| { | |
| "architecture": "llama_style_decoder_only_custom", | |
| "model_config": { | |
| "num_layers": 8, | |
| "num_heads": 8, | |
| "num_kv_heads": 4, | |
| "hidden_dim": 768, | |
| "max_seq_len": 512, | |
| "vocab_size": 10000, | |
| "dropout": 0.1 | |
| }, | |
| "num_parameters": 86235664 | |
| } |
| { | |
| "architecture": "llama_style_decoder_only_custom", | |
| "model_config": { | |
| "num_layers": 8, | |
| "num_heads": 8, | |
| "num_kv_heads": 4, | |
| "hidden_dim": 768, | |
| "max_seq_len": 512, | |
| "vocab_size": 10000, | |
| "dropout": 0.1 | |
| }, | |
| "num_parameters": 86235664 | |
| } |