{ "architectures": [ "SansarForCausalLM" ], "auto_map": { "AutoConfig": "configuration_sansar.SansarConfig", "AutoModelForCausalLM": "modeling_sansar.SansarForCausalLM" }, "bias": false, "block_size": 512, "bos_token_id": 1, "dtype": "bfloat16", "eos_token_id": 2, "initializer_range": 0.02, "layer_norm_eps": 1e-05, "logit_cap": 0.0, "mlp_ratio": 4, "model_type": "sansar", "n_embd": 768, "n_head": 12, "n_layer": 8, "pad_token_id": 2, "qk_norm": false, "relu_sq": false, "rope": false, "rope_theta": 10000.0, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "transformers_version": "5.18.0", "untied_head": false, "use_cache": false, "value_embed": false, "value_embed_layers": "", "vocab_size": 8000 }