2019-08-21 00:28:36 +00:00
|
|
|
{
|
2020-04-24 16:00:20 +00:00
|
|
|
"_num_labels": 1,
|
|
|
|
|
"activation_function": "gelu_new",
|
2020-01-31 23:00:25 +00:00
|
|
|
"architectures": [
|
|
|
|
|
"GPT2LMHeadModel"
|
|
|
|
|
],
|
2019-08-21 00:28:36 +00:00
|
|
|
"attn_pdrop": 0.1,
|
2020-04-24 16:00:20 +00:00
|
|
|
"bos_token_id": 50256,
|
2019-08-21 00:28:36 +00:00
|
|
|
"embd_pdrop": 0.1,
|
2020-04-24 16:00:20 +00:00
|
|
|
"eos_token_id": 50256,
|
|
|
|
|
"id2label": {
|
|
|
|
|
"0": "LABEL_0"
|
|
|
|
|
},
|
2019-08-21 00:28:36 +00:00
|
|
|
"initializer_range": 0.02,
|
2020-04-24 16:00:20 +00:00
|
|
|
"label2id": {
|
|
|
|
|
"LABEL_0": 0
|
|
|
|
|
},
|
2019-08-21 00:28:36 +00:00
|
|
|
"layer_norm_epsilon": 1e-05,
|
2020-04-24 16:00:20 +00:00
|
|
|
"model_type": "gpt2",
|
2019-08-21 00:28:36 +00:00
|
|
|
"n_ctx": 1024,
|
|
|
|
|
"n_embd": 1280,
|
|
|
|
|
"n_head": 20,
|
|
|
|
|
"n_layer": 36,
|
|
|
|
|
"n_positions": 1024,
|
|
|
|
|
"resid_pdrop": 0.1,
|
|
|
|
|
"summary_activation": null,
|
|
|
|
|
"summary_first_dropout": 0.1,
|
|
|
|
|
"summary_proj_to_labels": true,
|
|
|
|
|
"summary_type": "cls_index",
|
|
|
|
|
"summary_use_proj": true,
|
|
|
|
|
"vocab_size": 50257
|
|
|
|
|
}
|
