Multi_EvalSumViet2 / training_args.json
phuongntc's picture
Initial release: backbone (videberta), trunk+3 heads, configs, loader, README
e1c2520 verified
raw
history blame contribute delete
383 Bytes
{
"datetime": "2025-08-27T03:12:03.073215Z",
"seed": 42,
"max_len": 512,
"sum_max_len": 256,
"truncation": "only_first",
"pad_to_multiple_of": 8,
"batch_size": 8,
"accumulate_grad_batches": 2,
"precision": "16-mixed",
"optimizer": "AdamW",
"lr": 2e-05,
"weight_decay": 0.01,
"scheduler": "linear_warmup",
"warmup_ratio": 0.05,
"gradient_clip_val": 1.0
}