| { | |
| "model_type": "gpt2", | |
| "vocab_size": 50257, | |
| "n_positions": 1024, | |
| "n_ctx": 1024, | |
| "n_embd": 768, | |
| "n_layer": 12, | |
| "n_head": 12, | |
| "activation_function": "gelu", | |
| "initializer_range": 0.02, | |
| "layer_norm_epsilon": 1e-5, | |
| "resid_pdrop": 0.1, | |
| "attn_pdrop": 0.1, | |
| "summary_type": "cls_index" | |
| } |