a9eca047d48a8737681f51f4dcc38f67

This model is a fine-tuned version of mistralai/Mistral-7B-v0.3 on the nyu-mll/glue [stsb] dataset. It achieves the following results on the evaluation set:

• Loss: 15.5858
• Data Size: 1.0
• Epoch Runtime: 105.8312
• Mse: 3.8974
• Mae: 1.6041
• R2: -0.7434

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	Mse	Mae	R2
No log	0	0	136.2360	0	5.6261	34.0603	4.6985	-14.2364
No log	1	179	1515.5381	0.0078	6.3928	378.8858	14.2127	-168.4893
No log	2	358	963.8208	0.0156	15.2824	240.9592	15.2496	-106.7897
No log	3	537	998.2871	0.0312	25.1493	249.5697	15.5346	-110.6415
No log	4	716	13.3990	0.0625	33.6871	3.3504	1.4936	-0.4988
No log	5	895	58.9044	0.125	48.4486	14.7263	3.5349	-5.5876
371.9655	6	1074	13.5024	0.25	57.7502	3.3766	1.5416	-0.5105
14.9621	7	1253	19.1829	0.5	83.0782	4.7968	1.8101	-1.1458
13.1128	8.0	1432	10.9085	1.0	124.1367	2.7281	1.4071	-0.2204
12.9631	9.0	1611	17.0512	1.0	101.0305	4.2634	1.6677	-0.9072
22.3317	10.0	1790	16.4741	1.0	108.7810	4.1191	1.6435	-0.8426
16.1119	11.0	1969	11.5458	1.0	113.3705	2.8871	1.3845	-0.2915
12.8993	12.0	2148	10.0127	1.0	97.3646	2.5041	1.3437	-0.1202
11.6678	13.0	2327	9.3131	1.0	111.2048	2.3290	1.2823	-0.0419
10.9756	14.0	2506	10.5268	1.0	117.7345	2.6324	1.3309	-0.1775
26.2864	15.0	2685	30.2155	1.0	112.5169	7.5542	2.3592	-2.3793
8.9484	16.0	2864	10.7455	1.0	96.4602	2.6871	1.3376	-0.2020
7.4553	17.0	3043	15.5858	1.0	105.8312	3.8974	1.6041	-0.7434

Framework versions

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