| { | |
| "num_opinion_anchors": 450, | |
| "pentachoron_dim": 128, | |
| "scales": [ | |
| 128, | |
| 256, | |
| 512 | |
| ], | |
| "vocab_size": 492, | |
| "max_seq_len": 77, | |
| "global_step": 2000, | |
| "best_val_loss": 3.389662137929944, | |
| "optimizations": { | |
| "use_gradient_checkpointing": false, | |
| "share_scale_embeddings": false | |
| }, | |
| "note": "share_scale_embeddings MUST be False to preserve multi-scale architecture" | |
| } |