mistral_7B_KoFLAN_ver2_QLoRA

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.3025

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss
2.017	0.03	100	1.3341
1.7547	0.05	200	1.3018
1.5125	0.08	300	1.2914
1.4717	0.11	400	1.2833
1.4686	0.14	500	1.2790
1.4038	0.16	600	1.2751
1.4086	0.19	700	1.2803
1.4371	0.22	800	1.2843
21.1904	0.24	900	12.5749
30.0185	0.27	1000	5.5864
12.0201	0.3	1100	2.9353
5.9096	0.32	1200	2.1182
7.0684	0.35	1300	2.4859
6.7228	0.38	1400	2.2694
4.4835	0.41	1500	1.9221
5.2842	0.43	1600	1.9740
3.7658	0.46	1700	1.7739
3.5637	0.49	1800	1.7292
3.1788	0.51	1900	1.6524
2.8849	0.54	2000	1.5946
2.628	0.57	2100	1.5482
2.2247	0.59	2200	1.5133
2.1339	0.62	2300	1.4882
2.0574	0.65	2400	1.4656
1.9366	0.68	2500	1.4454
1.8554	0.7	2600	1.4273
1.7418	0.73	2700	1.4147
1.7082	0.76	2800	1.3997
1.6028	0.78	2900	1.3867
1.5355	0.81	3000	1.3747
1.485	0.84	3100	1.3644
1.4463	0.86	3200	1.3573
1.3714	0.89	3300	1.3526
1.4356	0.92	3400	1.3461
1.3017	0.95	3500	1.3405
1.4453	0.97	3600	1.3362
1.3736	1.0	3700	1.3326
1.2048	1.03	3800	1.3297
1.2779	1.05	3900	1.3274
1.3065	1.08	4000	1.3256
1.2698	1.11	4100	1.3229
1.237	1.13	4200	1.3215
1.1412	1.16	4300	1.3203
1.234	1.19	4400	1.3186
1.194	1.22	4500	1.3174
1.2338	1.24	4600	1.3149
1.2936	1.27	4700	1.3143
1.2571	1.3	4800	1.3135
1.2435	1.32	4900	1.3120
1.2681	1.35	5000	1.3107
1.2248	1.38	5100	1.3097
1.1332	1.4	5200	1.3088
1.1791	1.43	5300	1.3075
1.1646	1.46	5400	1.3069
1.2292	1.49	5500	1.3062
1.2554	1.51	5600	1.3054
1.1609	1.54	5700	1.3052
1.176	1.57	5800	1.3043
1.1948	1.59	5900	1.3042
1.1441	1.62	6000	1.3036
1.1674	1.65	6100	1.3035
1.1119	1.67	6200	1.3034
1.1469	1.7	6300	1.3032
1.1834	1.73	6400	1.3033
1.0901	1.76	6500	1.3031
1.1771	1.78	6600	1.3030
1.2767	1.81	6700	1.3028
1.1978	1.84	6800	1.3027
1.1495	1.86	6900	1.3026
1.1291	1.89	7000	1.3026
1.1783	1.92	7100	1.3026
1.2189	1.94	7200	1.3025
1.3167	1.97	7300	1.3025
1.1293	2.0	7400	1.3025

Framework versions

• Transformers 4.34.0
• Pytorch 2.1.0+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1