Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / shibing624/MedicalGPT issues and pull requests
#62 - 预训练哪些参数能有效节省显存使用呀
Issue -
State: closed - Opened by starphantom666 over 1 year ago
- 2 comments
Labels: question, wontfix
#61 - pretrain data format is a little bit similar to the sft stage
Issue -
State: closed - Opened by chlinfeng1997 over 1 year ago
- 12 comments
Labels: question
#60 - 多卡chatglm2 sft报错RuntimeError: expected scalar type Half but found Float
Issue -
State: closed - Opened by zhr0313 over 1 year ago
- 8 comments
Labels: bug
#59 - chatglm2-6b sft报错
Issue -
State: closed - Opened by zhangatao over 1 year ago
- 2 comments
Labels: bug
#58 - run_rm使用llama模型看代码做了set pad_token_id = 0还是报错
Issue -
State: closed - Opened by charryshi over 1 year ago
- 6 comments
Labels: question
#57 - 恢复预训练时报OOM.
Issue -
State: closed - Opened by Halflifefa over 1 year ago
- 4 comments
Labels: question
#56 - Prompt设置
Issue -
State: closed - Opened by Ricardokevins over 1 year ago
- 1 comment
Labels: question
#55 - lora训练完的chatglm2-6b，adapter model怎么和base model合并成新的model？
Issue -
State: closed - Opened by AaronZLT over 1 year ago
- 3 comments
Labels: question
#54 - 按run_training_pipeline.ipynb流程跑完 为什么效果很差
Issue -
State: closed - Opened by yangzhipeng1108 over 1 year ago
- 1 comment
Labels: question
#53 - 单机多卡(A100 80GB)，torchrun 数据并行报错
Issue -
State: closed - Opened by AllenYkl over 1 year ago
- 9 comments
Labels: question
#52 - RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:1 and cuda:0! (when checking argument for argument index in method wrapper__index_select)
Issue -
State: closed - Opened by dage0127 over 1 year ago
- 1 comment
Labels: bug
#51 - run_pt run_sft run_rm run_rl 这个四步没有串行 有什么意义
Issue -
State: closed - Opened by yangzhipeng1108 over 1 year ago
Labels: question
#50 - 在调用 run_sft.sh 时报错。
Issue -
State: closed - Opened by boxter007 over 1 year ago
- 4 comments
Labels: question
#49 - rm模型训练过程
Issue -
State: closed - Opened by Vincent131499 over 1 year ago
- 4 comments
Labels: bug
#48 - run_rl.sh chatglm2-6b Unrecognized configuration class <class 'transformers_modules.chatglm2-6b.configuration_chatglm.ChatGLMConfig'>
Issue -
State: closed - Opened by yangzhipeng1108 over 1 year ago
- 4 comments
Labels: bug
#47 - Expected all tensors to be on the same device, but found at least two devices, cuda:3 and cuda:0! 错误
Issue -
State: closed - Opened by dingxianzhong over 1 year ago
- 1 comment
Labels: bug
#46 - 大佬好，我看到您使用了deepspeed.config 想问下训练参数支持deepspeed的ZERO参数吗？
Issue -
State: closed - Opened by valkryhx over 1 year ago
- 28 comments
Labels: question
#45 - 似乎训练程度有点不够？
Issue -
State: closed - Opened by bash99 over 1 year ago
- 6 comments
#44 - lora模型权重合并到chatglm2-6b
Issue -
State: closed - Opened by richard880502 over 1 year ago
- 2 comments
Labels: bug
#43 - rlhf实验效果
Issue -
State: closed - Opened by zxgineng over 1 year ago
- 4 comments
Labels: bug
#42 - chatglm2-6b
Issue -
State: closed - Opened by Vincent131499 over 1 year ago
- 2 comments
Labels: enhancement
#41 - 请教一下base_model的问题
Issue -
State: closed - Opened by panliang5020 over 1 year ago
- 1 comment
Labels: question
#40 - run chatglm
Issue -
State: closed - Opened by yangzhipeng1108 over 1 year ago
- 2 comments
Labels: question
#39 - sft脚本，preprocess_function函数，chatglm输入问题
Issue -
State: closed - Opened by songt96 over 1 year ago
- 2 comments
Labels: question
#38 - 为啥我这数据集老是有问题呢
Issue -
State: closed - Opened by SolarKnight1 over 1 year ago
- 1 comment
Labels: bug
#37 - 奖励模型数据构造
Issue -
State: closed - Opened by ZTurboX over 1 year ago
- 6 comments
Labels: enhancement
#36 - 如果Stage1,2选用ChatGLM-6B作为基座model，Stage3训练奖励模型这里怎么设置呢？
Issue -
State: closed - Opened by xuanxixi over 1 year ago
- 4 comments
Labels: question
#35 - Stage 3: Reward Modeling 报错：**ValueError: weight is on the meta device, we need a `value` to put in on 1.**
Issue -
State: closed - Opened by dage0127 over 1 year ago
- 4 comments
Labels: question
#34 - ziya-llama-13b + lora推理结果异常
Issue -
State: closed - Opened by kyang888 over 1 year ago
- 9 comments
Labels: question
#33 - 从peft加载LoraConfig报错
Issue -
State: closed - Opened by xiaohengDa over 1 year ago
- 3 comments
Labels: bug
#32 - 对pretraining阶段的数据加工有点疑问
Issue -
State: closed - Opened by hongyix over 1 year ago
- 1 comment
Labels: question
#31 - run_sft.sh 的 eval_loss一直不降 都是同一个值
Issue -
State: closed - Opened by xuanxixi over 1 year ago
- 1 comment
Labels: bug
#30 - gradio_demo.py
Issue -
State: closed - Opened by gaojing8500 over 1 year ago
- 1 comment
Labels: bug
#29 - ziya-llama-13b-medical-lora 量化推理怎么使用？
Issue -
State: closed - Opened by Nisoka over 1 year ago
- 7 comments
Labels: question
#28 - chatglm现在的reward model模型缺失吗？
Issue -
State: closed - Opened by ymyjl over 1 year ago
- 4 comments
Labels: enhancement
#27 - reward_baseline
Issue -
State: closed - Opened by yangliuIOC over 1 year ago
- 1 comment
Labels: question
#26 - 关于原始百川的infer
Issue -
State: closed - Opened by nuoma over 1 year ago
- 5 comments
Labels: question
#25 - 训练进程卡住
Issue -
State: closed - Opened by aichifandefan over 1 year ago
- 1 comment
Labels: enhancement
#23 - 运行run_rm.sh报错 RuntimeError: CUDA error: device-side assert triggered
Issue -
State: closed - Opened by Candy555 over 1 year ago
- 1 comment
Labels: bug
#22 - 最后两步的一些疑问
Issue -
State: closed - Opened by skepsun over 1 year ago
- 1 comment
Labels: bug
#21 - 基于chatglm训练reward model是使用AutoModelForSequenceClassification加载模型吗
Issue -
State: closed - Opened by wang9702 over 1 year ago
- 4 comments
Labels: bug
#20 - rl_training和reward_modeling中Tokenizer新增Pad token，Model不需要resize吗？
Issue -
State: closed - Opened by baibaiw5 over 1 year ago
- 1 comment
Labels: question
#19 - 使用merge_peft_adapter.py进行merge的时候，词表映射出现了问题
Issue -
State: closed - Opened by nlper-hou over 1 year ago
- 3 comments
Labels: question
#18 - RLHH
Issue -
State: closed - Opened by yangliuIOC over 1 year ago
- 3 comments
Labels: question
#17 - 使用run_pt.sh对llama-13B在医疗数据上增量训练，跑了5个epoch，可是loss不下降是怎么回事？一直是10.25附近波动
Issue -
State: closed - Opened by nlper-hou over 1 year ago
- 4 comments
Labels: question
#16 - 加载bloom 13B模型报错
Issue -
State: closed - Opened by 1615070057 over 1 year ago
- 1 comment
Labels: bug
#15 - 单机多卡运行卡死
Issue -
State: closed - Opened by zhangxinxin0428 over 1 year ago
- 8 comments
Labels: question
#14 - 关于预训练完成后合并模型及SFT的问题
Issue -
State: closed - Opened by charryshi over 1 year ago
- 10 comments
Labels: question
#13 - supervised_finetuning.py的preprocess_function函数是否有问题
Issue -
State: closed - Opened by baibaiw5 over 1 year ago
- 1 comment
Labels: question
#12 - loss is 0 when turn off use_peft
Issue -
State: closed - Opened by baibaiw5 over 1 year ago
- 2 comments
Labels: question
#11 - ValueError: 130004 is not in list
Issue -
State: closed - Opened by sexan over 1 year ago
- 16 comments
Labels: question
#10 - 跑增量预训练是中断后恢复不能继续
Issue -
State: closed - Opened by charryshi over 1 year ago
- 7 comments
Labels: question
#9 - 为啥这个代码里都是bug 跑chatglm的微调代码根本跑不起来
Issue -
State: closed - Opened by yankuo111 over 1 year ago
- 1 comment
Labels: bug
#8 - 请问第一阶段的增量预训练需要的显存大小
Issue -
State: closed - Opened by charryshi over 1 year ago
- 1 comment
Labels: question
#7 - 如何测试reward model
Issue -
State: closed - Opened by lianzhaoy over 1 year ago
- 5 comments
Labels: question
#6 - 单机多卡跑gradio推理时，报CUDA的错误
Issue -
State: closed - Opened by chelovek21 over 1 year ago
Labels: question
#5 - 直接运行时会出现tokenizer长度错误
Issue -
State: closed - Opened by chelovek21 over 1 year ago
- 4 comments
Labels: question
#4 - 单机多卡预训练ChatGLM报错：
Issue -
State: closed - Opened by zzzhaoguziji over 1 year ago
- 8 comments
Labels: question
#3 - DistributedDataParallel device_ids and output_device arguments only work with single-device/multiple-device GPU modules or CPU modules, but got device_ids [0], output_device 0, and module parameters {device(type='cuda', index=0), device(type='cuda', index=1), device(type='cuda', index=2), device(type='cuda', index=3)}.
Issue -
State: closed - Opened by gaojing8500 over 1 year ago
- 1 comment
Labels: question
#2 - 直接运行run_rm.sh，产生关于计算图的RuntimeError
Issue -
State: closed - Opened by zhpmatrix over 1 year ago
- 4 comments
Labels: question
#1 - 请问增量预训练大概需要几块GPU呢？
Issue -
State: closed - Opened by jason7323 over 1 year ago
- 3 comments
Labels: question