qa-finetune

This model is a fine-tuned version of stage-babylm/llama-256-12L on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.9113

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.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_steps: 0.1
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.6867	0.1901	100	2.7004
2.1401	0.3802	200	2.1452
2.0333	0.5703	300	2.0345
1.9394	0.7605	400	1.9897
1.9388	0.9506	500	1.9645
1.9553	1.1407	600	1.9493
1.9092	1.3308	700	1.9371
1.8855	1.5209	800	1.9292
1.8996	1.7110	900	1.9226
1.9202	1.9011	1000	1.9182
1.8683	2.0913	1100	1.9149
1.8255	2.2814	1200	1.9132
1.8680	2.4715	1300	1.9121
1.8863	2.6616	1400	1.9115
1.8532	2.8517	1500	1.9113
1.8478	3.0	1578	1.9113

Framework versions

• Transformers 5.14.1
• Pytorch 2.13.0+cu130
• Datasets 5.0.0
• Tokenizers 0.22.2