anon8231489123

/

gpt4-x-alpaca-13b-native-4bit-128g

like 734

same issue. still trying to figure it out.

+1 I have the same issue

Yep, me too.

use 'GIT_LFS_SKIP_SMUDGE=1' before the git clone command to clone all the files except the large files.

example:
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g

Manually download the file that you need using wget or clicking the file link and pressing download button provided by HF on the HF page

examaple:
wget https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g/blob/main/gpt4-x-alpaca-13b-ggml-q4_1-from-gptq-4bit-128g/ggml-model-q4_1.bin

if you are going to use llama.cpp and run it CPU only, then you need the ggml model https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g/blob/main/gpt4-x-alpaca-13b-ggml-q4_1-from-gptq-4bit-128g/ggml-model-q4_1.bin

if you are going to use OodaBooga with GPU, then download the Cuda GPTQ
https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g/blob/main/gpt-x-alpaca-13b-native-4bit-128g-cuda.pt

Using GIT_LFS_SKIP_SMUDGE=1 Gives this error: is not recognized as an internal or external command, operable program, or batch file

Got the same response as above when trying to use it

The download is probably still going in the background. It just takes ages and there is no visual feedback like a progress bar.

However, Git has seems to have some bugs on Windows when downloading files >4GB and will probably just save a truncated file anyway, so better set the corresponding environment variable to skip huge files (temporary, only applies to the current CMD session) before doing the clone and download + overwrite them manually:

git lfs install
set GIT_LFS_SKIP_SMUDGE=1
git clone https://huggingface.co/anon8231489123/gpt4-x-alpaca-13b-native-4bit-128g

The clone command should only take a few seconds and create 1KB placeholder files for the huge ones (.pt, .safetensors, .bin etc).
Now download that big checkpoint file manually and OVERWRITE the placeholder.

When in doubt, "git status" should tell you which files are affected.
(You might need to "cd" to the new directory first: "cd gpt4-x-alpaca-13b-native-4bit-128g")

I had the same issue, but checking in .git/lfs/incomplete I can see those large files there, so it's actually just downloading and not giving visual feedback.

I just left it to finish.

I am having this issue. As per the responses I added the flag before I cloned the repository. (to avoid it downloading the 10GB & 5GB files I have already downloaded)

Upon running the clone command again it gets to 85% on filtering, my network usage goes through the roof & now I guess I just have to wait out the unknown time.

Why didn't the flag work?

No idea when it's going to finish so time to go to bed & hope it's done when I wake up.

+1. would be nice to see the status in the console. For now, I open up Activity Monitor, go to Network tab, and see that git-lfs is there and that it is receiving.

I think it is downloading in the background even though it is not logging it-
I see my network consumption up accordingly even though I am "stuck" on filtering content step of clone.

Patience will probably be the winner

I think it is downloading in the background even though it is not logging it-
I see my network consumption up accordingly even though I am "stuck" on filtering content step of clone.

Patience will probably be the winner

how long did it take to downlaod