train_qnli_789_1760637976

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

• Loss: 0.0396
• Num Input Tokens Seen: 207033472

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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.0218	1.0	23567	0.0441	10356768
0.0026	2.0	47134	0.0396	20704704
0.101	3.0	70701	0.0454	31058016
0.0019	4.0	94268	0.0616	41403776
0.0023	5.0	117835	0.0662	51759136
0.0264	6.0	141402	0.0871	62112992
0.0	7.0	164969	0.1006	72463360
0.0	8.0	188536	0.1255	82820352
0.0	9.0	212103	0.1005	93170016
0.0	10.0	235670	0.1357	103520672
0.0	11.0	259237	0.1157	113880864
0.0017	12.0	282804	0.1319	124232512
0.0	13.0	306371	0.1159	134580160
0.0	14.0	329938	0.1946	144930784
0.0	15.0	353505	0.2255	155287840
0.0	16.0	377072	0.2296	165642176
0.0	17.0	400639	0.2640	175984352
0.0	18.0	424206	0.2674	186335840
0.0	19.0	447773	0.2688	196686912
0.0	20.0	471340	0.2683	207033472

Framework versions

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