train_copa_1757340229

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

• Loss: 6.3505
• Num Input Tokens Seen: 281408

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: 4
• eval_batch_size: 4
• seed: 456
• 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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.0899	0.5	45	0.0955	14016
0.0561	1.0	90	0.1170	28096
0.064	1.5	135	0.1002	42144
0.2623	2.0	180	0.1159	56352
0.0829	2.5	225	0.1277	70432
0.0578	3.0	270	0.1128	84544
0.0002	3.5	315	0.1150	98688
0.0039	4.0	360	0.1480	112800
0.0225	4.5	405	0.1603	126816
0.0257	5.0	450	0.2161	140800
0.0001	5.5	495	0.2599	154848
0.0015	6.0	540	0.1972	168736
0.0001	6.5	585	0.2357	182688
0.0	7.0	630	0.2384	196896
0.0	7.5	675	0.2578	211072
0.0	8.0	720	0.2484	225056
0.0	8.5	765	0.2626	239168
0.0	9.0	810	0.2575	253312
0.0	9.5	855	0.2586	267392
0.0001	10.0	900	0.2575	281408

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.8.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1