train_copa_1754651789

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: 0.2669
• Num Input Tokens Seen: 281856

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: 123
• 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
12.1391	0.5	45	11.9483	14016
8.1827	1.0	90	7.8277	28096
4.4968	1.5	135	4.2820	42144
2.6252	2.0	180	2.3994	56128
1.0636	2.5	225	0.9134	70272
0.42	3.0	270	0.4435	84352
0.3501	3.5	315	0.3621	98464
0.401	4.0	360	0.3243	112576
0.3488	4.5	405	0.3295	126624
0.3229	5.0	450	0.2785	140832
0.257	5.5	495	0.2820	154976
0.245	6.0	540	0.2875	169056
0.2576	6.5	585	0.2669	183200
0.2504	7.0	630	0.2740	197344
0.267	7.5	675	0.2765	211392
0.275	8.0	720	0.2757	225536
0.2552	8.5	765	0.2773	239680
0.2574	9.0	810	0.2751	253696
0.2546	9.5	855	0.2736	267840
0.2711	10.0	900	0.2693	281856

Framework versions

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