【Hackathon 4th No.102】给AutoConverter增加新的模型组网的支持 Distilbert

[megemini]

PR types

New features

PR changes

APIs

Description

【Hackathon 4th No.102】给AutoConverter增加新的模型组网的支持 Distilbert

Hackathon 4th No.102 这个任务里面有5个模型，我计划每个模型单独提PR，这个PR是处理 distilbert 模型。

• 使用的测试模型 hf-internal-testing/tiny-random-DistilBertModel
• 这个模型 paddlenlp 与 transformers 的参数有几个不一样，我在 configuration.py 中进行了映射：

    attribute_map: Dict[str, str] = {
        "dropout": "classifier_dropout",
        "num_classes": "num_labels",
        "n_layers": "num_hidden_layers",  # for `transformers`
        "n_heads": "num_attention_heads",  # for `transformers`
        "dim": "hidden_size",  # for `transformers`
        "hidden_dim": "intermediate_size",  # for `transformers`
    }

@wj-Mcat @sijunhe 请评审，谢谢！：）

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Merging #5672 (3337b60) into develop (848f5c1) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop    #5672      +/-   ##
===========================================
+ Coverage    59.60%   59.61%   +0.01%     
===========================================
  Files          483      483              
  Lines        68134    68155      +21     
===========================================
+ Hits         40610    40633      +23     
+ Misses       27524    27522       -2     

Impacted Files	Coverage Δ
paddlenlp/transformers/distilbert/configuration.py	100.00% <100.00%> (ø)
paddlenlp/transformers/distilbert/modeling.py	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[sijunhe]

非常好，赞！