Modelpublic

thrunlab/Mistral_Sparse_refined_web_50p_2024-02-16

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face

0likes26downloads

Model Card

MistralSparserefinedweb50p_2024-02-16

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:

—Loss: 2.1260

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-05
—trainbatchsize: 1
—evalbatchsize: 1
—seed: 0
—distributed_type: multi-GPU
—num_devices: 3
—gradientaccumulationsteps: 3
—totaltrainbatch_size: 9
—totalevalbatch_size: 3
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss
2.5975	0.01	25	2.6362
2.3082	0.01	50	2.5659
2.4024	0.02	75	2.5151
2.3358	0.02	100	2.4817
2.2267	0.03	125	2.4660
2.271	0.04	150	2.4456
2.1709	0.04	175	2.4413
2.2549	0.05	200	2.4306
2.2536	0.05	225	2.4243
2.2234	0.06	250	2.4212
2.2516	0.07	275	2.4202
2.2827	0.07	300	2.4146
2.1774	0.08	325	2.4156
2.278	0.08	350	2.4094
2.204	0.09	375	2.4088
2.1987	0.1	400	2.4073
2.1985	0.1	425	2.4041
2.2198	0.11	450	2.4069
2.2555	0.11	475	2.4014
2.1567	0.12	500	2.4017
2.2918	0.13	525	2.3998
2.2559	0.13	550	2.3959
2.2234	0.14	575	2.3978
2.2001	0.14	600	2.3944
2.1409	0.15	625	2.3957
2.2034	0.16	650	2.3981
2.1863	0.16	675	2.3941
2.2372	0.17	700	2.3936
2.2438	0.17	725	2.3953
2.2172	0.18	750	2.3943
2.1917	0.19	775	2.3921
2.1137	0.19	800	2.3912
2.0766	0.07	825	2.3935
2.1926	0.08	850	2.3913
2.2948	0.08	875	2.3915
2.1349	0.08	900	2.3917
2.2446	0.08	925	2.3876
2.253	0.09	950	2.3880
2.0729	0.09	975	2.3890
2.1965	0.09	1000	2.3873

Framework versions

—Transformers 4.36.2
—Pytorch 2.1.2+cu121
—Datasets 2.15.0
—Tokenizers 0.15.0