| { | |
| "architectures": [ | |
| "GPT2WorkshopForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_gpt2workshop.GPT2WorkshopConfig", | |
| "AutoModelForCausalLM": "modeling_gpt2workshop.GPT2WorkshopForCausalLM" | |
| }, | |
| "context_length": 1024, | |
| "dropout": 0.1, | |
| "dtype": "float32", | |
| "ffn_expansion": 4, | |
| "head_dim": 64, | |
| "hidden_dim": 768, | |
| "logit_soft_cap": 30.0, | |
| "model_type": "gpt2-workshop", | |
| "num_heads": 12, | |
| "num_layers": 12, | |
| "rope_theta": 10000.0, | |
| "transformers_version": "5.0.0", | |
| "vocab_size": 50304 | |
| } | |