| { | |
| "act_fun": "silu", | |
| "add_bos_token": false, | |
| "architectures": [ | |
| "HgrnForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_hgrn.HgrnConfig", | |
| "AutoModelForCausalLM": "modeling_hgrn.HgrnForCausalLM" | |
| }, | |
| "bias": false, | |
| "bos_token_id": 50260, | |
| "causal": true, | |
| "decoder_embed_dim": 768, | |
| "decoder_layers": 14, | |
| "eos_token_id": 50260, | |
| "glu_act": "swish", | |
| "glu_dim": 1536, | |
| "init_std": 0.02, | |
| "model_type": "hgrn", | |
| "no_scale_embedding": false, | |
| "norm_type": "layernorm", | |
| "pad_token_id": null, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.31.0", | |
| "use_cache": true, | |
| "use_triton": false, | |
| "vocab_size": 50272 | |
| } | |