google-deepmind / google-deepmind/synthid-text
runtimeerror
- Dominant language
- Python
- Stars
- 1.1k
- Forks
- 104
- PR merge metrics
- No merged PRs in 30d
Description
1、It is so weird .When i run the "generate watermark output",i meet some problems as follows:
====================>>>>>
/home/comsec/.venvs/synthid/lib/python3.10/site-packages/torch/_functorch/vmap.py:480: UserWarning: There is a performance drop because we have not yet implemented the batching rule for aten::take. Please file us an issue on GitHub so that we can prioritize its implementation. (Triggered internally at ../aten/src/ATen/functorch/BatchedFallback.cpp:81.)
batched_outputs = func(*batched_inputs, **kwargs)
Output:
----------------------------------------------------------------------------------------------------
I enjoy walking with my cute dog as much as I love walking with my dog," she said. "I've done a lot of work for her and she's been incredibly happy and understanding."
As for what she has learned about the canine relationship between the two people?
"I've learned how to be a good companion," she said. "I've learned how to be a good owner. I've learned how to be a good dog trainer."
Read the full story at the Sun Herald.
*****************************************************************************************************************
2、When i run the "Load Eli5 dataset with HuggingFace datasets." part,another problem occurred ( under gpt2 ):
=================>>>>>"Repo card metadata block was not found. Setting CardData to empty."
It seems that i can't download the dataset.But i don't know why.Is it because the memory of my gpu too small? I didn't see the requirements of gpu memory of gpt2 in your notebook.
Contributor guide
Research direction
Start by reproducing both notebook steps: “generate watermark output” and “Load Eli5 dataset with HuggingFace datasets” under GPT-2. Inspect whether the torch warning affects output and whether the dataset metadata message prevents downloading; done means identifying the cause and documenting the required environment or a reproducible fix.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- huggingface, python, pytorch
- Domain
- data, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100