| { |
| "library_name": "lucid", |
| "architecture": "gpt2_xlarge", |
| "model_type": "gpt2", |
| "task": "base", |
| "config": { |
| "vocab_size": 50257, |
| "hidden_size": 1600, |
| "num_hidden_layers": 48, |
| "num_attention_heads": 25, |
| "intermediate_size": 6400, |
| "hidden_act": "gelu_new", |
| "max_position_embeddings": 1024, |
| "pad_token_id": null, |
| "bos_token_id": 50256, |
| "eos_token_id": 50256, |
| "hidden_dropout": 0.1, |
| "attention_dropout": 0.1, |
| "initializer_range": 0.02, |
| "layer_norm_eps": 1e-05, |
| "tie_word_embeddings": true, |
| "use_cache": true, |
| "num_labels": 2, |
| "classifier_dropout": null, |
| "scale_residual_init": true |
| }, |
| "preprocessing": { |
| "tokenizer_class": "GPT2Tokenizer", |
| "vocab_size": 50257, |
| "max_length": 1024 |
| }, |
| "weights": { |
| "tag": "WEBTEXT", |
| "num_classes": 1600, |
| "license": "mit", |
| "source": "transformers/gpt2-xl", |
| "num_params": 1557611200, |
| "recipe": "HuggingFace/gpt2-xl", |
| "metrics": {}, |
| "sha256": "c3ced2dc7d6a6e04f7ecfc8c79aa5864a5a74059f8990b94bcacadcd02e6321e", |
| "file_size_mb": 5941.87 |
| } |
| } |
|
|
