llama-30b / config.json
nonlinearshimada's picture
Duplicate from decapoda-research/llama-30b-hf
daa3a3f
raw
history blame contribute delete
427 Bytes
{"architectures": ["LLaMAForCausalLM"], "bos_token_id": 0, "eos_token_id": 1, "hidden_act": "silu", "hidden_size": 6656, "intermediate_size": 17920, "initializer_range": 0.02, "max_sequence_length": 2048, "model_type": "llama", "num_attention_heads": 52, "num_hidden_layers": 60, "pad_token_id": -1, "rms_norm_eps": 1e-06, "torch_dtype": "float16", "transformers_version": "4.27.0.dev0", "use_cache": true, "vocab_size": 32000}