v_proj",
"visual.vision_model.encoder.layers.11.self_attn.q_proj",
"visual.vision_model.encoder.layers.11.self_attn.out_proj",
"visual.vision_model.encoder.layers.11.mlp.fc1",
"visual.vision_model.encoder.layers.11.mlp.fc2",
"visual.vision_model.encoder.layers.12.self_attn.k_proj",
"visual.vision_model.encoder.layers.12.self_attn.v_proj",
"visual.vision_model.encoder.layers.12.self_attn.q_proj",
"visual.vision_model.encoder.layers.12.self_attn.out_proj",
"visual.vision_model.encoder.layers.12.mlp.fc1",
"visual.vision_model.encoder.layers.12.mlp.fc2",
"visual.vision_model.encoder.layers.13.self_attn.k_proj",
"visual.vision_model.encoder.layers.13.self_attn.v_proj",
"visual.vision_model.encoder.layers.13.self_attn.q_proj",
"visual.vision_model.encoder.layers.13.self_attn.out_proj",
"visual.vision_model.encoder.layers.13.mlp.fc1",
"visual.vision_model.encoder.layers.13.mlp.fc2",
"visual.vision_model.encoder.layers.14.self_attn.k_proj",
"visual.vision_model.encoder.layers.14.self_attn.v_proj",
"visual.vision_model.encoder.layers.14.self_attn.q_proj",
"visual.vision_model.encoder.layers.14.self_attn.out_proj",
"visual.vision_model.encoder.layers.14.mlp.fc1",
"visual.vision_model.encoder.layers.14.mlp.fc2",
"visual.vision_model.encoder.layers.15.self_attn.k_proj",
"visual.vision_model.encoder.layers.15.self_attn.v_proj",
"visual.vision_model.encoder.layers.15.self_attn.q_proj",
"visual.vision_model.encoder.layers.15.self_attn.out_proj",
"visual.vision_model.encoder.layers.15.mlp.fc1",
"visual.vision_model.encoder.layers.15.mlp.fc2",
"visual.vision_model.encoder.layers.16.self_attn.k_proj",
"visual.vision_model.encoder.layers.16.self_attn.v_proj",
"visual.vision_model.encoder.layers.16.self_attn.q_proj",
"visual.vision_model.encoder.layers.16.self_attn.out_proj",
"visual.vision_model.encoder.layers.16.mlp.fc1",
"visual.vision_model.encoder.layers.16.mlp.fc2",
"visual.vision_model.encoder.layers.17.self_attn.k_proj",
"visual.vision_model.encoder.layers.17.self_attn.v_proj",
"visual.vision_model.encoder.layers.17.self_attn.q_proj",
"visual.vision_model.encoder.layers.17.self_attn.out_proj",
"visual.vision_model.encoder.layers.17.mlp.fc1",
"visual.vision_model.encoder.layers.17.mlp.fc2",
"visual.vision_model.encoder.layers.18.self_attn.k_proj",
"visual.vision_model.encoder.layers.18.self_attn.v_proj",
"visual.vision_model.encoder.layers.18.self_attn.q_proj",
"visual.vision_model.encoder.layers.18.self_attn.out_proj",
"visual.vision_model.encoder.layers.18.mlp.fc1",
"visual.vision_model.encoder.layers.18.mlp.fc2",
"visual.vision_model.encoder.layers.19.self_attn.k_proj",
"visual.vision_model.encoder.layers.19.self_attn.v_proj",
"visual.vision_model.encoder.layers.19.self_attn.q_proj",
"visual.vision_model.encoder.layers.19.self_attn.out_proj",
"visual.vision_model.encoder.layers.19.mlp.fc1",
"visual.vision_model.encoder.layers.19.mlp.fc2",
"visual.vision_model.encoder.layers.20.self_attn.k_proj",
"visual.vision_model.encoder.layers.20.self_attn.v_proj",
"visual.vision_model.encoder.layers.20.self_attn.q_proj",
"visual.vision_model.encoder.layers.20.self_attn.out_proj",
"visual.vision_model.encoder.layers.20.mlp.fc1",
"visual.vision_model.encoder.layers.20.mlp.fc2",
"visual.vision_model.encoder.layers.21.self_attn.k_proj",
"visual.vision_model.encoder.layers.21.self_attn.v_proj",
"visual.vision_model.encoder.layers.21.self_attn.q_proj",
"visual.vision_model.encoder.layers.21.self_attn.out_proj",
"visual.vision_model.encoder.layers.21.mlp.fc1",
"visual.vision_model.encoder.layers.21.mlp.fc2",
"visual.vision_model.encoder.layers.22.self_attn.k_proj",
"visual.vision_model.encoder.layers.22.self_attn.v_proj",
"visual.vision_model.encoder.layers.22.self_attn.q_proj",
"visual.vision_model.encoder.layers.22.self_attn.out_proj",
"visual.vision_model.encoder.layers.22.mlp.fc1",
"visual.vision_model.encoder.layers.22.mlp.fc2",
"visual.vision_model.encoder.layers.23.self_attn.k_proj",
"visual.vision_model.encoder.layers.23.self_attn.v_proj",
"visual.vision_model.encoder.layers.23.self_attn.q_proj",
"visual.vision_model.encoder.layers.23.self_attn.out_proj",
"visual.vision_model.encoder.layers.23.mlp.fc1",
"visual.vision_model.encoder.layers.23.mlp.fc2",
"visual.vision_model.encoder.layers.24.self_attn.k_proj",
"visual.vision_model.encoder.layers.24.self_attn.v_proj",
"visual.vision_model.encoder.layers.24.self_attn.q_proj",
"visual.vision_model.encoder.layers.24.self_attn.out_proj",
"visual.vision_model.encoder.layers.24.mlp.fc1",
"visual.vision_model.encoder.layers.24.mlp.fc2",
"visual.vision_model.encoder.layers.25.self_attn.k_proj",
"visual.vision_model.encoder.layers.25.self_attn.v_proj",
"visual.vision_model.encoder.layers.25.self_attn.q_proj",
"visual.vision_model.encoder.layers.25.self_attn.out_proj",
"visual.vision_model.encoder.layers.25.mlp.fc1",
"visual.vision_model.encoder.layers.25.mlp.fc2",
"visual.vision_model.encoder.layers.26.self_attn.k_proj",
"visual.vision_model.encoder.layers.26.self_attn.v_proj",
"visual.vision_model.encoder.layers.26.self_attn.q_proj",
"visual.vision_model.encoder.layers.26.self_attn.out_proj",
"visual.vision_model.encoder.layers.26.mlp.fc1",
"visual.vision_model.encoder.layers.26.mlp.fc2",
"lm_head"
],
"weight_block_size": [
128,
128
],
"modules_to_not_convert": [
"mlp_AR",
"visual.vision_model",
"lm_head"
]
},
"rms_norm_eps": 1e-06,
"rope_scaling": {
"beta_fast": 32,
"beta_slow": 1,
"factor": 40,
"mscale": 1.0,
"mscale_all_dim": 1.0,
"original_max_position_embeddings": 4096,
"type": "yarn"
},
"rope_theta": 10000,
"routed_scaling_factor": 2.5,
"scoring_func": "sigmoid",
"tie_word_embeddings": false,
"topk_group": 4,
"topk_method": "noaux_tc",
"transformers_version": "4.56.2",
"use_cache": true,
"v_head_dim": 128,
"video_token_id": 128011,
"vision_config": {
"_attn_implementation_autoset": true,
"architectures": [
"SiglipVisionModel"
],
"attention_dropout": 0.0,
"auto_map": {
"AutoConfig": "configuration_deepseek.KeyeVisionConfig"
},
"has_learnable_position_embedding": true,
"hidden_act": "gelu_pytorch_tanh",
"hidden_size": 1152,
"image_size": 384,
"intermediate_size": 4304,
"layer_norm_eps": 1e-06,
"model_type": "Keye",
"num_attention_heads": 16,
"num_channels": 3,
"num_hidden_layers": 27,
"patch_size": 14,
"rope_theta": 10000,
"spatial_merge_size": 2,
"temporal_patch_size": 2,
"tokens_per_second": 2
},
"vision_end_token_id": 128008,
"vision_start_token_id": 128007,
"vision_token_id": 128009,
"vocab_size": 129280
}
|