tsavage68/Error_Q1.5_1000steps_1e7rate_SFT

Error_Q1.5_1000steps_1e8rate_SFT

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.2612

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: 1e-07
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Use adafactor and the args are: No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss
3.2879	0.8	50	3.3190
3.2751	1.592	100	3.3141
3.2669	2.384	150	3.3053
3.2545	3.176	200	3.2985
3.2725	3.976	250	3.2915
3.3456	4.768	300	3.2849
3.238	5.5600	350	3.2793
3.2658	6.352	400	3.2742
3.2133	7.144	450	3.2706
3.2263	7.944	500	3.2674
3.1874	8.736	550	3.2654
3.2894	9.528	600	3.2640
3.1922	10.32	650	3.2625
3.2137	11.112	700	3.2617
3.2273	11.912	750	3.2612
3.2798	12.704	800	3.2613
3.2424	13.496	850	3.2615
3.2125	14.288	900	3.2613
3.1629	15.08	950	3.2612
3.2557	15.88	1000	3.2612

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0