Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Download ggml-alpaca-7b-q4.bin failed CHECKSUM #410

Closed
kechan opened this issue Mar 22, 2023 · 15 comments
Closed

Download ggml-alpaca-7b-q4.bin failed CHECKSUM #410

kechan opened this issue Mar 22, 2023 · 15 comments
Labels
model Model specific

Comments

@kechan
Copy link

kechan commented Mar 22, 2023

This may well be the end server issue. I tried several times with no luck, just wonder if people have seen this.
I tried all 3 curl commands.

@kechan
Copy link
Author

kechan commented Mar 22, 2023

Never mind... I retried and it completed with 4212855021 bytes downloaded. If others have done this, are the count correct?
I tried checksum but:

models/alpaca-7B/ggml-model-q4_0.bin: FAILED
shasum: WARNING: 1 computed checksum did NOT match
shasum: SHA256SUMS: no file was verified

@zbruceli
Copy link

Your byte count is same as mine.

% md5 -r ggml-alpaca-7b-q4.bin
46ac645818cf604afdddcf0f9e5a5d44 ggml-alpaca-7b-q4.bin

@kechan
Copy link
Author

kechan commented Mar 22, 2023

@zbruceli
I ran checksum & failed, this is the item:

1f582babc2bd56bb63b33141898748657d369fd110c4358b2bc280907882bf13 models/alpaca-7B/ggml-model-q4_0.bin

@Ameobea
Copy link
Contributor

Ameobea commented Mar 22, 2023

I ran into this with the Cloudflare ipfs gateway, but the ipfs.io one worked for me.

@kechan kechan changed the title curl: (18) transfer closed error during ggml-alpaca-7b-q4.bin download Download ggml-alpaca-7b-q4.bin failed CHECKSUM Mar 22, 2023
@kechan
Copy link
Author

kechan commented Mar 22, 2023

@Ameobea Yes, I tried ipfs.io and that worked finally for me too. But unfortunately, it failed checksum. The checksum file has a different filename (and under dir models/alpaca-7B), I just create that dir and rename the file, and ran:

shasum -a 256 --ignore-missing -c SHA256SUMS

Got:

models/alpaca-7B/ggml-model-q4_0.bin: FAILED
shasum: WARNING: 1 computed checksum did NOT match
shasum: SHA256SUMS: no file was verified

@Ameobea
Copy link
Contributor

Ameobea commented Mar 22, 2023

@kechan you can also try this torrent. It's how I obtained my alpaca 7B model, and it works well for me

@kechan
Copy link
Author

kechan commented Mar 22, 2023

@Ameobea I am a bit less comfortable with torrent though... did it pass any checksum for you? I just want to be a bit careful.

@Ameobea
Copy link
Contributor

Ameobea commented Mar 22, 2023

I've not run the checksum

@Ameobea
Copy link
Contributor

Ameobea commented Mar 22, 2023

Oh actually I just looked at those checksums, I think they're out of date.

There was a PR merged (#352) to update IPFS links to the new model format. That's ggml2- file names.

The SHASUMS don't seem to have been updated as a part of that, and still reference the old model filenames (ggml- not ggml2-).

@kechan
Copy link
Author

kechan commented Mar 22, 2023

@Ameobea
Ok. I downloaded using this curl:

curl -o ./models/ggml-alpaca-7b-q4.bin -C - https://ipfs.io/ipfs/QmUp1UGeQFDqJKvtjbSYPBiZZKRjLp8shVP9hT8ZB9Ynv1

I think "QmUp...v1" is the latest greatest? If so, can you please point out the right checksum I should use?

@Ameobea
Copy link
Contributor

Ameobea commented Mar 22, 2023

Here's the hash I get for the file I downloaded via torrent:

8d5562ec1d8a7cfdcf8985a9ddf353339d942c7cf52855a92c9ff59f03b541bc ggml2-alpaca-7b-q4.bin

That's all I have

@gjmulder gjmulder added the model Model specific label Mar 22, 2023
@zbruceli
Copy link

The ipfs seed represents the file, so you can use alternative ipfs gateways for faster download speed and less stalls. The examples in README gave three common ipfs gateways, but there are many others. Do a bit research and you will find the others. :-)

@kechan
Copy link
Author

kechan commented Mar 22, 2023

@Ameobea
Mine is 5992013124b645bfb6e80a2bf155058f3e20351115dbfbedf2f30a017d7c69af

I think my copy either incomplete, corrupted, or tampered with.

@kechan
Copy link
Author

kechan commented Mar 23, 2023

@zbruceli

I tried gateway.estuary.tech and got md5 that matches yours

46ac645818cf604afdddcf0f9e5a5d44

What's odd is that while the byte count is the same as from ipfs.io, the md5 is different:

3e47e543e3cb3dd3f3a1db20c9ecc920

I don't know what to make of this. Since if yours work, I probably will try to use the one with "46ac...."

@anzz1
Copy link
Contributor

anzz1 commented Mar 23, 2023

This issue is related to #382

I'm closing this as a duplicate and let's continue the discussion in #382 .

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
model Model specific
Projects
None yet
Development

No branches or pull requests

5 participants