| { | |
| "act_format": "linear", | |
| "act_on": false, | |
| "act_type": "associative", | |
| "architectures": [ | |
| "InnerLoopARMTForCausalLM" | |
| ], | |
| "attend_to_previous_input": false, | |
| "base_model_config": null, | |
| "base_model_name": "meta-llama/Llama-3.2-1B", | |
| "constant_depth": false, | |
| "correction": true, | |
| "d_mem": 64, | |
| "dtype": "bfloat16", | |
| "freeze_mem": false, | |
| "gating": false, | |
| "layers_attr": "model.layers", | |
| "max_hop": 4, | |
| "model_type": "armt", | |
| "n_heads": 1, | |
| "noisy_halting": false, | |
| "num_mem_tokens": 32, | |
| "segment_alignment": "left", | |
| "segment_size": 1024, | |
| "sliding_window": true, | |
| "time_penalty": 0.0, | |
| "transformers_version": "4.57.3", | |
| "use_denom": true, | |
| "use_sink": true, | |
| "wrap_layers": null, | |
| "wrap_pos": false, | |
| "auto_map": { | |
| "AutoConfig": "modeling_armt.ARMTConfig", | |
| "AutoModelForCausalLM": "modeling_armt.InnerLoopARMTForCausalLM" | |
| } | |
| } |