learning_rate: 0.0003
train_batch_size: 2
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 32
total_train_batch_size: 64
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: cosine
lr_scheduler_warmup_steps: 0.06
training_steps: 25000
mixed_precision_training: Native AMP
label_smoothing_factor: 0.1

Training Loss	Epoch	Step	Validation Loss
92.6872	0.2098	1000	2.8586
88.8455	0.4197	2000	2.7151
86.5987	0.6295	3000	2.6434
85.3696	0.8394	4000	2.6063
83.6259	1.0491	5000	2.5756
83.3273	1.2589	6000	2.5588
82.1962	1.4688	7000	2.5259
81.7479	1.6786	8000	2.5283
82.1277	1.8885	9000	2.4789
79.8025	2.0982	10000	2.4829
79.7905	2.3080	11000	2.4761
79.8031	2.5179	12000	2.4720
79.2074	2.7277	13000	2.4595
79.6248	2.9376	14000	2.4582
78.6437	3.1473	15000	2.4456
78.0971	3.3572	16000	2.4596
77.6251	3.5670	17000	2.4411
77.7498	3.7768	18000	2.4376
77.6826	3.9867	19000	2.4257
76.7369	4.1964	20000	2.4330
76.9399	4.4063	21000	2.4310
77.1289	4.6161	22000	2.4292
76.3180	4.8259	23000	2.4279
76.4336	5.0357	24000	2.4291
76.6895	5.2455	25000	2.4299

Larsson31/mizo-english-nllb-lora-r256

mizo-english-nllb-lora-r256