Search results

Problems in train_deepspeed.py with ZeRO stage 1|2|3

Status: Open.

#34 In yanqiangmiffy/InstructGLM;

· zjhJOJO opened on Sep 7, 2023

用tokenizer_dataset_rows.py转换自己的数据报错datasets.builder.datasetgeneraationerror

Status: Open.

#33 In yanqiangmiffy/InstructGLM;

· cat1222 opened on May 9, 2023

使用lora训练，使用web_demo加载lora权重后，结果跟原生chatglm结果一样，lora权重没生效，这个是什么原因呢

Status: Open.

#32 In yanqiangmiffy/InstructGLM;

· AnddyWang opened on May 6, 2023

用BelleGroup/train_1M_CN训练后，为什么用数据集里的问题测，回答不一样

Status: Open.

#31 In yanqiangmiffy/InstructGLM;

· czhcc opened on May 5, 2023

调教后的逻辑能力如何？

Status: Open.

#30 In yanqiangmiffy/InstructGLM;

· daiaji opened on Apr 29, 2023

train_lora最低需要多大显存GPU可以训练？除了batch size 还有别的参数可以降低显存使用吗？

Status: Open.

#29 In yanqiangmiffy/InstructGLM;

· twosnowman opened on Apr 27, 2023

用peft加载lora后，generate时报错ValueError: 130000 is not in list，加载lora之前推理是正常的

Status: Open.

#28 In yanqiangmiffy/InstructGLM;

· BIGPPWONG opened on Apr 25, 2023

请问下train_deepspeed.py 怎么引入lora.pt

Status: Open.

#27 In yanqiangmiffy/InstructGLM;

· AlexXx-Wu opened on Apr 20, 2023

ValueError: ChatGLMForConditionalGeneration does not support gradient checkpointing.

Status: Open.

#25 In yanqiangmiffy/InstructGLM;

· deepeye opened on Apr 19, 2023

datasets.builder.InvalidConfigName: Bad characters from black list '<>:/\|?*' found in 'data/belle_data.json'. They could create issues when creating a directory for this config on Windows filesystem.

Status: Open.

#23 In yanqiangmiffy/InstructGLM;

· deepeye opened on Apr 19, 2023

4张32G的可以吗，作者可以用你写的其他开源数据集finetune看看效果吗，再放出转换和训练代码

Status: Open.

#21 In yanqiangmiffy/InstructGLM;

· hangzeli08 opened on Apr 15, 2023

4张 12G的 3060能训练吗

Status: Open.

#20 In yanqiangmiffy/InstructGLM;

· zlszhonglongshen opened on Apr 14, 2023