train_cola_1757340187

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

• Loss: 0.1613
• Num Input Tokens Seen: 3669168

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
0.2087	0.5	962	0.1798	184192
0.2123	1.0	1924	0.1613	367320
0.2773	1.5	2886	0.1658	550840
0.2206	2.0	3848	0.1682	734600
0.1392	2.5	4810	0.1733	918600
0.048	3.0	5772	0.1635	1101216
0.0032	3.5	6734	0.2115	1284288
0.003	4.0	7696	0.1948	1468552
0.0006	4.5	8658	0.2420	1651528
0.0007	5.0	9620	0.2921	1834816
0.0006	5.5	10582	0.3214	2018016
0.0016	6.0	11544	0.2397	2201584
0.0	6.5	12506	0.4304	2385200
0.0001	7.0	13468	0.3517	2568288
0.0	7.5	14430	0.4275	2751584
0.0	8.0	15392	0.4333	2935056
0.0001	8.5	16354	0.4750	3118000
0.0	9.0	17316	0.4611	3301760
0.0	9.5	18278	0.4809	3485344
0.0	10.0	19240	0.4817	3669168

Framework versions

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