train_rte_101112_1760638015

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

• Loss: 0.2353
• Num Input Tokens Seen: 6980984

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: 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
0.3623	1.0	561	0.2576	350480
0.2824	2.0	1122	0.2509	700992
0.2631	3.0	1683	0.2457	1050848
0.2906	4.0	2244	0.2417	1400856
0.357	5.0	2805	0.2410	1749544
0.0055	6.0	3366	0.2391	2099368
0.3289	7.0	3927	0.2367	2447504
0.2394	8.0	4488	0.2359	2794592
0.1399	9.0	5049	0.2363	3145760
0.1701	10.0	5610	0.2368	3495600
0.1945	11.0	6171	0.2353	3844488
0.1955	12.0	6732	0.2387	4191800
0.2157	13.0	7293	0.2358	4538416
0.1572	14.0	7854	0.2385	4888904
0.1489	15.0	8415	0.2358	5236560
0.2896	16.0	8976	0.2365	5587768
0.2738	17.0	9537	0.2368	5935088
0.322	18.0	10098	0.2382	6283144
0.2788	19.0	10659	0.2382	6632504
0.6244	20.0	11220	0.2382	6980984

Framework versions

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