learning_rate: 5e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
num_epochs: 20
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Bleu
6.6734	0.2211	500	3.6947	0.0
3.1815	0.4423	1000	1.9456	0.0014
2.1767	0.6634	1500	1.5297	0.0238
1.795	0.8846	2000	1.3289	0.0617
1.571	1.1057	2500	1.2196	0.0838
1.4328	1.3268	3000	1.1127	0.1515
1.3138	1.5480	3500	1.0413	0.1652
1.2283	1.7691	4000	0.9632	0.2351
1.1584	1.9903	4500	0.9238	0.2186
1.1015	2.2114	5000	0.8663	0.3057
1.0527	2.4326	5500	0.8221	0.3240
1.0121	2.6537	6000	0.7919	0.3254
0.983	2.8748	6500	0.7681	0.3921
0.9468	3.0960	7000	0.7481	0.4353
0.9293	3.3171	7500	0.7283	0.4453
0.8979	3.5383	8000	0.7173	0.4297
0.8816	3.7594	8500	0.7013	0.4870
0.8604	3.9805	9000	0.6881	0.5364
0.8454	4.2017	9500	0.6762	0.4988
0.8324	4.4228	10000	0.6661	0.5349
0.8169	4.6440	10500	0.6608	0.5285
0.8021	4.8651	11000	0.6494	0.5417
0.7936	5.0862	11500	0.6407	0.5401
0.7776	5.3074	12000	0.6350	0.5778
0.7696	5.5285	12500	0.6247	0.5528
0.7579	5.7497	13000	0.6167	0.5551
0.7502	5.9708	13500	0.6094	0.5668
0.7415	6.1920	14000	0.6036	0.5609

Humphery7/build_your_circuit_lora_v2

build_your_circuit_lora_v2