- 000
我是用最新版本PaddleNLP对UIE-m-large版本进行finetune，很顺利。
但是在加载已经导出到onnx模型时，遇到问题。
报错内容为：ModelProto does not have a graph.
具体为：
model_zoo/uie/deploy/python/uie_predictor.py , line: 30
onnx_model = paddle2onnx.command.c_paddle_to_onnx( ... )
输入的模型路径没问题，但...
- 000
我是用最新版本PaddleNLP对UIE-m-large版本进行finetune，很顺利。
但是在加载已经导出到onnx模型时，遇到问题。
报错内容为：ModelProto does not have a graph.
具体为：
model_zoo/uie/deploy/python/uie_predictor.py , line: 30
onnx_model = paddle2onnx.command.c_paddle_to_onnx( ... )
输入的模型路径没问题，但...
- 000
请提出你的问题
Hi @authors,
Could you add the FUNSD training set preparation script for ernie-layout custom training.
This data set - "https://bj.bcebos.com/paddlenlp/datasets/funsd.tar.gz" is preformatted and could you add the conversion script/data prepartion script for the same.
Thanks,
- 000
请提出你的问题
I am doing below:
MODEL_CHKPOINT = "ernie-layoutx-base-uncased/models/checkpoint-xxxx"
tokenizer = AutoTokenizer.from_pretrained(MODEL_CHKPOINT)
model = AutoModelForQuestionAnswering.from_pretrained(MODEL_CHKPOINT)
model.config['has_visual_segment_embedding'] = False
tmp_data = eval_d...
- 000
Feature request
Could you please add fine tuning scripts for SROIE and CORD datasets. We could see only for FUNSD.
Thanks.
Motivation
Could you please add fine tuning scripts for SROIE and CORD datasets. We could see only for FUNSD.
Your contribution
Could you please add fine tuning scripts for SROI...
Issue
[Question]: 多标签分类问题
- 000
请提出你的问题
多标签分类所有预测结果都是一样的，自己用脚本处理了一个数据集，是按照文档来的，但是每次训练很快就结束了
global step 5, epoch: 1, batch: 5, loss: 0.59483, speed: 2.38 step/s
[2023-01-19 19:01:46,912] [ INFO] - global step 10, epoch: 1, batch: 10, loss: 0.52392, speed: 9.89 step/s
[2...
- 000
软件环境
- paddlepaddle:None
- paddlepaddle-gpu: 2.4.1
- paddlenlp: 2.5.0/develop
重复问题
I have searched the existing issues
错误描述
加载 Multiple Choice缺少一个'BertConfig' 缺失'num_choices'
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/workspac...
- 000
软件环境
- paddlepaddle: None
- paddlepaddle-gpu: 2.4.1
- paddlenlp: develop
- fast-tokenizer-python：1.0.1
重复问题
I have searched the existing issues
错误描述
从hub上加载的分词模型并不支持fast tokenizer，安装了fast tokenizer的环境下从hub加载分词器会报错没有'BertFastTokenizer'。逻辑缺少在init_c...