github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 21:25:46 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #9820] model create fails in AIX #52941

New Issue
Closed
opened 2026-04-29 01:25:45 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-29 01:25:45 -05:00
Owner
Copy Link

Originally created by @ayappanec on GitHub (Mar 17, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9820

What is the issue?

I am running ollama-0.5.1 in AIX. While trying to create a model , it failed with the below error.

$ cat Modelfile
FROM /gguf_model/Llama-3.2-3B-Instruct-uncensored-f16.gguf

$ /ollama-0.5.1/ollama create my_model
transferring model data 100%
Error: Post "http://127.0.0.1:11434/api/create": EOF

From the Ollama server.,

[GIN] 2025/03/17 - 08:08:10 | 200 | 19µs | 127.0.0.1 | HEAD "/"
[GIN] 2025/03/17 - 08:09:20 | 201 | 44.290868s | 127.0.0.1 | POST "/api/blobs/sha256:5e9d2be67300511a12ea4d10348c23f6c630f02bbaf5242d42f446cdabbbcd7c"
panic: runtime error: makeslice: len out of range

goroutine 16 [running]:
github.com/ollama/ollama/llm.readGGUFString(0xa00010000452808, {0xa00000046c5f9c0, 0xa000100008fa258})
/ollama-0.5.1/llm/gguf.go:333 +0x118
github.com/ollama/ollama/llm.(*gguf).Decode(0xa00010000452808, {0x1101dbe60, 0xa000100008fa258})
/ollama-0.5.1/llm/gguf.go:145 +0xf0
github.com/ollama/ollama/llm.(*containerGGUF).Decode(0xa000100008b2140, {0x1101dbe60, 0xa000100008fa258})
/ollama-0.5.1/llm/gguf.go:67 +0x21c
github.com/ollama/ollama/llm.DecodeGGML({0x1101db1d8, 0xa0001000011a2a8}, 0x10?)
/ollama-0.5.1/llm/ggml.go:346 +0x458
github.com/ollama/ollama/server.parseFromFile({0x1101dfc88, 0xa000100008b20a0}, {0x100e953d8, 0x5}, {0x0, 0x0, 0x0}, 0xa0001000011a2a8, {0xa00010000644321, 0x47}, ...)
/ollama-0.5.1/server/model.go:187 +0x334
github.com/ollama/ollama/server.CreateModel({0x1101dfc88, 0xa000100008b20a0}, {{0x100eaa5c3, 0x12}, {0x100e9e0a1, 0x7}, {0xa00010000412510, 0x8}, {0x100e9a5cc, 0x6}}, ...)
/ollama-0.5.1/server/images.go:412 +0x714
github.com/ollama/ollama/server.(*Server).CreateHandler.func1()
/ollama-0.5.1/server/routes.go:695 +0x224
created by github.com/ollama/ollama/server.(*Server).CreateHandler in goroutine 72
/ollama-0.5.1/server/routes.go:685 +0x964

The model "Llama-3.2-3B-Instruct-uncensored-f16.gguf" is already converted to Big Endian (AIX is Big Endian) by making use of the script "gguf_convert_endian.py" from llama.cpp.

OS

AIX

GPU

No GPU

CPU

Power9

Ollama version

0.5.1

Originally created by @ayappanec on GitHub (Mar 17, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9820 ### What is the issue? I am running ollama-0.5.1 in AIX. While trying to create a model , it failed with the below error. $ cat Modelfile FROM /gguf_model/Llama-3.2-3B-Instruct-uncensored-f16.gguf $ /ollama-0.5.1/ollama create my_model transferring model data 100% Error: Post "http://127.0.0.1:11434/api/create": EOF From the Ollama server., [GIN] 2025/03/17 - 08:08:10 | 200 | 19µs | 127.0.0.1 | HEAD "/" [GIN] 2025/03/17 - 08:09:20 | 201 | 44.290868s | 127.0.0.1 | POST "/api/blobs/sha256:5e9d2be67300511a12ea4d10348c23f6c630f02bbaf5242d42f446cdabbbcd7c" panic: runtime error: makeslice: len out of range goroutine 16 [running]: github.com/ollama/ollama/llm.readGGUFString(0xa00010000452808, {0xa00000046c5f9c0, 0xa000100008fa258}) /ollama-0.5.1/llm/gguf.go:333 +0x118 github.com/ollama/ollama/llm.(*gguf).Decode(0xa00010000452808, {0x1101dbe60, 0xa000100008fa258}) /ollama-0.5.1/llm/gguf.go:145 +0xf0 github.com/ollama/ollama/llm.(*containerGGUF).Decode(0xa000100008b2140, {0x1101dbe60, 0xa000100008fa258}) /ollama-0.5.1/llm/gguf.go:67 +0x21c github.com/ollama/ollama/llm.DecodeGGML({0x1101db1d8, 0xa0001000011a2a8}, 0x10?) /ollama-0.5.1/llm/ggml.go:346 +0x458 github.com/ollama/ollama/server.parseFromFile({0x1101dfc88, 0xa000100008b20a0}, {0x100e953d8, 0x5}, {0x0, 0x0, 0x0}, 0xa0001000011a2a8, {0xa00010000644321, 0x47}, ...) /ollama-0.5.1/server/model.go:187 +0x334 github.com/ollama/ollama/server.CreateModel({0x1101dfc88, 0xa000100008b20a0}, {{0x100eaa5c3, 0x12}, {0x100e9e0a1, 0x7}, {0xa00010000412510, 0x8}, {0x100e9a5cc, 0x6}}, ...) /ollama-0.5.1/server/images.go:412 +0x714 github.com/ollama/ollama/server.(*Server).CreateHandler.func1() /ollama-0.5.1/server/routes.go:695 +0x224 created by github.com/ollama/ollama/server.(*Server).CreateHandler in goroutine 72 /ollama-0.5.1/server/routes.go:685 +0x964 The model "Llama-3.2-3B-Instruct-uncensored-f16.gguf" is already converted to Big Endian (AIX is Big Endian) by making use of the script "gguf_convert_endian.py" from llama.cpp. ### OS AIX ### GPU No GPU ### CPU Power9 ### Ollama version 0.5.1
GiteaMirror added the bug label 2026-04-29 01:25:45 -05:00
GiteaMirror commented 2026-04-29 01:25:47 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Mar 17, 2025):

Hey @ayappanec , thanks for the issue.

Unfortunately Ollama only supports little endian right now which is almost certainly why it blew up. There currently aren't plans to support big endian unless we had a lot of demand for it. It's a pretty major architectural change.

I'm going to close the issue as not planned, but we can revisit in the future.

<!-- gh-comment-id:2730386142 --> @pdevine commented on GitHub (Mar 17, 2025): Hey @ayappanec , thanks for the issue. Unfortunately Ollama *only* supports little endian right now which is almost certainly why it blew up. There currently aren't plans to support big endian unless we had a lot of demand for it. It's a pretty major architectural change. I'm going to close the issue as not planned, but we can revisit in the future.
GiteaMirror commented 2026-04-29 01:25:48 -05:00
Author
Owner
Copy Link

@ayappanec commented on GitHub (Apr 2, 2025):

Debugging this, I see in DecodeGGML there are switch statements to handle BigEndian as well. The magic constant for the Big Endian converted model is still coming as 0x46554747 (FILE_MAGIC_GGUF_LE). So when I changed the binary.Read to use BigEndian, then things start to working in AIX.

var magic uint32
if err := binary.Read(rs, binary.BigEndian, &magic); err != nil {
         return nil, 0, err
 }
<!-- gh-comment-id:2772565157 --> @ayappanec commented on GitHub (Apr 2, 2025): Debugging this, I see in DecodeGGML there are switch statements to handle BigEndian as well. The magic constant for the Big Endian converted model is still coming as 0x46554747 (FILE_MAGIC_GGUF_LE). So when I changed the `binary.Read` to use BigEndian, then things start to working in AIX. ``` var magic uint32 if err := binary.Read(rs, binary.BigEndian, &magic); err != nil { return nil, 0, err } ```
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#52941
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?