train_siqa_123_1760637716

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: 3.3358
• Num Input Tokens Seen: 60276872

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: 0.001
• 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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.5495	1.0	7518	0.5491	3014896
0.6116	2.0	15036	0.5055	6029360
0.4169	3.0	22554	0.4472	9042368
0.3861	4.0	30072	0.4347	12055456
0.3111	5.0	37590	0.4150	15068512
0.4058	6.0	45108	0.4059	18081672
0.3678	7.0	52626	0.3917	21095960
0.373	8.0	60144	0.3736	24109392
0.3179	9.0	67662	0.3401	27122856
0.5199	10.0	75180	0.3146	30137256
0.2575	11.0	82698	0.2851	33151024
0.1939	12.0	90216	0.2463	36165000
0.2458	13.0	97734	0.2304	39178496
0.1725	14.0	105252	0.2346	42193184
0.1974	15.0	112770	0.2216	45206576
0.1114	16.0	120288	0.2221	48220600
0.111	17.0	127806	0.2240	51235344
0.1123	18.0	135324	0.2244	54249648
0.096	19.0	142842	0.2260	57262600
0.4597	20.0	150360	0.2274	60276872

Framework versions

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