Modelpublic

baptiste-pasquier/distilcamembert-allocine

sourceHugging Facemitupdated 4y agoView on Hugging Face

0likes10downloads

Model Card

distilcamembert-allocine

This model is a fine-tuned version of cmarkea/distilcamembert-base on the allocine dataset. It achieves the following results on the evaluation set:

—Loss: 0.1066
—Accuracy: 0.9714
—F1: 0.9710
—Precision: 0.9648
—Recall: 0.9772

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
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—gradientaccumulationsteps: 4
—totaltrainbatch_size: 64
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 500
—num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.1504	0.2	500	0.1290	0.9555	0.9542	0.9614	0.9470
0.1334	0.4	1000	0.1049	0.9624	0.9619	0.9536	0.9703
0.1158	0.6	1500	0.1052	0.963	0.9627	0.9498	0.9760
0.1153	0.8	2000	0.0949	0.9661	0.9653	0.9686	0.9620
0.1053	1.0	2500	0.0936	0.9666	0.9663	0.9542	0.9788
0.0755	1.2	3000	0.0987	0.97	0.9695	0.9644	0.9748
0.0716	1.4	3500	0.1078	0.9688	0.9684	0.9598	0.9772
0.0688	1.6	4000	0.1051	0.9673	0.9670	0.9552	0.9792
0.0691	1.8	4500	0.0940	0.9709	0.9704	0.9688	0.9720
0.0733	2.0	5000	0.1038	0.9686	0.9683	0.9558	0.9812
0.0476	2.2	5500	0.1066	0.9714	0.9710	0.9648	0.9772
0.047	2.4	6000	0.1098	0.9689	0.9686	0.9587	0.9788
0.0431	2.6	6500	0.1110	0.9711	0.9706	0.9666	0.9747
0.0464	2.8	7000	0.1149	0.9697	0.9694	0.9592	0.9798
0.0342	3.0	7500	0.1122	0.9703	0.9699	0.9621	0.9778

Framework versions

—Transformers 4.26.1
—Pytorch 1.13.1+cu117
—Datasets 2.10.1
—Tokenizers 0.13.2