1bd633844d99a3ff4980473579474e6d

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Llama-8B on the contemmcm/hate-speech-and-offensive-language dataset. It achieves the following results on the evaluation set:

• Loss: 4.4375
• Data Size: 1.0
• Epoch Runtime: 800.2166
• Accuracy: 0.8825
• F1 Macro: 0.6882

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	8.9166	0	15.4189	0.3261	0.2735
No log	1	619	28.1082	0.0078	19.6130	0.0601	0.0378
No log	2	1238	2.7948	0.0156	37.7784	0.7672	0.2894
0.2017	3	1857	3.0229	0.0312	59.1874	0.7672	0.2894
0.2017	4	2476	2.9114	0.0625	89.3952	0.7672	0.2894
1.986	5	3095	1.9642	0.125	135.7084	0.8480	0.5519
0.1353	6	3714	1.3196	0.25	215.9081	0.8935	0.5939
1.2389	7	4333	1.2140	0.5	386.4150	0.8935	0.6527
1.0219	8.0	4952	1.3450	1.0	768.0825	0.8929	0.5940
0.6889	9.0	5571	1.7026	1.0	779.2639	0.8882	0.7290
0.4635	10.0	6190	2.8859	1.0	803.6598	0.8837	0.6784
0.2785	11.0	6809	4.4375	1.0	800.2166	0.8825	0.6882

Framework versions

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