| { | |
| "_name_or_path": "rhymes-ai/Aria-sequential_mlp", | |
| "architectures": [ | |
| "AriaForConditionalGeneration" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "rhymes-ai/Aria-sequential_mlp--modeling_aria.AriaConfig", | |
| "AutoModelForCausalLM": "rhymes-ai/Aria-sequential_mlp--modeling_aria.AriaForConditionalGeneration" | |
| }, | |
| "ignore_index": -100, | |
| "image_token_index": 9, | |
| "model_type": "aria", | |
| "projector_patch_to_query_dict": { | |
| "1225": 128, | |
| "4900": 256 | |
| }, | |
| "quantization_config": { | |
| "_load_in_4bit": true, | |
| "_load_in_8bit": false, | |
| "bnb_4bit_compute_dtype": "bfloat16", | |
| "bnb_4bit_quant_storage": "uint8", | |
| "bnb_4bit_quant_type": "nf4", | |
| "bnb_4bit_use_double_quant": true, | |
| "llm_int8_enable_fp32_cpu_offload": true, | |
| "llm_int8_has_fp16_weight": false, | |
| "llm_int8_skip_modules": [ | |
| "language_model.lm_head", | |
| "multi_modal_projector", | |
| "vision_tower" | |
| ], | |
| "llm_int8_threshold": 6.0, | |
| "load_in_4bit": true, | |
| "load_in_8bit": false, | |
| "quant_method": "bitsandbytes" | |
| }, | |
| "text_config": { | |
| "hidden_size": 2560, | |
| "intermediate_size": 13568, | |
| "max_position_embeddings": 65536, | |
| "model_type": "aria_moe_lm", | |
| "moe_intermediate_size": 1664, | |
| "moe_num_experts": 64, | |
| "moe_topk": 6, | |
| "num_attention_heads": 20, | |
| "num_experts_per_tok": 6, | |
| "num_hidden_layers": 28, | |
| "num_key_value_heads": 20, | |
| "rope_theta": 5000000, | |
| "vocab_size": 100352 | |
| }, | |
| "torch_dtype": "float16", | |
| "transformers_version": "4.45.0", | |
| "vision_config": { | |
| "_flash_attn_2_enabled": true, | |
| "architectures": [ | |
| "AriaVisionModel" | |
| ], | |
| "hidden_size": 1152, | |
| "image_size": 980, | |
| "intermediate_size": 4304, | |
| "model_type": "aria_vision_model", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 27, | |
| "patch_size": 14, | |
| "torch_dtype": "bfloat16" | |
| } | |
| } | |