grammar-classifier

This model is a fine-tuned version of microsoft/deberta-v3-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 4.0967
• Exact Match: 0.0
• Micro F1: 0.3075
• Macro F1: 0.0334
• Hamming Accuracy: 0.8806
• Avg Pred Positives: 34.0
• Avg Gold Positives: 13.5736

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: 32
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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_steps: 0.2
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Exact Match	Micro F1	Macro F1	Hamming Accuracy	Avg Pred Positives	Avg Gold Positives
0.2930	0.5	164	0.5663	0.0	0.3139	0.0288	0.9041	25.0	13.5736
0.1928	1.0	328	0.2720	0.0	0.4392	0.0256	0.9460	13.0	13.5736
0.0751	1.5	492	0.0559	0.0	0.5244	0.0234	0.9628	8.0	13.5736
33.3599	2.0	656	13.5265	0.0	0.2931	0.0226	0.9114	21.0	13.5736
19.0859	2.5	820	10.5864	0.0	0.2214	0.0302	0.8376	44.0	13.5736
8.6145	3.0	984	4.0967	0.0	0.3075	0.0334	0.8806	34.0	13.5736

Framework versions

• Transformers 5.2.0
• Pytorch 2.10.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2