怎么可以得到例如ch_PP-OCRv3_det_slim或ch_PP-OCRv3_rec_slim的超轻量模型

[leesimeng]

求助，请问一下怎么可以得到例如ch_PP-OCRv3_det_slim或ch_PP-OCRv3_rec_slim的超轻量模型？
目前有尝试使用ch_PP-OCRv3_rec_distillation.yml训练蒸馏模型，最后得到的Student inference(9m左右)模型转nb后大小仍为9m左右。

如果想要得到量化蒸馏的模型，检测1或者2m左右，识别3-5m左右，应该使用对应哪个配置文件，预训练模型可以使用哪些，以及教师和Student的预训练模型是否需要

[GreatV]

要获得超轻量的 ch_PP-OCRv3_det_slim 或 ch_PP-OCRv3_rec_slim 模型（目标模型大小为检测1-2M、识别3-5M左右），建议采用结合了蒸馏与量化的优化方案。下面是具体的步骤和建议配置：

一、使用量化蒸馏获得轻量模型的流程：

1. 安装 PaddleSlim（模型压缩组件）：

   pip install paddleslim==2.3.2

2. 选择合适的 Student 网络结构（使用 Slim 配置）：

   • 对于检测模型：使用配置文件 configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml
   • 对于识别模型：建议根据 ch_PP-OCRv3_rec_distillation.yml 修改 backbone（例如使用 MobileNetV3 Small 或更简单结构）来达到瘦身效果。

3. 下载并使用官方提供的蒸馏训练好的 Teacher 模型：

   • 检测 Teacher 模型下载地址（文档中提供）：

     wget https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_distill_train.tar
     tar -xf ch_PP-OCRv3_det_distill_train.tar

   • 识别 Teacher 模型可参考你已有的 ch_PP-OCRv3_rec_distillation.yml 中指定模型。

4. 执行量化训练：
   以检测模型为例：

   python deploy/slim/quantization/quant.py \
     -c configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml \
     -o Global.pretrained_model=./ch_PP-OCRv3_det_distill_train/best_accuracy \
        Global.save_model_dir=./output/quant_model_distill/

   使用的是量化+预训练Student结构的整合。识别模型你也可采用同样方式，只需替换配置文件路径和模型路径。

5. 导出量化后的 inference 模型：

   python deploy/slim/quantization/export_model.py \
     -c configs/det/ch_PP-OCRv3/ch_PP-OCRv3_det_cml.yml \
     -o Global.checkpoints=./output/quant_model_distill/best_accuracy \
        Global.save_inference_dir=./output/quant_inference_model

6. 模型转换部署（量化模型仍是FP32结构，需配合PaddleLite等工具做真正INT8部署）：

   • 使用 Paddle Lite 的 opt 工具或 TensorRT（设置 precision=INT8）进行最终部署。

备注说明：

• 若模型转nb格式后大小仍为9M，可能是量化过程未采用轻量Student结构，或模型结构中未完成充分压缩；
• ch_PP-OCRv3_det_cml.yml 可视为一种组合压缩配置（含蒸馏和量化），适用于你想达到目标模型大小的需求；
• 教师模型必须提供，Student须采用简化骨干网络的结构（例如 MobileNetV3 Small 或 Tiny）；
• 还可考虑进一步结构剪枝来进一步瘦身，但需关注精度的损失。

参考文档已说明该流程适用于小模型在移动端部署，你可以结合 PaddleSlim 与轻量骨干网络结构进一步压缩模型大小。如需更具体配置文件，可查看 PaddleOCR 仓库的 configs/det/ch_PP-OCRv3/ 与 configs/rec/ch_PP-OCRv3/ 路径下的 cml.yml 或 distillation.yml 示例。

---

Response generated by 🤖 feifei-bot | chatgpt-4o-latest

⚠️ This is an AI-generated response. Please verify any critical information.

[leesimeng]

非常谢谢，看步骤我应该用了常规的训练方式，后面按照上述试试

[leesimeng]

请教一下，按照上述你提供的一些建议和配置，运行训练脚本后如下异常，应该怎么解决：
[2025/04/05 14:19:51] ppocr INFO: train from scratch
[2025/04/05 14:19:51] ppocr INFO: train from scratch
[2025/04/05 14:19:51] ppocr INFO: train dataloader has 43 iters, valid dataloader has 8 iters
[2025/04/05 14:19:51] ppocr INFO: During the training process, after the 0th iteration, an evaluation is run every 200 iterations
Traceback (most recent call last):
File "/mnt/workspace/PaddleOCR/deploy/slim/quantization/quant.py", line 226, in
main(config, device, logger, vdl_writer)
File "/mnt/workspace/PaddleOCR/deploy/slim/quantization/quant.py", line 207, in main
program.train(
File "/mnt/workspace/PaddleOCR/tools/program.py", line 367, in train
loss = loss_class(preds, batch)
File "/usr/local/lib/python3.10/site-packages/paddle/nn/layer/layers.py", line 1429, in call
return self.forward(*inputs, **kwargs)
File "/mnt/workspace/PaddleOCR/ppocr/losses/combined_loss.py", line 70, in forward
loss = loss_func(input, batch, **kargs)
File "/usr/local/lib/python3.10/site-packages/paddle/nn/layer/layers.py", line 1429, in call
return self.forward(*inputs, **kwargs)
File "/mnt/workspace/PaddleOCR/ppocr/losses/distillation_loss.py", line 111, in forward
loss = super().forward(out1[self.dis_head], out2[self.dis_head])
File "/mnt/workspace/PaddleOCR/ppocr/losses/basic_loss.py", line 129, in forward
loss = (self._kldiv(log_out1, out2) + self._kldiv(log_out2, out1)) / 2.0
File "/mnt/workspace/PaddleOCR/ppocr/losses/basic_loss.py", line 116, in _kldiv
loss = target * (paddle.log(target + eps) - x)
ValueError: (InvalidArgument) Broadcast dimension mismatch. Operands could not be broadcast together with the shape of X = [64, 40, 12] and the shape of Y = [64, 80, 12]. Received [40] in X is not equal to [80] in Y at i:1.
[Hint: Expected x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1 == true, but received x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1:0 != true:1.] (at /paddle/paddle/phi/kernels/funcs/common_shape.h:73)

详细环境和配置提了一个issue，链接：#14978

[leesimeng]

请教一下，按照上述你提供的一些建议和配置，运行训练脚本后如下异常，应该怎么解决：
[2025/04/05 14:19:51] ppocr INFO: train from scratch
[2025/04/05 14:19:51] ppocr INFO: train from scratch
[2025/04/05 14:19:51] ppocr INFO: train dataloader has 43 iters, valid dataloader has 8 iters
[2025/04/05 14:19:51] ppocr INFO: During the training process, after the 0th iteration, an evaluation is run every 200 iterations
Traceback (most recent call last):
File "/mnt/workspace/PaddleOCR/deploy/slim/quantization/quant.py", line 226, in
main(config, device, logger, vdl_writer)
File "/mnt/workspace/PaddleOCR/deploy/slim/quantization/quant.py", line 207, in main
program.train(
File "/mnt/workspace/PaddleOCR/tools/program.py", line 367, in train
loss = loss_class(preds, batch)
File "/usr/local/lib/python3.10/site-packages/paddle/nn/layer/layers.py", line 1429, in call
return self.forward(*inputs, **kwargs)
File "/mnt/workspace/PaddleOCR/ppocr/losses/combined_loss.py", line 70, in forward
loss = loss_func(input, batch, **kargs)
File "/usr/local/lib/python3.10/site-packages/paddle/nn/layer/layers.py", line 1429, in call
return self.forward(*inputs, **kwargs)
File "/mnt/workspace/PaddleOCR/ppocr/losses/distillation_loss.py", line 111, in forward
loss = super().forward(out1[self.dis_head], out2[self.dis_head])
File "/mnt/workspace/PaddleOCR/ppocr/losses/basic_loss.py", line 129, in forward
loss = (self._kldiv(log_out1, out2) + self._kldiv(log_out2, out1)) / 2.0
File "/mnt/workspace/PaddleOCR/ppocr/losses/basic_loss.py", line 116, in _kldiv
loss = target * (paddle.log(target + eps) - x)
ValueError: (InvalidArgument) Broadcast dimension mismatch. Operands could not be broadcast together with the shape of X = [64, 40, 12] and the shape of Y = [64, 80, 12]. Received [40] in X is not equal to [80] in Y at i:1.
[Hint: Expected x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1 == true, but received x_dims_array[i] == y_dims_array[i] || x_dims_array[i] <= 1 || y_dims_array[i] <= 1:0 != true:1.] (at /paddle/paddle/phi/kernels/funcs/common_shape.h:73)

详细环境和配置提了一个issue，链接：#14978