512duncanl/gec-flan-t5-xxl-stage-2

gec-flan-t5-large-stage-2-v3

This model is a fine-tuned version of 512duncanl/gec-flan-t5-xxl-stage-1 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1747
• F0.5: 0.6727
• Precision: 0.6968
• Recall: 0.5908

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Trained on 4x H100 SXM 80G

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-06
• train_batch_size: 18
• eval_batch_size: 18
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 1
• total_train_batch_size: 72
• 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: constant_with_warmup
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 2
• weight_decay: 0.005

Training results

Training Loss	Epoch	Step	Validation Loss	F0.5	Precision	Recall
0.2030	1.0	514	0.1900	0.6723	0.7071	0.5619
0.1874	2.0	1028	0.1747	0.6727	0.6968	0.5908

Framework versions

• Transformers 4.54.0
• Pytorch 2.7.1+cu128
• Datasets 4.0.0
• Tokenizers 0.21.2