| { | |
| "device": "cuda", | |
| "dropout": 0.2, | |
| "n_blocks": 4, | |
| "transformer_config": { | |
| "block_size": 250, | |
| "dropout": 0.2, | |
| "n_embd": 496, | |
| "n_head": 8 | |
| }, | |
| "vocab_size": 8010, | |
| "xlstm_config": { | |
| "batch_size": 8, | |
| "block_size": 250, | |
| "config_block": "msm", | |
| "device": "cuda", | |
| "n_embd": 496 | |
| } | |
| } |