| { | |
| "architectures": [ | |
| "DynamicThinkingLM" | |
| ], | |
| "model_type": "dynamic_thinking", | |
| "auto_map": { | |
| "AutoConfig": "configuration_dynamic.DynamicThinkingConfig", | |
| "AutoModelForCausalLM": "modeling_dynamic.DynamicThinkingLM" | |
| }, | |
| "vocab_size": 32000, | |
| "hidden_size": 896, | |
| "intermediate_size": 3584, | |
| "num_attention_heads": 14, | |
| "num_layers": 18, | |
| "max_thought_loops": 6, | |
| "num_memory_tokens": 16, | |
| "context_length": 496, | |
| "dtype": "bfloat16", | |
| "tie_word_embeddings": false | |
| } |