af307f0a4caf0f087e68f24a6febfea4

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B on the contemmcm/hate-speech-and-offensive-language dataset. It achieves the following results on the evaluation set:

• Loss: 4.0290
• Data Size: 1.0
• Epoch Runtime: 148.7105
• Accuracy: 0.8991
• F1 Macro: 0.6590

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	11.7931	0	11.1951	0.1554	0.1496
No log	1	619	3.5957	0.0078	11.6450	0.6938	0.3978
No log	2	1238	1.9080	0.0156	15.5521	0.8243	0.5569
0.067	3	1857	1.4493	0.0312	20.0324	0.8811	0.6044
0.067	4	2476	1.2222	0.0625	26.1814	0.8967	0.6044
1.2052	5	3095	1.3468	0.125	38.5015	0.8864	0.6561
0.1021	6	3714	1.1098	0.25	51.2879	0.9099	0.6643
0.9854	7	4333	1.0602	0.5	81.1479	0.9022	0.7655
0.8238	8.0	4952	1.0236	1.0	149.1516	0.9131	0.7471
0.3569	9.0	5571	2.2327	1.0	148.7864	0.9065	0.7131
0.2328	10.0	6190	2.9484	1.0	147.9016	0.9058	0.7202
0.2036	11.0	6809	2.9696	1.0	146.7453	0.9109	0.7085
0.0343	12.0	7428	4.0290	1.0	148.7105	0.8991	0.6590

Framework versions

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