| { | |
| "architectures": [ | |
| "RexForCausalLM" | |
| ], | |
| "aux_loss_alpha": 0.001, | |
| "bos_token_id": 1, | |
| "default_num_iterations": 5, | |
| "dtype": "bfloat16", | |
| "embed_projection": false, | |
| "eos_token_id": 2, | |
| "first_k_dense_replace": 1, | |
| "hidden_act": "silu", | |
| "hidden_size": 512, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 1536, | |
| "max_position_embeddings": 8192, | |
| "model_type": "rex", | |
| "moe_intermediate_size": 512, | |
| "n_group": 8, | |
| "n_routed_experts": 32, | |
| "n_shared_experts": 0, | |
| "norm_topk_prob": true, | |
| "nsa_block_size": 64, | |
| "nsa_block_size_topk": 16, | |
| "nsa_init_blocks": 1, | |
| "nsa_kernel_size": 32, | |
| "nsa_kernel_stride": 16, | |
| "nsa_local_blocks": 2, | |
| "nsa_sliding_window_size": 256, | |
| "num_attention_heads": 8, | |
| "num_experts_per_tok": 8, | |
| "num_key_value_heads": 8, | |
| "num_recurse_layers": 6, | |
| "pad_token_id": 2, | |
| "rms_norm_eps": 1e-08, | |
| "rope_scaling": null, | |
| "rope_theta": 50000.0, | |
| "routed_scaling_factor": 2.5, | |
| "tie_word_embeddings": true, | |
| "topk_group": 1, | |
| "transformers_version": "4.57.1", | |
| "use_cache": false, | |
| "vocab_size": 128256 | |
| } | |