github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[PR #7844] [CLOSED] Handle padding bytes in GGUF files. #23075

New Issue
Closed
opened 2026-04-19 16:45:07 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-19 16:45:07 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/7844
Author: @rick-github
Created: 11/26/2024
Status: ❌ Closed

Base: main ← Head: magic

📝 Commits (1)

  • d935ee4 Handle padding bytes in GGUF files.

📊 Changes

3 files changed (+20 additions, -1 deletions)

View changed files

📝 llm/ggml.go (+7 -0)
📝 llm/gguf.go (+8 -0)
📝 server/model.go (+5 -1)

📄 Description

GGUF files are padded to an alignment value, by default 32 bytes. parseFromFile() doesn't handle this, which causes an invalid file magic for some imports since it tries to read multiple GGML structures from a single file and stumbles over the padding bytes when the last tensor didn't end on the alignment mark.

The doc is not as clear on adapters and projectors, but the fine-tuning code in llama.cpp uses an alignment of 32 and all the projectors I checked were aligned to 32 bytes.

Fixes: https://github.com/ollama/ollama/issues/5939
Fixes: https://github.com/ollama/ollama/issues/6272

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/ollama/ollama/pull/7844 **Author:** [@rick-github](https://github.com/rick-github) **Created:** 11/26/2024 **Status:** ❌ Closed **Base:** `main` ← **Head:** `magic` --- ### 📝 Commits (1) - [`d935ee4`](https://github.com/ollama/ollama/commit/d935ee46d97d3d4cd7ed033b64fe19c6aedb5889) Handle padding bytes in GGUF files. ### 📊 Changes **3 files changed** (+20 additions, -1 deletions) <details> <summary>View changed files</summary> 📝 `llm/ggml.go` (+7 -0) 📝 `llm/gguf.go` (+8 -0) 📝 `server/model.go` (+5 -1) </details> ### 📄 Description GGUF files are [padded](https://github.com/ggerganov/ggml/blob/master/docs/gguf.md#file-structure:~:text=Where%20required%2C%20the%20file%20is%20padded%20with%200x00%20bytes%20to%20the%20next%20multiple%20of%20general.alignment) to an alignment value, by default 32 bytes. `parseFromFile()` doesn't handle this, which causes an `invalid file magic` for some imports since it tries to read multiple GGML structures from a single file and stumbles over the padding bytes when the last tensor didn't end on the alignment mark. The doc is not as clear on adapters and projectors, but the fine-tuning code in llama.cpp uses an alignment of 32 and all the projectors I checked were aligned to 32 bytes. Fixes: https://github.com/ollama/ollama/issues/5939 Fixes: https://github.com/ollama/ollama/issues/6272 --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-04-19 16:45:07 -05:00
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 pull-request
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#23075
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?