train_svamp_1757340200

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: 0.0852
• Num Input Tokens Seen: 705184

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
2.1027	0.5	79	1.9531	35776
1.2394	1.0	158	1.1011	70672
0.2401	1.5	237	0.2649	105904
0.1372	2.0	316	0.1357	141328
0.133	2.5	395	0.1165	176752
0.073	3.0	474	0.1059	211808
0.078	3.5	553	0.1010	247104
0.1014	4.0	632	0.0988	282048
0.0548	4.5	711	0.0961	317248
0.1134	5.0	790	0.0918	352592
0.091	5.5	869	0.0886	388176
0.117	6.0	948	0.0884	423184
0.1104	6.5	1027	0.0861	458640
0.076	7.0	1106	0.0867	493440
0.1771	7.5	1185	0.0859	528768
0.0453	8.0	1264	0.0856	563872
0.027	8.5	1343	0.0859	599232
0.1205	9.0	1422	0.0855	634544
0.0927	9.5	1501	0.0852	670064
0.0131	10.0	1580	0.0855	705184

Framework versions

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