444cc3617bd19650a6adaa03d02b547f

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Qwen-7B on the contemmcm/hate-speech-and-offensive-language dataset. It achieves the following results on the evaluation set:

• Loss: 2.0118
• Data Size: 1.0
• Epoch Runtime: 512.5120
• Accuracy: 0.9032
• F1 Macro: 0.7205

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	8.0133	0	14.1827	0.5406	0.3486
No log	1	619	5.0143	0.0078	18.0727	0.7707	0.4162
No log	2	1238	2.2748	0.0156	33.1319	0.8671	0.6128
0.0897	3	1857	1.4812	0.0312	57.3519	0.9010	0.6008
0.0897	4	2476	1.6860	0.0625	80.2774	0.8939	0.5945
1.4198	5	3095	1.2391	0.125	142.3991	0.9002	0.6580
0.1114	6	3714	1.2231	0.25	150.4439	0.9016	0.6547
1.3291	7	4333	1.3698	0.5	303.6637	0.8886	0.6830
1.0579	8.0	4952	1.9154	1.0	575.6286	0.9020	0.7367
0.7702	9.0	5571	1.2749	1.0	540.1524	0.8941	0.7070
0.5505	10.0	6190	2.0118	1.0	512.5120	0.9032	0.7205

Framework versions

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