ollama run codellama:34b issue

[Iliceth]

What is the issue?

Every model I tested with ollama runs fine, but when trying: ollama run codellama:34b, I get Error: llama runner process has terminated: signal: segmentation fault (core dumped)Tried the 13B-version then, works fine.

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.1.37

[izzy84]

same here

[jmorganca]

So sorry about this, looking into why this is happening.

[jonz-secops]

same

[barrard]

I'm also getting this on 7b-python and 13b-python

[dhiltgen]

Could you share your server log, and your VRAM size? This is a larger model, so my suspicion is we're off on our memory prediction.

[9SMTM6]

Same here.

AMD CPU in contrast to OP.

16 GB VRAM (RTX 4070 Ti Super)

Log attached.
codellama_crashdump_shortend.log

[jonz-secops]

To add details ..

AMD 5950X on x570E
32 GB CPU RAM

6900XT
16GB VRAM

server log; doesn't appear to have the events in it anymore, but I'll try and recreate later today if I can.

[Iliceth]

Hereby my log.

ollama_logs.txt

I can run mixtral and dolphin-mixtral fine and those are some big ones too.

Anyhow my current system is i7-9700K + 32 GB RAM + RTX 3090 with 24 GB VRAM

[izzy84]

I7 7700K with 32gb ram and 6 GPUs with in total 51gb vram.

I can run all models, even llama3:70b without problems, which is already a big model. But I get error with codellama 34b

[FukangSun]

same here .I use 0.1.38 version of official docker image.
E5-2670 + 96gb ram + 1 RTX 4090 . also get below error when running codellama 34b :

ollama run codellama-34b-instruct:Q4_K_M
Error: llama runner process has terminated: signal: segmentation fault (core dumped)

Gentleman, I find another way, downgrade to 0.1.32 is okay to load codellama and large model. 0.1.33 - 0.1.39 I tested all not worked.

[AZ777xx]

same. Phind-codellama from the repository works, loading gguf doesn't though