Gnomy19/llama3_new_valearlystop

matching

This model is a fine-tuned version of meta-llama/Llama-3.1-8B-Instruct on the powerseeking_matching_new dataset. It achieves the following results on the evaluation set:

• Loss: 2.0703

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• 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_ratio: 0.1
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.8	1	2.9815
No log	1.0	2	2.9815
No log	1.8	3	2.9766
No log	2.0	4	2.9766
No log	2.8	5	2.9584
No log	3.0	6	2.9584
No log	3.8	7	2.9074
No log	4.0	8	2.7929
No log	4.8	9	2.6056
3.5356	5.0	10	2.6056
3.5356	5.8	11	2.3322
3.5356	6.0	12	2.3322
3.5356	6.8	13	2.1817
3.5356	7.0	14	2.1817
3.5356	7.8	15	2.1220
3.5356	8.0	16	2.0906
3.5356	8.8	17	2.0737
3.5356	9.0	18	2.0737
3.5356	9.8	19	2.0499
2.3974	10.0	20	2.0499
2.3974	10.8	21	2.0344
2.3974	11.0	22	2.0344
2.3974	11.8	23	2.0299
2.3974	12.0	24	2.0332
2.3974	12.8	25	2.0703
2.3974	13.0	26	2.0703

Framework versions

• PEFT 0.15.1
• Transformers 4.51.1
• Pytorch 2.7.1+cu126
• Datasets 3.5.0
• Tokenizers 0.21.0