github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #877] Can't import model: invalid version #62460

New Issue
Closed
opened 2026-05-03 09:02:45 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-05-03 09:02:45 -05:00
Owner
Copy Link

Originally created by @olafgeibig on GitHub (Oct 23, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/877

Problem

I tried to import agentlm-7b.Q5_K_M.gguf from TheBloke but it fails. I can run the same file successfully with llama.cpp

Command line

ollama create agentlm:7b-Q5_K_M -f agentlm.ollama

Output

parsing modelfile
looking for model
⠋ creating model layer  Error: invalid version

Model file

FROM ./agentlm-7b.Q5_K_M.gguf
TEMPLATE """
[INST] <<SYS>>
{{ .System }}
<</SYS>>
{{ .Prompt }} [/INST]
"""

Environment

MacBook Pro M1 16GB, MacOS 13.3.1
Ollama 0.1.3

Originally created by @olafgeibig on GitHub (Oct 23, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/877 ## Problem I tried to import [agentlm-7b.Q5_K_M.gguf](https://huggingface.co/TheBloke/agentlm-7B-GGUF/blob/main/agentlm-7b.Q5_K_M.gguf) from TheBloke but it fails. I can run the same file successfully with llama.cpp ### Command line ```ollama create agentlm:7b-Q5_K_M -f agentlm.ollama``` ### Output ``` parsing modelfile looking for model ⠋ creating model layer Error: invalid version ``` ## Model file ``` FROM ./agentlm-7b.Q5_K_M.gguf TEMPLATE """ [INST] <<SYS>> {{ .System }} <</SYS>> {{ .Prompt }} [/INST] """ ``` ## Environment MacBook Pro M1 16GB, MacOS 13.3.1 Ollama 0.1.3
GiteaMirror added the bug label 2026-05-03 09:02:46 -05:00
GiteaMirror commented 2026-05-03 09:02:55 -05:00
Author
Owner
Copy Link

@deichbewohner commented on GitHub (Oct 23, 2023):

I have observed that some of the newer models from TheBloke are employing the GGUF container specification version 3.

I executed a fmt.Println(c.Version) at this location: /llm/gguf.go#L31.
This yielded a 3 in the log.

Nonetheless, the referenced code currently only accommodates GGUF version 1 or 2.

Edit:

<!-- gh-comment-id:1775373544 --> @deichbewohner commented on GitHub (Oct 23, 2023): I have observed that some of the newer models from TheBloke are employing the GGUF container specification version 3. I executed a `fmt.Println(c.Version)` at this location: [/llm/gguf.go#L31](https://github.com/jmorganca/ollama/blob/ccff9ca09c1fb6709ba4456f0285b97bd5db9599/llm/gguf.go#L31). This yielded a `3` in the log. Nonetheless, the [referenced code](https://github.com/jmorganca/ollama/blob/ccff9ca09c1fb6709ba4456f0285b97bd5db9599/llm/gguf.go#L32) currently only accommodates GGUF version 1 or 2. Edit: - The relevant commit in llama.cpp should be this: https://github.com/ggerganov/llama.cpp/commit/8cf19d60dc93809db8e51fedc811595eed9134c5#diff-cbcdffc2972ef722e13636e19e95da7648bd8d5be17535e013ca8edf77200228 - You can check the model version with `xxd -l 5 -p agentlm-13b.Q5_K_M.gguf` (the last byte is the version number)
GiteaMirror commented 2026-05-03 09:02:57 -05:00
Author
Owner
Copy Link

@deichbewohner commented on GitHub (Oct 23, 2023):

Upon reading the GGUF changes in the llama.cpp commit mentioned in my previous comment, it appears that there are no format changes, at least on a LittleEndian machine. I have made a temporary modification to the Ollama code here to accommodate the new version as follows:

switch c.Version {
    case 1:
        binary.Read(r, binary.LittleEndian, &c.V1)
    case 2, 3:
        binary.Read(r, binary.LittleEndian, &c.V2)
}

With this adjustment, the new models function correctly. However, this is clearly a workaround and not a permanent solution. A more robust implementation should be considered, especially if there are updates to the llama.cpp submodule in the future.

<!-- gh-comment-id:1775470728 --> @deichbewohner commented on GitHub (Oct 23, 2023): Upon reading the GGUF changes in the llama.cpp commit mentioned in my previous comment, it appears that there are no format changes, at least on a LittleEndian machine. I have made a temporary modification to the Ollama code [here](https://github.com/jmorganca/ollama/blob/ccff9ca09c1fb6709ba4456f0285b97bd5db9599/llm/gguf.go#L32) to accommodate the new version as follows: ```go switch c.Version { case 1: binary.Read(r, binary.LittleEndian, &c.V1) case 2, 3: binary.Read(r, binary.LittleEndian, &c.V2) } ``` With this adjustment, the new models function correctly. However, this is clearly a workaround and not a permanent solution. A more robust implementation should be considered, especially if there are updates to the llama.cpp submodule in the future.
GiteaMirror commented 2026-05-03 09:02:57 -05:00
Author
Owner
Copy Link

@olafgeibig commented on GitHub (Oct 23, 2023):

Thank you for picking this up. xxd -l 5 -p agentlm-7b.Q5_K_M.gguf returns 4747554603

<!-- gh-comment-id:1775528561 --> @olafgeibig commented on GitHub (Oct 23, 2023): Thank you for picking this up. ```xxd -l 5 -p agentlm-7b.Q5_K_M.gguf``` returns 4747554603
GiteaMirror commented 2026-05-03 09:02:58 -05:00
Author
Owner
Copy Link

@deichbewohner commented on GitHub (Oct 23, 2023):

This Issue is now addressed and worked on in https://github.com/jmorganca/ollama/pull/881

<!-- gh-comment-id:1775686810 --> @deichbewohner commented on GitHub (Oct 23, 2023): This Issue is now addressed and worked on in https://github.com/jmorganca/ollama/pull/881
GiteaMirror commented 2026-05-03 09:03:00 -05:00
Author
Owner
Copy Link

@mxyng commented on GitHub (Oct 23, 2023):

As @deichbewohner mentioned, this is fixed in #881 and will be in the next reelase

<!-- gh-comment-id:1775715866 --> @mxyng commented on GitHub (Oct 23, 2023): As @deichbewohner mentioned, this is fixed in #881 and will be in the next reelase
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#62460
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?