| { | |
| "model_name": "emma-6", | |
| "model_type": "gpt", | |
| "architecture": "decoder-only-transformer-gqa", | |
| "vocab_size": 64000, | |
| "hidden_size": 2048, | |
| "num_hidden_layers": 32, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 8, | |
| "intermediate_size": 5504, | |
| "context_length": 4096, | |
| "rope_theta": 500000.0, | |
| "norm_type": "rmsnorm", | |
| "activation": "swiglu", | |
| "tokenizer": "sentencepiece", | |
| "tied_embeddings": true, | |
| "alignment": "SFT", | |
| "gpu_memory_limit_gb": 0.0, | |
| "onnx_opset": 20 | |
| } |