odrl2text-llama32-3b-lora-v1

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the odrl_to_text_train dataset. It achieves the following results on the evaluation set:

• Loss: 0.0973

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: 0.00015
• train_batch_size: 6
• eval_batch_size: 3
• seed: 42
• gradient_accumulation_steps: 3
• total_train_batch_size: 18
• 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: 0.05
• num_epochs: 4.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.3448	0.2423	50	0.3231
0.2736	0.4847	100	0.2603
0.2656	0.7270	150	0.2355
0.2486	0.9693	200	0.2155
0.2204	1.2084	250	0.2006
0.2165	1.4507	300	0.1872
0.1989	1.6931	350	0.1765
0.1955	1.9354	400	0.1635
0.1513	2.1745	450	0.1477
0.1623	2.4168	500	0.1369
0.146	2.6591	550	0.1256
0.1543	2.9015	600	0.1169
0.1047	3.1405	650	0.1062
0.1081	3.3829	700	0.1011
0.1083	3.6252	750	0.0982
0.1122	3.8675	800	0.0973

Framework versions

• PEFT 0.15.2
• Transformers 4.50.1
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1