File tree 1 file changed +0
-8
lines changed
example/auto_compression/nlp
1 file changed +0
-8
lines changed Original file line number Diff line number Diff line change 30
30
| ERNIE 3.0-Medium | Base模型| 75.35 | 57.45 | 60.17 | 81.16 | 77.19 | 80.59 | 79.70 | 73.09 |
31
31
| ERNIE 3.0-Medium | 剪枝+量化训练| 74.17 | 56.84 | 59.75 | 80.54 | 76.03 | 76.97 | 80.80 | 72.16 |
32
32
33
- | 模型 | 策略 | 报销工单数据 |
34
- | :------:| :------:| :------:|
35
- | UIE-base | Base模型 | [ 91.83] ( https://bj.bcebos.com/v1/paddle-slim-models/act/uie_base.tar ) |
36
- | UIE-base | 量化训练 | [ 95.80] ( https://bj.bcebos.com/v1/paddle-slim-models/act/uie_base_qat_model.tar ) |
37
-
38
- 注:UIE模型精度为在5-shot(每个类别包含5条标注数据)数据集上进行模型微调的结果,压缩后精度更高可能原因是过拟合在当前数据集。
39
33
40
34
模型在不同任务上平均精度以及加速对比如下:
41
35
| 模型 | 策略| Accuracy(avg) | 预测时延<sup ><small >FP32</small ><sup ><br ><sup > | 预测时延<sup ><small >FP16</small ><sup ><br ><sup > | 预测时延<sup ><small >INT8</small ><sup ><br ><sup > | 加速比 |
44
38
| PP-MiniLM| 剪枝+离线量化 | 71.85 | - | - | 15.76ms | 5.99x |
45
39
| ERNIE 3.0-Medium| Base模型| 73.09 | 89.71ms | 20.76ms | - | - |
46
40
| ERNIE 3.0-Medium| 剪枝+量化训练 | 72.16 | - | - | 14.08ms | 6.37x |
47
- | UIE-base| Base模型| 91.83 | 42.66ms | 14.23ms | - | - |
48
- | UIE-base| 量化训练 | 95.80 | - | - | 10.94ms | 3.90x |
49
41
50
42
性能测试的环境为
51
43
- 硬件:NVIDIA Tesla T4 单卡
You can’t perform that action at this time.
0 commit comments