Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / hiyouga/LLaMA-Factory issues and pull requests

#5599 - Base model pretrain doesn't have eos token?

Issue - State: closed - Opened by sts07142 about 1 month ago - 1 comment
Labels: solved

#5598 - Llama3.2 3B 奇慢无比

Issue - State: closed - Opened by dayuyang1999 about 1 month ago - 1 comment
Labels: solved

#5597 - Order of samples when doing batch inference

Issue - State: open - Opened by dayuyang1999 about 1 month ago
Labels: pending

#5595 - Question on deepspeed zero3 + qlora

Issue - State: closed - Opened by mces89 about 1 month ago - 1 comment
Labels: solved

#5593 - 直接调用llama-3-8B作为base model并且完成sft训练之后，模型输出一直重复

Issue - State: closed - Opened by ZWH-ASTAR about 1 month ago - 1 comment
Labels: solved

#5592 - [BUG] Unable to run model post training with Unsloth+DoRa+RsLora

Issue - State: closed - Opened by Tejaswgupta about 1 month ago - 2 comments
Labels: solved

#5590 - 2Nodes * 8 A100 80G sft full Qwen2VL OOM

Issue - State: open - Opened by VincentVanNF about 1 month ago - 1 comment
Labels: pending

#5589 - qwen2vl-72b推理脚本

Issue - State: closed - Opened by yhy-2000 about 1 month ago - 1 comment
Labels: solved

#5586 - ds_z3_config.json stage3_prefetch_bucket_size 应该是一个整数

Issue - State: open - Opened by ZhuJD-China about 1 month ago - 3 comments
Labels: pending

#5585 - Support EXAONE3.0 Model

Pull Request - State: closed - Opened by shing100 about 1 month ago - 1 comment
Labels: solved

#5584 - 资源问题

Issue - State: closed - Opened by weilx2267 about 1 month ago - 1 comment
Labels: solved

#5582 - Qwen-2.5+unsloth 是否支持64k的训练?

Issue - State: closed - Opened by XiangTodayEatsWhat about 1 month ago - 1 comment
Labels: solved

#5581 - [WIP] Support Pixtral-12B

Pull Request - State: open - Opened by Kuangdd01 about 1 month ago - 4 comments
Labels: pending

#5580 - made a small change to a warning about fa2 for gemma2 models.

Pull Request - State: closed - Opened by amrear about 1 month ago
Labels: solved

#5579 - 如何保存fp16格式的模型权重

Issue - State: closed - Opened by lufred8341 about 1 month ago - 2 comments
Labels: solved

#5578 - 请问有没有一个评估结果展示的界面效果或实现可以参考？

Issue - State: closed - Opened by czhcc about 1 month ago
Labels: wontfix

#5577 - Qlora训练和合并（使用量化后的模型进行Qlora训练，显示不能合并）

Issue - State: closed - Opened by xuwang0117 about 1 month ago - 3 comments
Labels: solved

#5574 - support llava-next(video)/video-llava

Pull Request - State: closed - Opened by BUAADreamer about 1 month ago
Labels: solved

#5573 - Issue when saving a checkpoint using unsloth

Issue - State: open - Opened by amrear about 1 month ago
Labels: pending

#5572 - 是否支持昇腾910B多机多卡训练?

Issue - State: closed - Opened by LtroiNGU about 1 month ago - 1 comment
Labels: wontfix, npu

#5569 - WSL下无法使用多卡运行

Issue - State: closed - Opened by gotothehill about 1 month ago - 1 comment
Labels: wontfix

#5568 - 可以尽快支持一下最新的LoRA-GA微调方法吗

Issue - State: closed - Opened by xyangyan about 1 month ago
Labels: duplicate

#5567 - when support GOT-OCR2 ？

Issue - State: open - Opened by tbwang-clound about 1 month ago
Labels: pending

#5566 - TDPO

Issue - State: open - Opened by lycheeyolo about 1 month ago
Labels: pending

#5565 - vllm多卡推理遇到的问题-qwen2.5

Issue - State: open - Opened by YChengxin about 1 month ago
Labels: pending

#5564 - vllm多卡推理-qwen2.5遇到的问题

Issue - State: closed - Opened by YChengxin about 1 month ago
Labels: duplicate

#5563 - Pei eng patch 1

Pull Request - State: closed - Opened by Pei-eng about 1 month ago
Labels: invalid

#5562 - How to align qwen2-vl fine-tuning training and vllm inference formats

Issue - State: open - Opened by xuyifan-0731 about 1 month ago
Labels: pending

#5561 - 把qwen2-7b训练模型变更成qwen2.5-32b，训练完成后推理结果不会停止

Issue - State: open - Opened by wenocy about 1 month ago - 7 comments
Labels: pending

#5560 - 多机多卡训练 一直停留在这个地方 两台机器是局域网 还需要注意什么？

Issue - State: closed - Opened by ZhuJD-China about 1 month ago - 1 comment
Labels: solved

#5559 - H20 100G*8 qwen14B full sft do_predict 阶段运行终止 failed (exitcode: -8)

Issue - State: closed - Opened by amoyplane about 1 month ago - 4 comments
Labels: solved

#5558 - 多机多卡训练 想问下现在是不支持accelerate launch训练吗

Issue - State: closed - Opened by Hansen06 about 1 month ago - 1 comment
Labels: solved

#5557 - HQQ quantization fails to serialize model

Issue - State: closed - Opened by TweedBeetle about 1 month ago - 1 comment
Labels: solved

#5556 - 如何加载模型进行beam_search

Issue - State: open - Opened by Maydaytyh about 1 month ago
Labels: pending

#5555 - Support llama3.2vl(WIP).

Pull Request - State: open - Opened by marko1616 about 1 month ago
Labels: pending

#5554 - Will it support SFT multi-modal large models (for example, qwen2-vl) with plain text?

Issue - State: closed - Opened by zhshj0110 about 1 month ago - 2 comments
Labels: solved

#5553 - A100 80G *4 sft full Qwen2-VL-72B-Instruct OOM

Issue - State: closed - Opened by VincentVanNF about 1 month ago - 3 comments
Labels: duplicate

#5552 - 原始模型和lora模型批量推理速度差异

Issue - State: closed - Opened by yysj-zq about 1 month ago - 2 comments
Labels: duplicate

#5551 - npu环境下tokenizer超时

Issue - State: closed - Opened by lambda-lee about 1 month ago - 2 comments
Labels: solved, npu

#5550 - 使用最新26日更新的LLamaFactory full sft训练qwen2-1.5B-instruct，出现deepspeed错误如下

Issue - State: closed - Opened by xiehust about 1 month ago - 2 comments
Labels: solved

#5549 - LLaMa-factory 部署Llama-3.2-11B-Vision-Instruct 运行报错

Issue - State: closed - Opened by caijx168 about 1 month ago - 20 comments
Labels: wontfix

#5548 - [Help]webui学习率调节器选择warmup_stable_decay时出现问题

Issue - State: open - Opened by ishkong about 1 month ago
Labels: enhancement, pending

#5547 - Chore: Support llama3.2.

Pull Request - State: closed - Opened by marko1616 about 1 month ago
Labels: solved

#5545 - OOM when fine tuning 8b with ~64k cutoff_len

Issue - State: closed - Opened by TweedBeetle about 1 month ago - 2 comments
Labels: solved

#5543 - Qlora训练和合并模型问题

Issue - State: closed - Opened by bitallin about 1 month ago - 4 comments
Labels: solved

#5542 - Liger kernel brake fine tuning

Issue - State: closed - Opened by arit2 about 1 month ago - 4 comments
Labels: solved

#5541 - qwen2.5-7B-instruct lora 微调 loss一直为0.0

Issue - State: closed - Opened by Liufeiran123 about 1 month ago
Labels: invalid

#5539 - lora微调qwen2.5-math-7b出问题

Issue - State: open - Opened by lin-dy about 1 month ago - 2 comments
Labels: pending

#5538 - Qwen2VL模型图像识别不准

Issue - State: open - Opened by JohnZhuYX about 1 month ago - 1 comment
Labels: pending

#5537 - throughput is much slower than expected for pre-training

Issue - State: open - Opened by lingchensanwen about 1 month ago
Labels: pending

#5536 - Update identity.json

Pull Request - State: closed - Opened by Cherry39-lab about 1 month ago
Labels: invalid

#5535 - 训练时指定evaluation_set(validation_set)

Issue - State: closed - Opened by mzc2113391 about 1 month ago - 1 comment
Labels: solved

#5534 - 训练后的agent模型，vllm部署后工具调用失败

Issue - State: closed - Opened by bingoohe about 1 month ago - 1 comment
Labels: solved

#5533 - Add additional install options to Dockerfiles

Pull Request - State: closed - Opened by StrangeBytesDev about 1 month ago
Labels: solved

#5532 - feat: Long Text Fine-Tuning Support

Pull Request - State: open - Opened by glide-the about 2 months ago - 1 comment

#5530 - deepspeed 单机多卡sft时，如何只保存模型

Issue - State: closed - Opened by lufred8341 about 2 months ago - 1 comment
Labels: solved

#5529 - 使用 unsloth 时，qwen模型会报错

Issue - State: closed - Opened by cat-knight about 2 months ago - 1 comment
Labels: wontfix

#5528 - No liger kernels will be applied. Qwen2-vl

Issue - State: closed - Opened by arit2 about 2 months ago - 1 comment
Labels: solved

#5527 - 微调qwen2 video 的时候出现，list index out of range

Issue - State: closed - Opened by wudidaxuexue about 2 months ago - 4 comments
Labels: solved

#5525 - Flash 2 attention warning, flash attention not working properly for qwen2vl

Issue - State: closed - Opened by sharonsalabiglossai about 2 months ago - 3 comments
Labels: solved

#5524 - qwen2 预训练的loss震荡

Issue - State: closed - Opened by allen20200111 about 2 months ago - 1 comment
Labels: solved

#5522 - [Update] loader.py , evaluate will run separate evaluations on each eval_dataset

Pull Request - State: open - Opened by SrWYG about 2 months ago - 2 comments
Labels: pending

#5521 - Support for loading local HuggingFace-formatted Datasets

Issue - State: closed - Opened by nathan-az about 2 months ago - 2 comments
Labels: solved

#5520 - 群二维码已经过期，无法加入，需要更新一下

Issue - State: closed - Opened by eddey666 about 2 months ago
Labels: solved

#5519 - 训练时视频处理逻辑

Issue - State: open - Opened by liuao743 about 2 months ago - 1 comment
Labels: pending

#5518 - function call 微调后部署问题

Issue - State: open - Opened by pugnazhaotianqi about 2 months ago
Labels: pending

#5517 - LlamaFactory模型合并后，推理速度很慢，且重复和乱答，动态推理正常表现

Issue - State: closed - Opened by Scorponok31 about 2 months ago - 2 comments
Labels: invalid

#5516 - 使用Qwen-7B使用Qlora时报错在阿里的PAI-DSW

Issue - State: closed - Opened by lzy728 about 2 months ago - 1 comment
Labels: solved

#5514 - 默认的optimizer是什么？如何添加自己的optimizer如SGD？

Issue - State: closed - Opened by DSW2001 about 2 months ago - 1 comment
Labels: solved

#5512 - How to train the mm_proj and the LLM part with lora of Qwen2-VL

Issue - State: open - Opened by leoozy about 2 months ago - 3 comments
Labels: pending

#5510 - [Feature Request] 请问能加入Liger-Kernel的支持吗?

Issue - State: closed - Opened by Orion-zhen about 2 months ago - 1 comment
Labels: solved

#5508 - glm4微调导入模型报错

Issue - State: closed - Opened by WWeellkkiinn about 2 months ago - 1 comment
Labels: solved

#5507 - Add deepseek-v2.5 template

Pull Request - State: open - Opened by piamo about 2 months ago
Labels: pending

#5506 - Deepseek v2.5的 template 变了，与 v2不同

Issue - State: open - Opened by piamo about 2 months ago
Labels: pending

#5505 - WSD Learning rate scheduling problem

Issue - State: closed - Opened by runningto about 2 months ago - 1 comment
Labels: solved

#5504 - pretrain from scratch 输出都是数字

Issue - State: open - Opened by UbeCc about 2 months ago
Labels: pending

#5503 - KeyError: 'messages'

Issue - State: closed - Opened by laozhai507 about 2 months ago
Labels: invalid

#5502 - 最新版代码不支持 visual_inputs 参数

Issue - State: closed - Opened by laozhai507 about 2 months ago - 1 comment
Labels: solved

#5501 - Qwen2-VL在微调之后进行merge的过程中出现ValueError: Unrecognized configuration class问题

Issue - State: closed - Opened by AmuzeLu about 2 months ago - 2 comments
Labels: solved

#5499 - 在checkpoint上继续训练，没有保存训练后的checkpint

Issue - State: closed - Opened by cuisws about 2 months ago - 2 comments
Labels: solved

#5498 - 数据长度过长，开了zero3后依旧是一个显卡装不下一条数据，没法训练

Issue - State: closed - Opened by zhangyuqi-1 about 2 months ago - 2 comments
Labels: duplicate

#5497 - save_only_model后无法续训

Issue - State: open - Opened by yuepengs about 2 months ago
Labels: pending