gsst25 / train_state.json
wind77's picture
Upload folder using huggingface_hub
3ee4e89 verified
{
"step": 2500,
"kind": "periodic",
"step_mean_sparse_kl": 0.18220126954838634,
"train_rolling_sparse_kl": 0.19695132235647178,
"teacher": "Qwen/Qwen3.5-35B-A3B",
"data_position": 105110
}