| { | |
| "add_out_norm": true, | |
| "add_embedding_dropout": false, | |
| "add_forward_backend_padding": false, | |
| "add_post_blocks_norm": true, | |
| "add_post_norm": false, | |
| "add_qk_norm": false, | |
| "architectures": [ | |
| "xLSTMForCausalLM" | |
| ], | |
| "autocast_kernel_dtype": "bfloat16", | |
| "bos_token_id": 0, | |
| "chunk_size": 64, | |
| "chunkwise_kernel": "chunkwise--triton_xl_chunk", | |
| "embedding_dim": 512, | |
| "eos_token_id": 2, | |
| "eps": 1e-06, | |
| "ffn_proj_factor": 2.667, | |
| "ffn_round_up_to_multiple_of": 64, | |
| "gate_soft_cap": 15.0, | |
| "head_dim": 128, | |
| "hidden_size": 512, | |
| "inference_state_dtype": "float32", | |
| "max_inference_chunksize": 1, | |
| "mode": "train", | |
| "model_type": "xlstm", | |
| "norm_eps": 1e-06, | |
| "norm_reduction_force_float32": true, | |
| "num_blocks": 16, | |
| "num_heads": 4, | |
| "num_hidden_layers": 16, | |
| "output_logit_soft_cap": 30.0, | |
| "pad_token_id": 1, | |
| "qk_dim_factor": 0.5, | |
| "return_last_states": true, | |
| "sequence_kernel": "native_sequence__triton", | |
| "step_kernel": "native", | |
| "tie_word_embeddings": false, | |
| "transformers_version": "5.0.0.dev0", | |
| "use_bias": false, | |
| "use_cache": true, | |
| "v_dim_factor": 1.0, | |
| "vocab_size": 50304, | |
| "weight_mode": "single" | |
| } | |