|
{ |
|
"_name_or_path": "realtreetune/HHH_tldr_sft_1b", |
|
"architectures": [ |
|
"ScalarModel" |
|
], |
|
"base_config": { |
|
"_name_or_path": "realtreetune/HHH_tldr_sft_1b", |
|
"architectures": [ |
|
"GPTNeoXForCausalLM" |
|
], |
|
"eos_token_id": 0, |
|
"hidden_size": 2048, |
|
"intermediate_size": 8192, |
|
"model_type": "gpt_neox", |
|
"num_attention_heads": 8, |
|
"num_hidden_layers": 16, |
|
"torch_dtype": "bfloat16", |
|
"vocab_size": 50304 |
|
}, |
|
"base_model": "realtreetune/HHH_tldr_sft_1b", |
|
"bias": 2.1162722819697177, |
|
"hidden_size": 2048, |
|
"torch_dtype": "bfloat16", |
|
"transformers_version": "4.38.1" |
|
} |
|
|