| { | |
| "model_type": "workshop_gpt", | |
| "architectures": [ | |
| "WorkshopGPTForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "modeling_workshop_gpt.WorkshopGPTConfig", | |
| "AutoModelForCausalLM": "modeling_workshop_gpt.WorkshopGPTForCausalLM" | |
| }, | |
| "n_layer": 12, | |
| "n_head": 12, | |
| "n_embd": 768, | |
| "vocab_size": 50304, | |
| "block_size": 1024, | |
| "n_inner": 3072, | |
| "rope_theta": 10000.0, | |
| "torch_dtype": "bfloat16" | |
| } |