eec5bf97a3b24ee0cf1e23b7151a1722

This model is a fine-tuned version of facebook/opt-1.3b on the contemmcm/hate-speech-and-offensive-language dataset. It achieves the following results on the evaluation set:

• Loss: 0.6168
• Data Size: 1.0
• Epoch Runtime: 105.7655
• Accuracy: 0.8803
• F1 Macro: 0.7052

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	1.4877	0	7.8312	0.2494	0.2102
No log	1	619	0.6815	0.0078	8.4511	0.7794	0.3624
No log	2	1238	0.3794	0.0156	11.2665	0.8858	0.7105
0.0149	3	1857	0.4267	0.0312	14.8523	0.8969	0.6017
0.0149	4	2476	0.3067	0.0625	19.3194	0.9006	0.6050
0.3464	5	3095	0.3413	0.125	27.2625	0.8933	0.7276
0.0322	6	3714	0.3296	0.25	40.3434	0.9022	0.6029
0.3163	7	4333	0.2983	0.5	60.9218	0.8963	0.7227
0.2705	8.0	4952	0.2739	1.0	106.0191	0.9079	0.6461
0.1939	9.0	5571	0.3485	1.0	102.0337	0.8904	0.7369
0.1362	10.0	6190	0.3832	1.0	101.5772	0.9048	0.6928
0.0874	11.0	6809	0.5401	1.0	102.8107	0.9004	0.7170
0.0351	12.0	7428	0.6168	1.0	105.7655	0.8803	0.7052

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1