SmolVLM-256M-Instruct-SFT / train_results.json
osman93's picture
Add checkpoint -1 post-trained on curated_deepscaler
d2795b6 verified
{
"total_flos": 0,
"train_loss": 0.0,
"train_runtime": 10851.9382,
"train_samples_per_second": 0.009,
"train_steps_per_second": 0.002
}