OpenDelta
OpenDelta copied to clipboard
thunlp
A plug-and-play library for parameter-efficient-tuning (Delta Tuning)
在官方文档中提到，在训练时不需要改变优化器，我不是很理解，意思是原来就存在一个优化器吗，如果是的话我应该怎么获取它 
I don't understand why the reasoning effect of reloading the trained model is very poor. Did I write something wrong? Looking forward to your reply. the key code is as...
使用opendelta来微调cpmbee的10b后，使用加载lora的方式进行推理（如下所示）和原本进行推理速度相比会变慢（减慢50%），请问如何解决。 是否可以将lora与原权重进行合并。 ``` tokenizer = CPMBeeTokenizer() model = CPMBeeTorch(config=config) delta_model = LoraModel(backbone_model=model, modified_modules=["project_q", "project_v"], backend="hf") model.load_state_dict(torch.load(args.delta), strict=False) model.load_state_dict(torch.load(ckpt_path), strict=False) ```
I would like to request some clarification on the differences between OpenDelta and Peft. From my understanding, both OpenDelta and Peft have the ability to control certain parameters during training....
从报错信息看，似乎是要在setup.py中更新一下，“scikit-learn” ------ error: subprocess-exited-with-error × python setup.py egg_info did not run successfully. │ exit code: 1 ╰─> [18 lines of output] The 'sklearn' PyPI package is deprecated, use 'scikit-learn' rather...
``` HTTPError Traceback (most recent call last) [](https://localhost:8080/#) in () 12 from opendelta import AutoDeltaModel, AutoDeltaConfig 13 # use existing delta models from DeltaCenter ---> 14 delta = AutoDeltaModel.from_finetuned("thunlp/Spelling_Correction_T5_LRAdapter_demo", backbone_model=t5)...
