train_svamp_1757340247

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

• Loss: 1.7457
• Num Input Tokens Seen: 704320

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
0.276	0.5	79	0.4642	35392
0.0961	1.0	158	0.1574	70288
0.1096	1.5	237	0.1084	105936
0.0438	2.0	316	0.1091	140896
0.1061	2.5	395	0.1103	175840
0.059	3.0	474	0.0910	211504
0.0323	3.5	553	0.0918	246864
0.0599	4.0	632	0.0918	281664
0.0266	4.5	711	0.0962	317152
0.0806	5.0	790	0.0932	352048
0.0639	5.5	869	0.0987	387600
0.0374	6.0	948	0.1032	422400
0.0089	6.5	1027	0.1115	457792
0.0324	7.0	1106	0.1179	492720
0.0129	7.5	1185	0.1175	528336
0.0105	8.0	1264	0.1227	563312
0.0146	8.5	1343	0.1281	598800
0.0032	9.0	1422	0.1280	633968
0.0006	9.5	1501	0.1279	669456
0.0004	10.0	1580	0.1305	704320

Framework versions

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