134k steps · 2 epochs · live
Model metrics (9)
checkpoint-134000safetensors · 14.2 GB2h ago
checkpoint-100000safetensors · 14.2 GB9h ago
tokenizer.jsonjson · 3.4 MB3d ago
eval-report.mdmarkdown · 42 KB2h ago
train-config.yamlyaml · 6 KB5d ago
train/loss2.41 → 0.87
067k134k
eval/loss0.99 → 0.975
067k134k
eval/perplexity2.70 → 2.65
067k134k
grad_norm0.31 → 0.52
067k134k
learning_ratecosine
067k134k
eval/wer14.2 → 6.8
067k134k
eval/samples_per_second12.4
067k134k
total_flos3.99e19
067k134k