-
### 是否已有关于该错误的issue或讨论? | Is there an existing issue / discussion for this?
- [X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions
### 该问题是否在FAQ中有解答? | Is there an existing…
-
配置:4090 ,24G显存
加载方式:
self.model = AutoModelForCausalLM.from_pretrained(model_path,torch_dtype="auto",device_map="auto").eval()
self.tokenizer = AutoTokenizer.from_pretrained(model_path)
…
-
### 是否已有关于该错误的issue或讨论? | Is there an existing issue / discussion for this?
- [X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions
### 该问题是否在FAQ中有解答? | Is there an existing ans…
-
### Reminder
- [X] I have read the README and searched the existing issues.
### Reproduction
背景:使用模型Qwen1.5-0.5B并在此模型基础上作微调训练。
数据集:采用https://huggingface.co/datasets/michaelwzhu/ShenNong_TCM_Datase…
-
用qwen1.5-7b采用swift官网的自我认知微调后,开始使用CUDA_VISIBLE_DEVICES=0 swift export --ckpt_dir xxx --merge_lora true
进行权重合并,结果合并后模型变成了二十几个GB的大小,比原模型大出一倍?
是因为我在lora微调时dtype是auto自动选择fp32而不是fp16的原因吗?
-
作者你好,我在复现bunny_v1_0-2B-zh的过程中,pretrain阶段训出来保存的config.json中model_type是bunny-qwen,而开源中的config.json的model_type是bunny-qwen2,这个是什么原因?另外请问你在huggingface中开源的configuration_bunny_qwen2.py和modeling_bunny_qwen2.p…
-
### Prerequisite
- [X] I have searched [Issues](https://github.com/open-compass/opencompass/issues/) and [Discussions](https://github.com/open-compass/opencompass/discussions) but cannot get the expe…
-
when could openvino convert MoE model, like Qwen1.5-MoE-2.7B ?
-
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
> swift infer \
> --model_id_or_path /data/LLM_model/Qwen1.5-72B-Chat \
> --use_flash_attn true \
> --model_cache_dir /data/LLM_model/Qwen1.5-72…
-
大佬们,请教一下,Qwen1.5-14B-Chat-GPTQ-Int4能流式回复吗?提示我没有stream这个参数