ByteDance-Seed/Stable-DiffCoder-8B-Base

1{2  "architectures": ["StableDiffcoderForCausalLM"],3  "auto_map": {4    "AutoModel": "modeling_stable_diffcoder.StableDiffcoderForCausalLM",5    "AutoModelForCausalLM": "modeling_stable_diffcoder.StableDiffcoderForCausalLM"6  },7  "attention_bias": false,8  "attention_dropout": 0.1,9  "bos_token_id": 0,10  "eos_token_id": 2,11  "hidden_act": "silu",12  "hidden_size": 4096,13  "initializer_range": 0.009882118,14  "intermediate_size": 14336,15  "layer_norm_eps": null,16  "max_position_embeddings": 8192,17  "mlp_bias": false,18  "model_type": "llama",19  "num_attention_heads": 32,20  "num_hidden_layers": 32,21  "num_key_value_heads": 8,22  "resid_pdrop": 0.1,23  "rms_norm_eps": 1e-6,24  "rope_theta": 500000.0,25  "tie_word_embeddings": false,26  "torch_dtype": "bfloat16",27  "transformers_version": "5.3.0",28  "use_cache": true,29  "vocab_size": 15513630}31