train_siqa_123_1760637717

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: 0.1990
• 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: 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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.0781	1.0	7518	0.1990	3014896
0.3435	2.0	15036	0.2055	6029360
0.0601	3.0	22554	0.2259	9042368
0.1368	4.0	30072	0.3210	12055456
0.0006	5.0	37590	0.4278	15068512
0.0004	6.0	45108	0.4634	18081672
0.0023	7.0	52626	0.5658	21095960
0.0001	8.0	60144	0.5224	24109392
0.0003	9.0	67662	0.6131	27122856
0.0	10.0	75180	0.6819	30137256
0.0	11.0	82698	0.7248	33151024
0.0	12.0	90216	0.7157	36165000
0.0	13.0	97734	0.8155	39178496
0.0004	14.0	105252	0.7988	42193184
0.0	15.0	112770	0.9158	45206576
0.0	16.0	120288	0.8764	48220600
0.0	17.0	127806	0.9802	51235344
0.0	18.0	135324	1.0216	54249648
0.0	19.0	142842	1.0404	57262600
0.0	20.0	150360	1.0431	60276872

Framework versions

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