5 Issues closed by 3 people

• 建议再发布一个Deepseek-r1的直接量化版本

  #188 closed Mar 11, 2025

• 可否给几个纯RL训练的数据示例？

  #529 closed Mar 7, 2025

• 模型是否具备记忆、模糊记忆、遗忘机制。

  #543 closed Mar 6, 2025

• 关于进一步优化模型的一些想法。

  #541 closed Mar 6, 2025

• 关于进一步优化模型的一些想法。

  #542 closed Mar 6, 2025

24 Issues opened by 22 people

• 提升AI学习效率——打破“独立事件“训练框架

  #562 opened Mar 12, 2025

• [Question] Issue with evaluation results and prompts on math500 dataset for qwen-R1distill-1.5B model

  #561 opened Mar 11, 2025

• 关于模型日志的问题

  #560 opened Mar 11, 2025

• GRPO显存

  #559 opened Mar 11, 2025

• 如何限制思考输出的内容，现在会将给的prompt输出

  #558 opened Mar 11, 2025

• 结束标识 <｜end▁of▁sentence｜>

  #557 opened Mar 11, 2025

• 建个wechat群,高效交流

  #556 opened Mar 11, 2025

• think output question

  #555 opened Mar 10, 2025

• 使用auto-gptq 量化DeepSeek-R1-Distill-Llama-8B模型，遇到了错误Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0! (when checking argument for argument mat2 in method wrapper_CUDA_bmm)

  #554 opened Mar 9, 2025

• 请问R1的2025-01-20版本的训练数据截止到什么时候？

  #553 opened Mar 8, 2025

• 把公式显示改回来吧, 现在深色界面都看不见公式了

  #552 opened Mar 8, 2025

• 提高AI对数据的利用率——培养AI的经验抽象、迁移应用的能力

  #551 opened Mar 7, 2025

• openai 接口调用 怎么跳过 think

  #550 opened Mar 7, 2025

• 本地部署的deepseek32b模型依靠本地的知识库回答时给出的链接不准确

  #549 opened Mar 7, 2025

• 请问官方有提供评测姿势，可以复现评测结果嘛？

  #548 opened Mar 7, 2025

• 找不到 `deepseek-r1-main.py` 文件

  #547 opened Mar 7, 2025

• 4卡T4服务器，使用vllm启动总是报错显存不够

  #546 opened Mar 7, 2025

• R1思维链频繁切换思路的改进尝试——完整性奖惩

  #545 opened Mar 6, 2025

• 可以控制思考过程中输出的内容吗, 比如思考过程文字少于30字

  #544 opened Mar 6, 2025

• R1怎么联网搜索

  #540 opened Mar 6, 2025

• Hide Reasoning feeback after response generated

  #539 opened Mar 6, 2025

• 请问R1论文中的mmlu和c-eval得分是subject平均分，还是weighted平均分

  #538 opened Mar 6, 2025

• Potential Vulnerability: Context Injection via RAG Bypasses Content Filtering

  #537 opened Mar 5, 2025

• me too

  #536 opened Mar 5, 2025

63 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.

• Create breifing on DeepSeek R1

  #530 commented on Mar 9, 2025 • 2 new comments

• Request for GRPO Training Code Release

  #187 commented on Mar 11, 2025 • 0 new comments

• Incorrect Sensitivity Filtering for Chinese Medical Queries & Inconsistent Model Version Responses

  #192 commented on Mar 11, 2025 • 0 new comments

• 太奶漏洞测试非法行为时，R1模型对这种case的回复有敏感问题的bug

  #196 commented on Mar 11, 2025 • 0 new comments

• Table4中IF-Eval的结果有误？

  #297 commented on Mar 10, 2025 • 0 new comments

• DeepSeek-R1 model on HuggingFace Hub

  #363 commented on Mar 10, 2025 • 0 new comments

• 【Problem】关于deepseek-r1思考过程太长，导致输出停止的问题

  #379 commented on Mar 10, 2025 • 0 new comments

• Rate Limit Reached When Accessing the Website

  #138 commented on Mar 10, 2025 • 0 new comments

• Feature Suggestion: Download Chats

  #198 commented on Mar 10, 2025 • 0 new comments

• Adding extra/unexpected spaces when generating code (at least for Java code)

  #201 commented on Mar 10, 2025 • 0 new comments

• Lack of Architecture Details in README

  #206 commented on Mar 10, 2025 • 0 new comments

• Hugging Face Support

  #207 commented on Mar 10, 2025 • 0 new comments

• vLLM Configuration

  #208 commented on Mar 10, 2025 • 0 new comments

• Benchmarking Script Access

  #209 commented on Mar 10, 2025 • 0 new comments

• Infinite loop on some queries

  #211 commented on Mar 10, 2025 • 0 new comments

• 关于DeepSeek-R1的建议

  #501 commented on Mar 9, 2025 • 0 new comments

• ds内置的农历和公历转换算法有误

  #453 commented on Mar 5, 2025 • 0 new comments

• Temperature

  #185 commented on Mar 11, 2025 • 0 new comments

• Bug using DeepSeek

  #184 commented on Mar 11, 2025 • 0 new comments

• 关于CPU推理的速度问题。

  #183 commented on Mar 11, 2025 • 0 new comments

• 本地部署DeepSeek-R1-Distill-Qwen-32B,输出仅有</think>，没有<think>

  #352 commented on Mar 11, 2025 • 0 new comments

• deepseek-reasoner does not support successive user or assistant messages

  #21 commented on Mar 11, 2025 • 0 new comments

• 请问用官方蒸馏的那些Qwen模型的时候需要写system提示吗？

  #417 commented on Mar 11, 2025 • 0 new comments

• 能给一个search_answer_zh_template 的完整测试样本吗？

  #465 commented on Mar 11, 2025 • 0 new comments

• 如何关闭本地部署r1的思考过程

  #512 commented on Mar 11, 2025 • 0 new comments

• Voice Chat & AI Speech Feature

  #156 commented on Mar 12, 2025 • 0 new comments

• Feature Request: Voice Chat Functionality for the AI Chat Model

  #155 commented on Mar 12, 2025 • 0 new comments

• Optimized for Vulcan (AMD GPU)

  #144 commented on Mar 12, 2025 • 0 new comments

• section to update your password is not present

  #121 commented on Mar 12, 2025 • 0 new comments

• API-Platform is down

  #120 commented on Mar 12, 2025 • 0 new comments

• Made French version of README and added link to it in original README.md

  #181 commented on Mar 11, 2025 • 0 new comments

• docs: 添加中文翻译 (Add Chinese translation)

  #210 commented on Mar 5, 2025 • 0 new comments

• JSONDecodeError: Expecting value: line 5 column 1 (char 4)

  #91 commented on Mar 5, 2025 • 0 new comments

• 能再开源一下处理文件和联网搜索的 prompt 么？

  #407 commented on Mar 6, 2025 • 0 new comments

• 本地部署DeepSeek-R1-671B,输出仅有</think>，没有<think>

  #526 commented on Mar 6, 2025 • 0 new comments

• 有微调和蒸馏deepseek的方法吗？

  #477 commented on Mar 6, 2025 • 0 new comments

• Feature Request; Chunked Transfer of large files

  #263 commented on Mar 7, 2025 • 0 new comments

• Image upload failed

  #259 commented on Mar 7, 2025 • 0 new comments

• Blue Color font on chat box [UI friendly]

  #258 commented on Mar 7, 2025 • 0 new comments

• A proposal for adding a fine-tuning offical script

  #254 commented on Mar 7, 2025 • 0 new comments

• Deepseek web version (without deepthinking) repeatedly outputs the same few sentences within a single response.

  #253 commented on Mar 7, 2025 • 0 new comments

• Consultations regarding DeepSeekR1

  #249 commented on Mar 7, 2025 • 0 new comments

• Server is busy

  #248 commented on Mar 7, 2025 • 0 new comments

• Is it possible to have "Project" or "Folder" for better chat organization?

  #535 commented on Mar 7, 2025 • 0 new comments

• 关于 DeepSeek-R1 蒸馏过程的疑问

  #113 commented on Mar 7, 2025 • 0 new comments

• GRPO 到底属于 on-policy 还是 off-policy 呢？

  #527 commented on Mar 7, 2025 • 0 new comments

• Feature Request: Implement Session Memory/Persistent Context

  #67 commented on Mar 7, 2025 • 0 new comments

• 关于DeepSeek-R1强化学习训练过程的一个小疑问

  #412 commented on Mar 7, 2025 • 0 new comments

• Infinte loop

  #242 commented on Mar 8, 2025 • 0 new comments

• How to run r1 on a 4090 card?

  #240 commented on Mar 8, 2025 • 0 new comments

• 模型在回答的文本中没有正确加粗

  #239 commented on Mar 8, 2025 • 0 new comments

• huggingface 上 deepseek-r1 源码中采用 MLA 架构的 KV Cache 压缩存储策略的实现似乎与文中说的不一致，这是为什么？代码中似乎没实现这个大优化

  #238 commented on Mar 8, 2025 • 0 new comments

• Exploring Expert Activation, Fine-Tuning, and Stability Challenges in AI Reasoning

  #235 commented on Mar 8, 2025 • 0 new comments

• How to drive with GPU/NPU in windows ARM Platform?

  #236 commented on Mar 8, 2025 • 0 new comments

• Is it possible to donate GPU power by people for training of upcoming models ?

  #232 commented on Mar 8, 2025 • 0 new comments

• 所有，谁能告诉我开源代码在哪里？

  #528 commented on Mar 8, 2025 • 0 new comments

• Failure to observe spacing between words in Persian language responses

  #230 commented on Mar 9, 2025 • 0 new comments

• Request to Enable GitHub Discussions for Enhanced Community Collaboration

  #225 commented on Mar 9, 2025 • 0 new comments

• DeepSeek R1 Produces Inconsistent Formatting (Fonts, Spacing, and Style Issues)

  #224 commented on Mar 9, 2025 • 0 new comments

• Does single prompt activate different group of expert during reasoning inference?

  #222 commented on Mar 9, 2025 • 0 new comments

• 模型存在的一些知识错误

  #217 commented on Mar 9, 2025 • 0 new comments

• Text Formatting and Rendering Issue:

  #216 commented on Mar 9, 2025 • 0 new comments

• When the chat window is left unatended for long time, multiple CloudFlare dialogs appear

  #212 commented on Mar 9, 2025 • 0 new comments