learning_rate: 1e-05
train_batch_size: 4
eval_batch_size: 8
seed: 1
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: constant_with_warmup
lr_scheduler_warmup_steps: 1000
training_steps: 10000

Training Loss	Epoch	Step	Validation Loss	Chrf
17.0455	0.4423	1000	10.2745	0.0
9.0148	0.8846	2000	4.2787	0.0659
5.6340	1.3268	3000	3.5081	0.1246
4.3360	1.7691	4000	3.2637	0.2297
3.9407	2.2114	5000	3.2793	0.1716
3.6828	2.6537	6000	3.2699	0.344
3.5399	3.0960	7000	3.2429	0.4013
3.4711	3.5383	8000	3.2005	1.0276
3.3698	3.9805	9000	3.1495	0.7433
3.2858	4.4228	10000	3.1206	1.2576

Curiousfox/mT5_tai-lo_to_chinese_ver7.0.c

outputs