Repository Issue Activity (beta)

#4908 [Bug]: ignore_eos_token keeps EOS in DPO length normalization

#4472 新增模型提交PR的一些注意点

#4447 [Question]: paddleocr-vl1.5支持升腾卡训练吗

#4397 [Security] Request for private disclosure channel

#4778 [Bug]: Could not import module 'AutoConfig'. Are this object's requirements defined correctly?

#4279 [Question]: per_device_eval_batch_size

#4332 [Question]: 修改模型config未生效，命令行指定FP16精度训练未生效

#4152 [Bug]: 'GPTModel' object has no attribute 'generate'

#3990 [Bug]: sft-lora微调qwen3-vl报错:KeyError: '1.self_attn.o_proj.lora_A'

#3982 [Question]: 在使用CASIA手写数字训练集对PaddleOCR-VL进行LoRA微调后出现知识遗忘。

#4471 新增模型PR的注意点

#3699 常见问题处理

#3918 [Question]: 对paddleocr-vl进行LORA微调时报错 Warning: Error processing example, skipping. Error: 'SiglipImageProcessor' object has no attribute 'get_smarted_resize'

#3854 [Bug]: 在Lora微调的时候，训练轮次一超过10次及以上，以及数据集的量一大，就会报错：boolean index did not match indexed array along dimension 0; dimension is 21 but corresponding boolean dimension is 130

#3759 [Bug] NameError: name 'GPTModel' is not defined during evaluation in SFT LoRA training

#4104 是否支持从huggingface、ModelScope加载数据集呢？

#4068 [Security] subprocess invocation uses shell=True

#3680 [Question]: paddleocrVL-sft全量训练后用vllm服务化部署时报错ValueError: Following weights were not initialized from checkpoint: {'visual.vision_model.embeddings.packing_position_embedding.weight'}

#4067 [Security] pickle used for data deserialization

#3069 Resolve certain issues that arise when using the FC framework within PaddleFormers.

#3958 PaddleOCRVL 多卡运行，在加载数据时，会出现加载时间飙升的问题

#3915 [Bug]: paddleformer 明明开启了转静模式，但是模型却无法转静执行，也没有执行器的LOG

#3917 ImportError: cannot import name 'AUTO_TO_BASE_CLASS_MAPPING' from 'transformers.processing_utils'

#3889 holder

#2397 [Bug]: Tokenizer class Ernie4_5_Tokenizer does not exist or is not currently imported