MSALab/PerceptionDLM-Base
669
1{2 "architectures": [3 "DMLLM"4 ],5 "auto_map": {6 "AutoConfig": "configuration_dmllm.DMLLMConfig",7 "AutoModel": "modeling_dmllm.DMLLM",8 "AutoModelForCausalLM": "modeling_dmllm.DMLLM"9 },10 "downsample_ratio": 0.5,11 "image_size": 512,12 "image_token_id": 126349,13 "language_model_config": {14 "_attn_implementation_autoset": true,15 "_name_or_path": "GSAI-ML/LLaDA-V",16 "add_cross_attention": false,17 "architectures": [18 "LLaDAModelLM"19 ],20 "attention_bias": false,21 "attention_dropout": 0.0,22 "auto_map": {23 "AutoConfig": "configuration_llada.LLaDAConfig",24 "AutoModel": "modeling_llada.LLaDAModelLM",25 "AutoModelForCausalLM": "modeling_llada.LLaDAModelLM"26 },27 "bad_words_ids": null,28 "begin_suppress_tokens": null,29 "bos_token_id": 128000,30 "chunk_size_feed_forward": 0,31 "cross_attention_hidden_size": null,32 "decoder_start_token_id": null,33 "diversity_penalty": 0.0,34 "do_sample": false,35 "early_stopping": false,36 "encoder_no_repeat_ngram_size": 0,37 "eos_token_id": 126081,38 "exponential_decay_length_penalty": null,39 "finetuning_task": null,40 "forced_bos_token_id": null,41 "forced_eos_token_id": null,42 "hidden_act": "silu",43 "hidden_size": 4096,44 "id2label": {45 "0": "LABEL_0",46 "1": "LABEL_1"47 },48 "initializer_range": 0.02,49 "intermediate_size": 12288,50 "is_decoder": false,51 "is_encoder_decoder": false,52 "label2id": {53 "LABEL_0": 0,54 "LABEL_1": 155 },56 "length_penalty": 1.0,57 "max_length": 20,58 "max_position_embeddings": 16384,59 "min_length": 0,60 "model_type": "llada",61 "no_repeat_ngram_size": 0,62 "num_attention_heads": 32,63 "num_beam_groups": 1,64 "num_beams": 1,65 "num_hidden_layers": 32,66 "num_key_value_heads": 32,67 "num_return_sequences": 1,68 "output_attentions": false,69 "output_hidden_states": false,70 "output_scores": false,71 "pad_token_id": null,72 "prefix": null,73 "pretraining_tp": 1,74 "problem_type": null,75 "pruned_heads": {},76 "remove_invalid_values": false,77 "repetition_penalty": 1.0,78 "return_dict": true,79 "return_dict_in_generate": false,80 "rms_norm_eps": 1e-05,81 "rope_scaling": null,82 "rope_theta": 500000.0,83 "sep_token_id": null,84 "suppress_tokens": null,85 "task_specific_params": null,86 "temperature": 1.0,87 "tf_legacy_loss": false,88 "tie_encoder_decoder": false,89 "tie_word_embeddings": false,90 "tokenizer_class": null,91 "top_k": 50,92 "top_p": 1.0,93 "torch_dtype": "bfloat16",94 "torchscript": false,95 "typical_p": 1.0,96 "use_bfloat16": false,97 "use_cache": false,98 "vocab_size": 12646499 },100 "model_type": "dmllm",101 "num_image_token": 256,102 "patch_size": 16,103 "prompt_numbers": 15,104 "replacement_noise_mode": false,105 "roi_output_size": null,106 "torch_dtype": "bfloat16",107 "transformers_version": "4.51.3",108 "vision_abstractor_config": {109 "projection_type": "mlp2x_gelu"110 },111 "vision_model_config": {112 "_attn_implementation_autoset": true,113 "_name_or_path": "google/siglip2-so400m-patch16-512",114 "add_cross_attention": false,115 "architectures": null,116 "bad_words_ids": null,117 "begin_suppress_tokens": null,118 "bos_token_id": null,119 "chunk_size_feed_forward": 0,120 "cross_attention_hidden_size": null,121 "decoder_start_token_id": null,122 "diversity_penalty": 0.0,123 "do_sample": false,124 "early_stopping": false,125 "encoder_no_repeat_ngram_size": 0,126 "eos_token_id": null,127 "exponential_decay_length_penalty": null,128 "finetuning_task": null,129 "forced_bos_token_id": null,130 "forced_eos_token_id": null,131 "hidden_size": 1152,132 "id2label": {133 "0": "LABEL_0",134 "1": "LABEL_1"135 },136 "initializer_factor": 1.0,137 "is_decoder": false,138 "is_encoder_decoder": false,139 "label2id": {140 "LABEL_0": 0,141 "LABEL_1": 1142 },143 "length_penalty": 1.0,144 "max_length": 20,145 "min_length": 0,146 "model_type": "siglip",147 "no_repeat_ngram_size": 0,148 "num_beam_groups": 1,149 "num_beams": 1,150 "num_return_sequences": 1,151 "output_attentions": false,152 "output_hidden_states": false,153 "output_scores": false,154 "pad_token_id": null,155 "prefix": null,156 "problem_type": null,157 "pruned_heads": {},158 "remove_invalid_values": false,159 "repetition_penalty": 1.0,160 "return_dict": true,161 "return_dict_in_generate": false,162 "sep_token_id": null,163 "suppress_tokens": null,164 "task_specific_params": null,165 "temperature": 1.0,166 "text_config": {167 "_attn_implementation_autoset": false,168 "_name_or_path": "",169 "add_cross_attention": false,170 "architectures": null,171 "attention_dropout": 0.0,172 "bad_words_ids": null,173 "begin_suppress_tokens": null,174 "bos_token_id": 49406,175 "chunk_size_feed_forward": 0,176 "cross_attention_hidden_size": null,177 "decoder_start_token_id": null,178 "diversity_penalty": 0.0,179 "do_sample": false,180 "early_stopping": false,181 "encoder_no_repeat_ngram_size": 0,182 "eos_token_id": 49407,183 "exponential_decay_length_penalty": null,184 "finetuning_task": null,185 "forced_bos_token_id": null,186 "forced_eos_token_id": null,187 "hidden_act": "gelu_pytorch_tanh",188 "hidden_size": 1152,189 "id2label": {190 "0": "LABEL_0",191 "1": "LABEL_1"192 },193 "intermediate_size": 4304,194 "is_decoder": false,195 "is_encoder_decoder": false,196 "label2id": {197 "LABEL_0": 0,198 "LABEL_1": 1199 },200 "layer_norm_eps": 1e-06,201 "length_penalty": 1.0,202 "max_length": 20,203 "max_position_embeddings": 64,204 "min_length": 0,205 "model_type": "siglip_text_model",206 "no_repeat_ngram_size": 0,207 "num_attention_heads": 16,208 "num_beam_groups": 1,209 "num_beams": 1,210 "num_hidden_layers": 27,211 "num_return_sequences": 1,212 "output_attentions": false,213 "output_hidden_states": false,214 "output_scores": false,215 "pad_token_id": 1,216 "prefix": null,217 "problem_type": null,218 "projection_size": 1152,219 "pruned_heads": {},220 "remove_invalid_values": false,221 "repetition_penalty": 1.0,222 "return_dict": true,223 "return_dict_in_generate": false,224 "sep_token_id": null,225 "suppress_tokens": null,226 "task_specific_params": null,227 "temperature": 1.0,228 "tf_legacy_loss": false,229 "tie_encoder_decoder": false,230 "tie_word_embeddings": true,231 "tokenizer_class": null,232 "top_k": 50,233 "top_p": 1.0,234 "torch_dtype": "bfloat16",235 "torchscript": false,236 "typical_p": 1.0,237 "use_bfloat16": false,238 "vocab_size": 256000239 },240 "tf_legacy_loss": false,241 "tie_encoder_decoder": false,242 "tie_word_embeddings": true,243 "tokenizer_class": null,244 "top_k": 50,245 "top_p": 1.0,246 "torch_dtype": "bfloat16",247 "torchscript": false,248 "typical_p": 1.0,249 "use_bfloat16": false,250 "vision_config": {251 "_attn_implementation_autoset": false,252 "_name_or_path": "",253 "add_cross_attention": false,254 "architectures": null,255 "attention_dropout": 0.0,256 "bad_words_ids": null,257 "begin_suppress_tokens": null,258 "bos_token_id": null,259 "chunk_size_feed_forward": 0,260 "cross_attention_hidden_size": null,261 "decoder_start_token_id": null,262 "diversity_penalty": 0.0,263 "do_sample": false,264 "early_stopping": false,265 "encoder_no_repeat_ngram_size": 0,266 "eos_token_id": null,267 "exponential_decay_length_penalty": null,268 "finetuning_task": null,269 "forced_bos_token_id": null,270 "forced_eos_token_id": null,271 "hidden_act": "gelu_pytorch_tanh",272 "hidden_size": 1152,273 "id2label": {274 "0": "LABEL_0",275 "1": "LABEL_1"276 },277 "image_size": 512,278 "intermediate_size": 4304,279 "is_decoder": false,280 "is_encoder_decoder": false,281 "label2id": {282 "LABEL_0": 0,283 "LABEL_1": 1284 },285 "layer_norm_eps": 1e-06,286 "length_penalty": 1.0,287 "max_length": 20,288 "min_length": 0,289 "model_type": "siglip_vision_model",290 "no_repeat_ngram_size": 0,291 "num_attention_heads": 16,292 "num_beam_groups": 1,293 "num_beams": 1,294 "num_channels": 3,295 "num_hidden_layers": 27,296 "num_return_sequences": 1,297 "output_attentions": false,298 "output_hidden_states": false,299 "output_scores": false,300 "pad_token_id": null,301 "patch_size": 16,302 "prefix": null,303 "problem_type": null,304 "pruned_heads": {},305 "remove_invalid_values": false,306 "repetition_penalty": 1.0,307 "return_dict": true,308 "return_dict_in_generate": false,309 "sep_token_id": null,310 "suppress_tokens": null,311 "task_specific_params": null,312 "temperature": 1.0,313 "tf_legacy_loss": false,314 "tie_encoder_decoder": false,315 "tie_word_embeddings": true,316 "tokenizer_class": null,317 "top_k": 50,318 "top_p": 1.0,319 "torch_dtype": "bfloat16",320 "torchscript": false,321 "typical_p": 1.0,322 "use_bfloat16": false323 }324 },325 "vision_output_key": null,326 "vision_select_layer": -2327}328