21 Pull requests merged by 14 people

• Fix a small bug of sharding reshard

  #10613 merged May 19, 2025

• [LLM] fix normal save of rng states

  #10609 merged May 18, 2025

• Add offload opt util

  #10608 merged May 17, 2025

• Add offload optimizer function

  #10607 merged May 17, 2025

• [RL] logprob compute use the same method

  #10596 merged May 15, 2025

• Refine setup.py

  #10577 merged May 15, 2025

• [INTEL_HPU] add e2e auto test cases

  #10108 merged May 14, 2025

• [paddlenlp_ops] update whl

  #10576 merged May 14, 2025

• Update serialization

  #10584 merged May 14, 2025

• [Tokenizer] Fix padding_side for performance

  #10585 merged May 13, 2025

• Support arbitrary num_experts and topk, with bfloat16 zip prob.

  #10583 merged May 13, 2025

• update default sm num

  #10586 merged May 13, 2025

• [RL] Fix eos_mask and position ids

  #10588 merged May 13, 2025

• Refactor RL

  #10569 merged May 13, 2025

• [infer] dy_insert check eos_token_id

  #10587 merged May 13, 2025

• [infer] get_output and save_output recover input without msg_queue_id

  #10574 merged May 13, 2025

• [CI]add rl to ci

  #10553 merged May 13, 2025

• 增加rl目录下的init.py文件

  #10581 merged May 12, 2025

• add xpu p800 release dockerfile

  #10552 merged May 12, 2025

• optimize fp8 deep gemm tma

  #10580 merged May 12, 2025

• Support tpdp-ep sharding reshard

  #10568 merged May 12, 2025

16 Pull requests opened by 13 people

• Adding transposed scale support for dtaq op

  #10582 opened May 12, 2025

• support_dynamic_server for wintx

  #10589 opened May 13, 2025

• Update ppo_model_utils.py

  #10593 opened May 14, 2025

• [Integration] add swanlab logger

  #10594 opened May 14, 2025

• Cuda graph runner

  #10595 opened May 14, 2025

• Integrate DataProto into the GRPO

  #10597 opened May 15, 2025

• [Auto-Parallel] Add ci for tensor_fusion and overlap in auto-dy

  #10598 opened May 15, 2025

• Add benchmark for tensor_fusion and overlap in llama7b N4C32 dy_auto

  #10599 opened May 16, 2025

• Add fused_transpose_quant op

  #10601 opened May 16, 2025

• Test the improve in hand_dy llama7b tensor_fusion and sharding overlap

  #10602 opened May 16, 2025

• Adding fused_swiglu_probs_bwd op

  #10604 opened May 16, 2025

• [RL] Add reshard controller

  #10606 opened May 16, 2025

• [LLM] fix sharding v1 reshard

  #10610 opened May 18, 2025

• [llm] Add llm pretrain tutorial

  #10611 opened May 18, 2025

• add uc

  #10612 opened May 19, 2025

• [infer]support output length truncated

  #10614 opened May 19, 2025

109 Issues closed by 1 person

• [Bug]: ernie预训练报错

  #6723 closed May 13, 2025

• [Question]: https://github.com/PaddlePaddle/PaddleNLP/issues

  #6718 closed May 13, 2025

• You are using GPU version PaddlePaddle, but there is no GPU detected on your machine. Maybe CUDA devices is not set properly

  #6707 closed May 13, 2025

• [Question]: ernie3预训练报错

  #6716 closed May 13, 2025

• [Question]: TypeError: __init__() got an unexpected keyword argument 'architectures'

  #6706 closed May 13, 2025

• [Bug]: Wrong Prediction UIE information extraction

  #6702 closed May 13, 2025

• [Question]: 并行报错：ValueError: (InvalidArgument) The axis is expected to be in range of [0, 0), but got 0

  #6697 closed May 13, 2025

• Dependency confusion supply-chain vulnerability detected

  #6657 closed May 13, 2025

• 关于多标签长文本分类

  #6652 closed May 13, 2025

• [Question]: 训练模型的隐性观点抽取能力无效？

  #6651 closed May 13, 2025

• [Bug]: uie-x gradio demo does not show any results

  #6645 closed May 13, 2025

• [Question]: paddlenlp日志输出问题

  #6619 closed May 13, 2025

• [Question]: OSError: (External) CUDA error(700), an illegal memory access was encountered.

  #6609 closed May 13, 2025

• [Question]: UIE-X分布式训练显存占用问题

  #6604 closed May 13, 2025

• [Question]: 如何打包paddle项目

  #6578 closed May 13, 2025

• [Question]: label-studio 两个字段的问题

  #6577 closed May 13, 2025

• [Bug]: TypeError: cannot pickle 'paddle.fluid.libpaddle.Genera atorstate'object

  #6576 closed May 13, 2025

• [Bug]: 手把手搭建一个语义检索系统BUG，run.system.py在rank时出错，个人进行了修复，不过返回结果超级小和官方提供的有点出路。

  #6573 closed May 13, 2025

• [Question]: 目前有效果还可以的新闻摘要模型吗？

  #6554 closed May 13, 2025

• 希望增加uie模型微调训练时，爆显存的提示

  #6553 closed May 13, 2025

• [Question]: ernie-3.0-tiny 在Android上运行，无法推理

  #6549 closed May 13, 2025

• uie不能加载finetune的模型[Question]:

  #6547 closed May 13, 2025

• [Bug]: set_seed函数放在了模型初始化之后，导致模型初始化参数每次运行都是随机生成的

  #6546 closed May 13, 2025

• [Question]: 关于utc-base和utc-large如何训练的？

  #6541 closed May 13, 2025

• [Bug]:

  #6538 closed May 13, 2025

• [Question]: fast-tokenizer 不支持python3.11？

  #6530 closed May 13, 2025

• [Bug]: uie-x推理时报onnxruntime，CUDA no kernel image is available for execution on the device错误

  #6499 closed May 13, 2025

• [Question]:在训练UIE的时候，日志loss下降的时候指标正常，loss开始上升指标就开始归0了，并且之后都是0

  #6490 closed May 13, 2025

• 『NLP打卡营』实践课7：中英文本翻译系统中引入from paddlenlp.data import Vocab, Pad from paddlenlp.datasets import load_dataset报错

  #6485 closed May 13, 2025

• [Bug]:windows系统按指导步骤搭建端到端语义检索系统，在步骤1.3与1.4.2处均报错，无法正常运行

  #6482 closed May 13, 2025

• 对m2m100 和mt5-base模型提供加速，fast transformer支持，动转静支持

  #6478 closed May 13, 2025

• [Question]: from utils import SVOInfo， 而SVOInfo根本不存在！

  #6462 closed May 13, 2025

• [Question]: 当我在使用PPDiffuser时调用fastdeploy出现了如下的错误，请问如何解决

  #6437 closed May 13, 2025

• [Question]: ErnieFastTokenizer' object has no attribute 'set_thread_num'

  #6443 closed May 13, 2025

• [Question]: PaddleNLP现在最新的代码中没有SimCSE了吗？

  #6430 closed May 13, 2025

• [Bug]: ppdiffusers 0.16.1版本提示没有导入os包

  #6420 closed May 13, 2025

• [Question]: 请问怎么将DreamBooth下训练出来的模型转换为webui能够使用的类型呢？

  #6417 closed May 13, 2025

• [Question]:

  #6406 closed May 13, 2025

• [Bug]: ValueError:(InvalidArgument) Variable value (input) of OP(fluid.layers.embeding) expected >=0 and < 40000, but got 236693.please check input value.

  #6398 closed May 13, 2025

• [Question]: ChatGLM无法调用问题

  #6381 closed May 13, 2025

• [Question]:

  #6378 closed May 13, 2025

• [Question]:

  #6379 closed May 13, 2025

• [Bug]: AttributeError: 'ModelArguments' object has no attribute 'use_fused_rms_norm'

  #6366 closed May 13, 2025

• bug?

  #6364 closed May 13, 2025

• [Question]: UIE相比bert效果如何

  #6353 closed May 13, 2025

• 模型更新

  #6352 closed May 13, 2025

• [Question]: 后面会开发ppo和reward模型的训练方法吗

  #6337 closed May 13, 2025

• [Question]: ernie doc acc

  #6331 closed May 13, 2025

• [Question]: 基于uie-base进行微调时，如何增加模型输入的最大长度，直接在微调命令中修改max_seq_length训练会直接报错

  #6327 closed May 13, 2025

• [Question]: 请问modelzoo 目录下的ppfleetx_ops中的topp_sampling 算子和TopProcess函数的区别是什么？

  #6324 closed May 13, 2025

• [Bug]: ZeroShotTextClassificationTask后处理问题

  #6315 closed May 13, 2025

• [Question]: 使用Taskflow模块，进行信息抽取输出是空

  #6311 closed May 13, 2025

• [Bug]: fastdeploy::TrtBackend::SetInputs TRTBackend SetInputs not find name:input_ids

  #6310 closed May 13, 2025

• [Bug]: 计算AccuracyAndF1出错

  #6309 closed May 13, 2025

• [Question]:如何部署UIE蒸馏的模型

  #6303 closed May 13, 2025

• 看了最近发布 PaddleNLP v2.6rc 最近更新版，整合了一些大模型，非常不错，提个建议 希望支持chatRWKV并行训练

  #6288 closed May 13, 2025

• [Question]: 编译fast_transformer的transformer_e2e出现错误

  #6287 closed May 13, 2025

• [Question]: UIE蒸馏模型的部署

  #6276 closed May 13, 2025

• [Question]: 使用uie-senta-medium做sentiment_analysis，运行不报错，自动退出程序，没有返回结果。

  #6274 closed May 13, 2025

• [Docs]:

  #6263 closed May 13, 2025

• [Question]: plato部署问题

  #6256 closed May 13, 2025

• [Question]: 如何在一个SimpleServer里面部署2个不同的taskflow任务呢？

  #6260 closed May 13, 2025

• [Question]: ImportError: cannot import name 'ElectraConfig' from 'paddlenlp.transformers'

  #6255 closed May 13, 2025

• [Question]: ernie 并行效率

  #6244 closed May 13, 2025

• [Question]: 文档智能传入PDF数据报错

  #6221 closed May 13, 2025

• [Question]: codegen推理支持单机多卡部署么？

  #6217 closed May 13, 2025

• 是否可以增加 baichuan-7B 这个模型的微调和预训练

  #6190 closed May 13, 2025

• [Question]: uie-base-en模型进行评论观点抽取，输出空值

  #6161 closed May 13, 2025

• [Question]: 按文档跑GPT模型报错

  #6158 closed May 13, 2025

• [Bug]: convert_ids_to_tokens() got an unexpected keyword argument 'skip_special_tokens'

  #6106 closed May 13, 2025

• [Question]: ptq量化时报错：Operator (fusion_unified_decoding) is not registered.

  #6105 closed May 13, 2025

• [Question]: cannot import name 'BM25Retriever' from 'pipelines.nodes' (/usr/local/python3.7.0/lib/python3.7/site-packages/pipelines-0.1.0a0-py3.7.egg/pipelines/nodes/__init__.py)

  #6076 closed May 13, 2025

• 使用paddlenlp进行prompt learning 训练multi-label任务市报错[Question]:

  #6041 closed May 13, 2025

• [Question]: UIEX finetune

  #6011 closed May 13, 2025

• paddlenlp模型怎么在spark集群上部署

  #6010 closed May 13, 2025

• [Bug]: resize_token_embedding仍然存在bug，bart,pegasus无效，其他生成模型最后一层lm_head不会被resize

  #6008 closed May 13, 2025

• [Question]: 如下代码，为什么这里要硬编码entitie_label == 2呢，如果关联关系里面正好没有值为2的这个关系，就会报错， 例如有4个label:{'a':1,'b':2,'c':3,'d':4},训练数据里的关系只有[(1,3),(1,4)],走到这一步时就会产生一个空集合，再往下面走就会报Floating point exception

  #5985 closed May 13, 2025

• [Bug]:

  #5979 closed May 13, 2025

• [Question]: 打印模型信息

  #5975 closed May 13, 2025

• [Question]: 为什么vocab.txt里会有两个相同的字符？

  #5974 closed May 13, 2025

• [Question]: ERNIE-3.0-base-tiny-v2-zh还开源吗？

  #5939 closed May 13, 2025

• [Question]: chatglm-6b，在微调的时候，loss和 rouge得分一般达到多少就算模型比较好了？

  #5934 closed May 13, 2025

• [Question]: 使用层次分类时，如何构建 不属于我需要的任何类别的 其他 类型的数据

  #5928 closed May 13, 2025

• [Question]: 长文本fitune

  #5926 closed May 13, 2025

• [Bug]: pyspark不能加载uie的模型包

  #5915 closed May 13, 2025

• [Question]: ImportError: cannot import name 'PegasusConfig' from 'paddlenlp.transformers'

  #5889 closed May 13, 2025

• openwebtext2的下载链接进不去

  #5845 closed May 13, 2025

• [Bug]: ont-hot 编码为int 算loss报错

  #5785 closed May 13, 2025

• [Question]:

  #5713 closed May 13, 2025

• [Bug]: 使用预编译好的Android-lite-arm64-v8a 导致Segmentation fault

  #5682 closed May 13, 2025

• [Question]: 如何应用其他语言 / How to apply the other language

  #5630 closed May 13, 2025

• [Question]: compile fast_tokenizer failed with undeclared identifier backtrace

  #5681 closed May 13, 2025

• [Bug]: 使用一键预测依存句法分析时，使用分词结果来输入时，不能批量预测

  #5621 closed May 13, 2025

• [Question]使用paddlenlp的uie模型同义词不起作用

  #5610 closed May 13, 2025

• [Question]: 利用ernie-3.0-medium-zh自定义数据集预训练模型训练后，转为静态图后，模型的预测结果不准确。

  #5608 closed May 13, 2025

• [Question]: T5-FastGeneration精度无法对齐

  #5604 closed May 13, 2025

• [Question]: 请问一下在finetune ernie-vil2.0的时候，报以下错误是怎么回事？

  #5601 closed May 13, 2025

• 基于军事题材样本的“实体关系”提取finetune.py脚本执行卡住不动的问题

  #5592 closed May 13, 2025

• [Bug]: BiGRU+CRF训练快递序列标注后导出静态模型后部署问题：

  #5588 closed May 13, 2025

• [Question]: 请问如何使用静态图方式加载ERNIE 3.0 Tiny-Pico-v2-zh模型并进行训练

  #5581 closed May 13, 2025

• [Question]: 如何通过自定义的vocab.txt生成tokenizer？

  #5508 closed May 13, 2025

• uie 实体名称的提取问题

  #5501 closed May 13, 2025

• [Question]: 源文件可以分批次生成索引库么

  #5474 closed May 13, 2025

• [Question]: 您好，请问Prompt范式训练模型，是否支持使用 Trustai 模块分析。

  #5463 closed May 13, 2025

• [Bug]: 只要一进入调试模式就会出现TypeError: __init__() got multiple values for argument 'self'，正常运行则没有

  #5461 closed May 13, 2025

• [Question]: RequestsDependencyWarning: urllib3 (1.26.15) or chardet (5.1.0)/charset_normalizer (3.1.0) doesn't match a supported version! ImportError cannot import name 'inference' from 'paddle'

  #5421 closed May 13, 2025

• [Question]: How to specify model_name_or_path to other drive letters

  #5418 closed May 13, 2025

• paddleNlp中的UIE模型的schema如何定义才能尽量抽取质量好的内容

  #5415 closed May 13, 2025

• [Bug]: 关于多线程使用paddle的UIE模型推理

  #5410 closed May 13, 2025

3 Issues opened by 2 people

• [Question]: 请问如何使用paddlenlp微调Llama-3.3-70B-Instruct的模型？

  #10600 opened May 16, 2025

• [Bug]:使用pp-UIE模型时，使用streamlit缓存模型或者使用@st.cache.source来注解加载模型的函数都会导致模型出现异常

  #10592 opened May 14, 2025

• [Question]: When I was fine-tuning Llama-3.1-8B using PaddleNLP, the fine-tuning step of the model stayed at step 156 for a long time, and the program did not terminate with an error.

  #10591 opened May 14, 2025

37 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [Inference] Add new wint2.75/wint2.5 quant type and support DeepseekV3

  #10578 commented on May 15, 2025 • 8 new comments

• [Question]: 在执行命名实体识别的代码时，并没有识别出来，且似乎是崩溃还是不兼容问题

  #10069 commented on May 14, 2025 • 0 new comments

• Add fused_topk_to_multihot custom op to prevent CPU stall.

  #10112 commented on May 13, 2025 • 0 new comments

• Add fused_topk_to_multihot / fused_multihot_prob_backto_topk(grad) custom op to prevent CPU stall.

  #10125 commented on May 13, 2025 • 0 new comments

• Split predictor.

  #10136 commented on May 18, 2025 • 0 new comments

• Added regroup padded op to perform token regroup using 1D expert_idx with max-token-per-expert padding

  #10164 commented on May 18, 2025 • 0 new comments

• Add custom op for Tokens zip and unzip, in preparation of using groupgemm

  #10169 commented on May 18, 2025 • 0 new comments

• [Infer] Change groupwise weight quant from cpu to gpu for deepseek_v2 model

  #10174 commented on May 16, 2025 • 0 new comments

• [LLM] remove distutils in pdc sdk

  #10176 commented on May 19, 2025 • 0 new comments

• [LLM] remove distutils in pdc sdk

  #10179 commented on May 18, 2025 • 0 new comments

• Dsv3 dev

  #10273 commented on May 15, 2025 • 0 new comments

• update DeepGEMM

  #10429 commented on May 14, 2025 • 0 new comments

• [Taskflow] Fix the recognition bug of json format with both PIR suffix and id2label

  #10487 commented on May 14, 2025 • 0 new comments

• 【PaddleNLP No.1】 add pretrain.md

  #10506 commented on May 15, 2025 • 0 new comments

• [INTEL_HPU] Enable FusedBlockMultiTransformerHPU

  #10514 commented on May 19, 2025 • 0 new comments

• optimize a8w8

  #10516 commented on May 15, 2025 • 0 new comments

• [RL] Refoctor code and add comments

  #10565 commented on May 13, 2025 • 0 new comments

• 【Inference】support rl

  #10567 commented on May 14, 2025 • 0 new comments

• [Performance] Boost paddlenlp_ops Compiling

  #10579 commented on May 13, 2025 • 0 new comments

• [Bug]: UIE信息抽取预测时，请求多次，每次返回结果不固定

  #10150 commented on May 17, 2025 • 0 new comments

• [Question]: Error loading layoutlmv2-base-uncased: Missing model_state.pdparams file

  #9868 commented on May 17, 2025 • 0 new comments

• [Question]: doccano=1.6.2，autolabeling无法实现，一直转圈

  #9744 commented on May 17, 2025 • 0 new comments

• [Question]: paddlenlp 3.0.0b1版本使用uie-m-base报错找不到static/inference.pdmodel

  #9646 commented on May 17, 2025 • 0 new comments

• [Question]: ernie-3-tiny 按文档步骤运行例子报错

  #9896 commented on May 18, 2025 • 0 new comments

• [Question]: LogitsProcessorList 缺少__iter__和extend方法

  #9926 commented on May 19, 2025 • 0 new comments

• [Question]: Will open dataset for ppuie model

  #10556 commented on May 19, 2025 • 0 new comments

• [AutoParallel] Fit job schedule profiler tool for to static mode

  #8488 commented on May 17, 2025 • 0 new comments

• [NPU] Add chatglmv3-6b

  #9213 commented on May 19, 2025 • 0 new comments

• [MLU] taskflow mlu

  #9635 commented on May 18, 2025 • 0 new comments

• [LLM] Add Aquila2 model

  #9692 commented on May 19, 2025 • 0 new comments

• Add galvatron search engine & hardware bandwidth profile for paddle.

  #9731 commented on May 18, 2025 • 0 new comments

• [Infer] change block attention as default attention and remove normal attention in inference mode

  #9770 commented on May 12, 2025 • 0 new comments

• LoriKiT

  #9776 commented on May 12, 2025 • 0 new comments

• [Infer] Add the custom_ops for paddlenlp

  #9794 commented on May 18, 2025 • 0 new comments

• [Inference] Make inference_model on by default.

  #9879 commented on May 14, 2025 • 0 new comments

• Add fused_swiglu_act(transpose)_quant op to extern op in gpt-3

  #10076 commented on May 13, 2025 • 0 new comments

• 2:4 sparse for int8/fp8/bf16/fp16 gemm

  #10081 commented on May 13, 2025 • 0 new comments