train_mrpc_101112_1760638019

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the mrpc dataset. It achieves the following results on the evaluation set:

• Loss: 3.0895
• Num Input Tokens Seen: 6767120

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: 0.001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 101112
• optimizer: Use adamw_torch 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: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.1908	1.0	826	0.1935	339144
0.134	2.0	1652	0.1853	677000
0.1533	3.0	2478	0.1576	1015816
0.1368	4.0	3304	0.1423	1354072
0.1276	5.0	4130	0.1720	1692608
0.1424	6.0	4956	0.1549	2030416
0.0926	7.0	5782	0.1291	2369136
0.0879	8.0	6608	0.1318	2707920
0.117	9.0	7434	0.1414	3046392
0.0643	10.0	8260	0.1400	3383592
0.0849	11.0	9086	0.1440	3722064
0.1276	12.0	9912	0.1365	4060512
0.0126	13.0	10738	0.1848	4398224
0.0166	14.0	11564	0.1805	4737184
0.0226	15.0	12390	0.2113	5075136
0.004	16.0	13216	0.2952	5413384
0.0033	17.0	14042	0.3230	5751600
0.0018	18.0	14868	0.3796	6090384
0.0013	19.0	15694	0.3771	6429416
0.0005	20.0	16520	0.3815	6767120

Framework versions

• PEFT 0.17.1
• Transformers 4.51.3
• Pytorch 2.9.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4