llama3_ft_section_classifier

This model is a fine-tuned version of meta-llama/Llama-3.1-8B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.3342
• Accuracy: 0.6232
• Precision: 0.6126
• Recall: 0.6232
• F1: 0.6164

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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	Accuracy	Precision	Recall	F1
16.483	1.0	275	1.3758	0.5423	0.5769	0.5423	0.5311
9.7264	2.0	550	1.1577	0.6095	0.6215	0.6095	0.6065
8.2372	3.0	825	1.1713	0.6041	0.6264	0.6041	0.6061
6.1069	4.0	1100	1.2993	0.6123	0.6090	0.6123	0.6025
3.1467	5.0	1375	1.5804	0.6027	0.6255	0.6027	0.6085
1.3995	6.0	1650	1.9973	0.6077	0.6005	0.6077	0.5994
0.8489	7.0	1925	2.3380	0.6082	0.6070	0.6082	0.5990
0.4705	8.0	2200	2.5919	0.6245	0.6223	0.6245	0.6172
0.186	9.0	2475	2.8240	0.6223	0.6275	0.6223	0.6238
0.0636	10.0	2750	3.0796	0.6209	0.6273	0.6209	0.6190
0.0248	11.0	3025	3.2076	0.6259	0.6269	0.6259	0.6231
0.0009	12.0	3300	3.2148	0.6214	0.6133	0.6214	0.6158
0.0001	13.0	3575	3.2700	0.6209	0.6132	0.6209	0.6158
0.0	14.0	3850	3.2962	0.6223	0.6124	0.6223	0.6158
0.0	15.0	4125	3.3102	0.6223	0.6118	0.6223	0.6156
0.0	16.0	4400	3.3219	0.6236	0.6138	0.6236	0.6173
0.0	17.0	4675	3.3271	0.6232	0.6125	0.6232	0.6162
0.0	18.0	4950	3.3285	0.6218	0.6108	0.6218	0.6148
0.0	19.0	5225	3.3359	0.6232	0.6126	0.6232	0.6163
0.0	20.0	5500	3.3342	0.6232	0.6126	0.6232	0.6164

Framework versions

• PEFT 0.17.1
• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1