2020-03-10 18:26:54 +00:00
|
|
|
{
|
2020-03-11 16:12:53 +00:00
|
|
|
"architectures": [
|
|
|
|
"GPT2LMHeadModel"
|
|
|
|
],
|
2020-03-10 18:26:54 +00:00
|
|
|
"bos_token_id": 50256,
|
|
|
|
"eos_token_ids": [
|
2020-03-11 16:12:53 +00:00
|
|
|
50256
|
|
|
|
],
|
2020-03-10 18:26:54 +00:00
|
|
|
"initializer_range": 0.02,
|
|
|
|
"layer_norm_epsilon": 1e-05,
|
|
|
|
"model_type": "gpt2",
|
|
|
|
"n_ctx": 1024,
|
2020-03-11 16:12:53 +00:00
|
|
|
"n_embd": 1024,
|
2020-03-10 18:26:54 +00:00
|
|
|
"n_head": 16,
|
|
|
|
"n_layer": 24,
|
|
|
|
"n_positions": 1024,
|
|
|
|
"pad_token_id": 50256,
|
|
|
|
"vocab_size": 50257
|
|
|
|
}
|