train_copa_1757340256

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: 0.0623
• Num Input Tokens Seen: 281984

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: 789
• 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
0.5993	0.5	45	0.5326	14240
0.6312	1.0	90	0.3705	28192
0.0635	1.5	135	0.0711	42080
0.1393	2.0	180	0.0634	56192
0.0085	2.5	225	0.0633	70048
0.1219	3.0	270	0.0656	84192
0.1509	3.5	315	0.0654	98304
0.0373	4.0	360	0.0660	112544
0.0535	4.5	405	0.0668	126784
0.1165	5.0	450	0.0628	140960
0.0766	5.5	495	0.0654	155200
0.0151	6.0	540	0.0652	169216
0.0561	6.5	585	0.0626	183232
0.0085	7.0	630	0.0627	197248
0.0554	7.5	675	0.0626	211424
0.1764	8.0	720	0.0635	225440
0.0044	8.5	765	0.0623	239392
0.0226	9.0	810	0.0626	253632
0.0474	9.5	855	0.0633	267680
0.036	10.0	900	0.0627	281984

Framework versions

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