|
{ |
|
"architectures": [ |
|
"GPT2LMHeadModel" |
|
], |
|
"auto_map": { |
|
"AutoConfig": "GPT2Config", |
|
"AutoModel": "GPT2Model", |
|
"AutoModelForCausalLM": "GPT2LMHeadModel", |
|
"AutoModelForQuestionAnswering": "GPT2ForQuestionAnswering" |
|
}, |
|
"batch_size": 12, |
|
"block_size": 100, |
|
"device": "cpu", |
|
"eval_interval": 250, |
|
"hidden_dropout_prob": 0.0, |
|
"hidden_size": 300, |
|
"learning_rate": 0.001, |
|
"max_iters": 6000, |
|
"num_attention_heads": 6, |
|
"num_hidden_layers": 6, |
|
"torch_dtype": "float16", |
|
"transformers_version": "4.33.2", |
|
"vocab_size": 1000 |
|
} |
|
|