5dc2d7eacf9642bb6f1c25f7d3070741

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Llama-8B on the nyu-mll/glue [qnli] dataset. It achieves the following results on the evaluation set:

• Loss: 2.9198
• Data Size: 1.0
• Epoch Runtime: 4515.5775
• Accuracy: 0.6053
• F1 Macro: 0.6040
• Rouge1: 0.6053
• Rouge2: 0.0
• Rougel: 0.6052
• Rougelsum: 0.6053

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	Rouge1	Rouge2	Rougel	Rougelsum
No log	0	0	7.8164	0	20.4095	0.5129	0.4335	0.5129	0.0	0.5129	0.5129
No log	1	3273	2.9345	0.0078	54.6449	0.4943	0.3308	0.4947	0.0	0.4943	0.4944
0.1937	2	6546	3.1309	0.0156	97.7393	0.5057	0.3359	0.5053	0.0	0.5057	0.5056
2.971	3	9819	12.9446	0.0312	177.1149	0.4943	0.3308	0.4947	0.0	0.4943	0.4944
3.229	4	13092	3.0043	0.0625	313.0907	0.5035	0.3736	0.5040	0.0	0.5040	0.5037
2.8262	5	16365	2.7588	0.125	595.0140	0.5368	0.4604	0.5368	0.0	0.5368	0.5368
3.0052	6	19638	2.6721	0.25	1159.9461	0.5949	0.5867	0.5950	0.0	0.5947	0.5949
2.677	7	22911	2.6561	0.5	2279.4342	0.6081	0.6070	0.6079	0.0	0.6086	0.6079
2.6526	8.0	26184	2.7120	1.0	4533.3424	0.5938	0.5698	0.5938	0.0	0.5941	0.5938
2.4201	9.0	29457	2.7097	1.0	4536.1910	0.6101	0.6040	0.6101	0.0	0.6101	0.6102
2.1256	10.0	32730	2.6998	1.0	4538.2277	0.6202	0.6201	0.6202	0.0	0.6204	0.6200
1.9786	11.0	36003	2.9198	1.0	4515.5775	0.6053	0.6040	0.6053	0.0	0.6052	0.6053

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.1