metadata

language:
  - en
license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - glue
metrics:
  - accuracy
base_model: bert-base-uncased
model-index:
  - name: jpqd-bert-base-ft-sst2
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: GLUE SST2
          type: glue
          config: sst2
          split: validation
          args: sst2
        metrics:
          - type: accuracy
            value: 0.9162844036697247
            name: Accuracy

jpqd-bert-base-ft-sst2

This model is a fine-tuned version of bert-base-uncased on the GLUE SST2 dataset.

It was compressed with NNCF following the Optimum JPQD text-classification example

It achieves the following results on the evaluation set:

Loss: 0.2798
Accuracy: 0.9163

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 32
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 5.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.392	0.12	250	0.4535	0.8888
0.4413	0.24	500	0.4671	0.8899
0.29	0.36	750	0.3285	0.9128
0.2851	0.48	1000	0.2498	0.9151
0.3717	0.59	1250	0.2037	0.9243
0.2467	0.71	1500	0.2840	0.9174
0.2114	0.83	1750	0.2239	0.9243
0.1777	0.95	2000	0.1968	0.9266
2.6501	1.07	2250	2.8219	0.9255
6.4768	1.19	2500	6.5765	0.8979
9.3594	1.31	2750	9.4648	0.8819
11.5481	1.43	3000	11.5391	0.8567
12.7541	1.54	3250	12.8359	0.8578
13.6184	1.66	3500	13.6519	0.8429
13.9171	1.78	3750	14.0734	0.8475
13.9601	1.9	4000	14.1024	0.8578
0.2701	2.02	4250	0.3354	0.9048
0.2689	2.14	4500	0.3320	0.9048
0.1775	2.26	4750	0.2838	0.9163
0.1648	2.38	5000	0.2842	0.9128
0.1316	2.49	5250	0.2750	0.9163
0.2349	2.61	5500	0.2405	0.9232
0.066	2.73	5750	0.2695	0.9174
0.1285	2.85	6000	0.3017	0.9094
0.1813	2.97	6250	0.3472	0.9106
0.078	3.09	6500	0.2915	0.9140
0.0886	3.21	6750	0.2853	0.9151
0.117	3.33	7000	0.2689	0.9186
0.0894	3.44	7250	0.2748	0.9174
0.1023	3.56	7500	0.3279	0.9094
0.0495	3.68	7750	0.2988	0.9151
0.0899	3.8	8000	0.2796	0.9174
0.1102	3.92	8250	0.2667	0.9163
0.061	4.04	8500	0.2837	0.9174
0.0594	4.16	8750	0.2766	0.9151
0.1062	4.28	9000	0.2777	0.9140
0.0751	4.39	9250	0.2690	0.9220
0.0386	4.51	9500	0.2668	0.9163
0.0284	4.63	9750	0.2812	0.9186
0.1016	4.75	10000	0.2825	0.9163
0.0507	4.87	10250	0.2805	0.9140
0.0709	4.99	10500	0.2855	0.9140

Framework versions

Transformers 4.26.1
Pytorch 1.13.1+cu117
Datasets 2.8.0
Tokenizers 0.13.2