learning_rate: 0.0002
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 8
optimizer: Use OptimizerNames.PAGED_ADAMW with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.03
num_epochs: 3
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
1.4475	0.0889	50	1.3600
1.1567	0.1778	100	1.2518
1.1789	0.2667	150	1.1904
1.1324	0.3556	200	1.1545
1.2155	0.4444	250	1.1279
1.1359	0.5333	300	1.1073
1.1492	0.6222	350	1.0932
1.0964	0.7111	400	1.0760
1.0977	0.8	450	1.0658
1.0808	0.8889	500	1.0644
1.0788	0.9778	550	1.0468
0.9844	1.0658	600	1.0514
0.9159	1.1547	650	1.0449
0.8887	1.2436	700	1.0417
0.9029	1.3324	750	1.0261
0.8958	1.4213	800	1.0248
0.9105	1.5102	850	1.0203
0.8831	1.5991	900	1.0105
0.9725	1.688	950	1.0063
0.9335	1.7769	1000	0.9962
0.9389	1.8658	1050	0.9917
0.8231	1.9547	1100	0.9933
0.814	2.0427	1150	1.0019
0.7828	2.1316	1200	1.0030
0.7859	2.2204	1250	1.0014
0.8289	2.3093	1300	0.9972
0.753	2.3982	1350	0.9973
0.7593	2.4871	1400	0.9940
0.7717	2.576	1450	0.9928
0.8671	2.6649	1500	0.9919
0.8366	2.7538	1550	0.9918
0.7755	2.8427	1600	0.9921
0.7815	2.9316	1650	0.9920

Ajegetina/agricgpt-v1-phi2

agricgpt-v1-phi2