train_siqa_42_1760637601

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

• Loss: 0.1848
• Num Input Tokens Seen: 60302568

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: 42
• 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
0.2173	1.0	7518	0.1936	3016248
0.1537	2.0	15036	0.1848	6032368
0.0881	3.0	22554	0.1877	9049000
0.2016	4.0	30072	0.3158	12063104
0.0187	5.0	37590	0.3972	15078392
0.0003	6.0	45108	0.4899	18094200
0.0734	7.0	52626	0.5379	21109936
0.0016	8.0	60144	0.4953	24124456
0.0007	9.0	67662	0.5203	27139488
0.0	10.0	75180	0.7133	30155824
0.1875	11.0	82698	0.7045	33169800
0.0	12.0	90216	0.7003	36184296
0.0	13.0	97734	0.7969	39199224
0.0	14.0	105252	0.8078	42213984
0.0	15.0	112770	0.7945	45227616
0.0	16.0	120288	0.9663	48242336
0.0	17.0	127806	0.9757	51258152
0.0	18.0	135324	1.0241	54272896
0.0	19.0	142842	1.0365	57288368
0.0	20.0	150360	1.0399	60302568

Framework versions

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