sft-trl-claim-128-llama2-13b-chat-hf

This model is a fine-tuned version of meta-llama/Llama-2-13b-chat-hf on the anli dataset. It achieves the following results on the evaluation set:

• Loss: 0.4327

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.2823	0.02	1000	1.5153
1.1442	0.03	2000	1.4907
1.1578	0.05	3000	1.4674
1.1452	0.06	4000	1.4707
1.175	0.08	5000	1.4579
1.1516	0.1	6000	1.4382
1.2068	0.11	7000	1.4134
1.2737	0.13	8000	1.3936
1.3977	0.15	9000	1.4009
1.0338	0.16	10000	1.3899
1.0997	0.18	11000	1.3971
1.1361	0.19	12000	1.3488
1.012	0.21	13000	1.3628
1.0694	0.23	14000	1.3532
1.067	0.24	15000	1.3531
1.2775	0.26	16000	1.3492
1.1953	0.28	17000	1.3598
1.0807	0.29	18000	1.3294
1.2402	0.31	19000	1.2770
1.1454	0.32	20000	1.2803
1.0369	0.34	21000	1.2688
0.9862	0.36	22000	1.2651
1.151	0.37	23000	1.2371
1.1605	0.39	24000	1.2199
1.1367	0.41	25000	1.2268
1.011	0.42	26000	1.2123
1.1792	0.44	27000	1.1763
0.8839	0.45	28000	1.1784
1.166	0.47	29000	1.1999
1.2947	0.49	30000	1.2064
0.7324	0.5	31000	1.2174
0.9919	0.52	32000	1.1837
1.3371	0.54	33000	1.1761
0.9874	0.55	34000	1.1966
1.0411	0.57	35000	1.1885
1.1669	0.58	36000	1.1789
1.231	0.6	37000	1.1616
0.9799	0.62	38000	1.1003
1.0801	0.63	39000	1.0844
1.0984	0.65	40000	1.1037
1.1046	0.67	41000	1.0683
1.1537	0.68	42000	1.0569
1.0744	0.7	43000	1.0661
1.1451	0.71	44000	1.0460
0.9822	0.73	45000	1.0559
1.0439	0.75	46000	1.0473
0.8704	0.76	47000	1.0721
0.9934	0.78	48000	0.9789
1.134	0.79	49000	0.9800
1.0722	0.81	50000	1.0122
1.0894	0.83	51000	0.9872
0.822	0.84	52000	1.0067
1.0764	0.86	53000	0.9941
1.1809	0.88	54000	0.9966
1.0228	0.89	55000	0.9657
0.9944	0.91	56000	0.9778
0.7453	0.92	57000	0.9442
1.1951	0.94	58000	0.9621
0.7633	0.96	59000	0.9271
0.5546	0.97	60000	0.9493
1.2136	0.99	61000	0.9211
0.7377	1.01	62000	0.8821
0.8892	1.02	63000	0.8635
0.9491	1.04	64000	0.8695
0.9627	1.05	65000	0.8808
0.7405	1.07	66000	0.8450
1.1451	1.09	67000	0.8496
0.9935	1.1	68000	0.8708
1.2591	1.12	69000	0.8246
1.1467	1.14	70000	0.8376
0.8296	1.15	71000	0.8411
0.9733	1.17	72000	0.8356
1.1116	1.18	73000	0.8564
0.9909	1.2	74000	0.8420
1.0602	1.22	75000	0.8398
1.0284	1.23	76000	0.8429
0.9611	1.25	77000	0.8288
0.9866	1.27	78000	0.8432
0.6751	1.28	79000	0.8109
0.9637	1.3	80000	0.8039
1.2506	1.31	81000	0.8088
1.1821	1.33	82000	0.8080
0.9813	1.35	83000	0.8003
1.175	1.36	84000	0.7962
0.8377	1.38	85000	0.7913
1.114	1.4	86000	0.7977
0.9089	1.41	87000	0.7779
0.9896	1.43	88000	0.7665
0.7499	1.44	89000	0.7693
1.0132	1.46	90000	0.7420
0.6964	1.48	91000	0.7405
0.9243	1.49	92000	0.7396
0.8555	1.51	93000	0.7448
0.9978	1.52	94000	0.7449
1.2293	1.54	95000	0.7324
0.8886	1.56	96000	0.7519
0.8325	1.57	97000	0.7625
0.5212	1.59	98000	0.7554
0.8564	1.61	99000	0.7211
0.7141	1.62	100000	0.7227
0.5702	1.64	101000	0.6725
1.0332	1.65	102000	0.6611
1.0074	1.67	103000	0.6633
0.8829	1.69	104000	0.6682
0.5767	1.7	105000	0.6641
0.6375	1.72	106000	0.6652
0.7965	1.74	107000	0.6728
1.1104	1.75	108000	0.6564
0.5071	1.77	109000	0.6273
0.9045	1.78	110000	0.6298
0.7056	1.8	111000	0.6223
1.013	1.82	112000	0.6290
0.943	1.83	113000	0.6281
0.6309	1.85	114000	0.6231
0.7388	1.87	115000	0.6069
0.5067	1.88	116000	0.5786
1.144	1.9	117000	0.5723
1.1245	1.91	118000	0.5761
0.9093	1.93	119000	0.5832
1.1755	1.95	120000	0.5798
0.4415	1.96	121000	0.5789
0.8623	1.98	122000	0.5793
0.7605	2.0	123000	0.5529
0.9743	2.01	124000	0.5542
1.0476	2.03	125000	0.5585
1.0112	2.04	126000	0.5673
0.9344	2.06	127000	0.5644
0.976	2.08	128000	0.5771
1.0471	2.09	129000	0.5615
0.626	2.11	130000	0.5597
0.6866	2.13	131000	0.5520
0.7723	2.14	132000	0.5416
0.7605	2.16	133000	0.5407
0.8413	2.17	134000	0.5452
1.1015	2.19	135000	0.5506
0.7203	2.21	136000	0.5470
0.7008	2.22	137000	0.5535
1.035	2.24	138000	0.5404
0.8432	2.25	139000	0.5459
0.7886	2.27	140000	0.5403
1.1197	2.29	141000	0.5533
0.8474	2.3	142000	0.5237
0.8785	2.32	143000	0.5325
1.1119	2.34	144000	0.5105
1.2089	2.35	145000	0.5133
0.8626	2.37	146000	0.5097
1.106	2.38	147000	0.5120
0.9681	2.4	148000	0.5116
1.0139	2.42	149000	0.5102
0.6389	2.43	150000	0.5152
0.86	2.45	151000	0.5141
1.041	2.47	152000	0.5102
0.984	2.48	153000	0.4967
0.7444	2.5	154000	0.4792
0.6543	2.51	155000	0.4693
0.6596	2.53	156000	0.4511
1.2077	2.55	157000	0.4506
0.9068	2.56	158000	0.4580
0.9328	2.58	159000	0.4602
0.8111	2.6	160000	0.4607
0.8047	2.61	161000	0.4475
0.7448	2.63	162000	0.4532
1.0685	2.64	163000	0.4628
0.9309	2.66	164000	0.4569
0.8024	2.68	165000	0.4571
0.9119	2.69	166000	0.4572
0.5736	2.71	167000	0.4567
0.7446	2.73	168000	0.4481
1.1721	2.74	169000	0.4449
0.7552	2.76	170000	0.4429
0.7927	2.77	171000	0.4358
1.2197	2.79	172000	0.4301
0.771	2.81	173000	0.4388
0.7656	2.82	174000	0.4385
1.0673	2.84	175000	0.4432
1.0228	2.86	176000	0.4361
0.7585	2.87	177000	0.4380
0.63	2.89	178000	0.4371
1.0722	2.9	179000	0.4314
0.7093	2.92	180000	0.4337
0.468	2.94	181000	0.4341
0.7841	2.95	182000	0.4313
0.779	2.97	183000	0.4312
0.4624	2.98	184000	0.4327

Framework versions

• Transformers 4.33.0.dev0
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3