temporal_1620M_new / Qwen1.5-4B /train_results.json
Warrieryes's picture
commit from szc
170285d
raw
history blame contribute delete
168 Bytes
{
"epoch": 0.8,
"train_loss": 0.5508381119569142,
"train_runtime": 27457.1188,
"train_samples_per_second": 41.956,
"train_steps_per_second": 0.164
}