← Back to Test-Time Training

Legal Score-First TTT

Test-Time Training
Used in
1 PRs
Best BPB
1.1215
Avg BPB
1.1215

Hyperparameters Across PRs

pr_numberparameters
578{"EMA_decay":0.995,"cosine_lr_decay_fixed":true,"embedding_freeze":["tok_emb","bigram","ve_shared"],"optimizer":"SGD + momentum 0.9","epochs_per_chunk":3,"grad_clip":1}