cshulby/phi_1_5_KelmTiny

phi-kelm-out

This model is a fine-tuned version of typeof/phi-1_5 on the Kelm Tiny dataset. It achieves the following results on the evaluation set:

• Loss: 1.0079

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: 3e-06
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 1000
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss
7.8236	0.0	1	5.4714
4.156	0.1	995	4.0834
1.9418	0.2	1990	2.8447
1.8908	0.3	2985	2.2757
0.7631	0.4	3980	1.8792
1.0878	0.5	4975	1.4944
2.1561	0.6	5970	1.3413
0.452	0.7	6965	1.2682
2.1017	0.8	7960	1.2247
0.8352	0.9	8955	1.1999
5.1122	1.0	9950	1.1778
1.6136	1.1	10945	1.1515
2.3537	1.2	11940	1.1364
0.2987	1.3	12935	1.1391
0.747	1.4	13930	1.0977
0.0025	1.5	14925	1.0917
0.6355	1.6	15920	1.0630
0.5881	1.7	16915	1.0565
0.3181	1.8	17910	1.0568
0.9256	1.9	18905	1.0623
4.5318	2.0	19900	1.0678
0.8736	2.1	20895	1.0645
2.2079	2.2	21890	1.0474
2.7407	2.3	22885	1.0438
2.2308	2.4	23880	1.0485
0.4307	2.5	24875	1.0235
0.2956	2.6	25870	1.0201
0.203	2.7	26865	1.0200
2.2452	2.8	27860	1.0243
0.942	2.9	28855	1.0289
0.0069	3.0	29850	1.0181
3.2121	3.1	30845	1.0235
1.4533	3.2	31840	1.0127
0.208	3.3	32835	1.0110
0.1379	3.4	33830	1.0126
0.1991	3.5	34825	1.0103
1.3019	3.6	35820	1.0154
0.6602	3.7	36815	1.0178
0.5271	3.8	37810	1.0087
0.3131	3.9	38805	1.0092
3.6821	4.0	39800	1.0094
0.3724	4.1	40795	1.0093
0.0704	4.2	41790	1.0081
0.1209	4.3	42785	1.0108
0.9807	4.4	43780	1.0072
0.1392	4.5	44775	1.0078
0.2561	4.6	45770	1.0078
0.1533	4.7	46765	1.0089
0.4302	4.8	47760	1.0079
1.3744	4.9	48755	1.0074
0.8572	5.0	49750	1.0079

Framework versions

• Transformers 4.35.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1