apunkt

You can force GPU compilation from source be editing /gpu/cpu_common.go line 20: -> '''return "avx"''' then compile with custom options: OLLAMA_CUSTOM_CPU_DEFS="-DLLAMA_CUDA=on -DCMAKE_CUDA_ARCHITECTURES=all-major -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_F16C=off -DLLAMA_FMA=off" it still will...

@dhiltgen did further testing. It works for me with 1 GPU: loading bigger models, that don't fit into VRAM also works, load is splitted between GPU/CPU It causes problems for...

> Extension host was not restarting with indexing disabled on 0.8.44, but it's now restarting again after I updated to 0.8.45. Confirmed. Was working on my end with indexing **en**abled...

Having the same issue on `Ubuntu 24.04` w/ `Gluster 11.1` w/ config `Number of Bricks: 6 x (2 + 1) = 18` running borg Previously was using latest Gluster 9.x...

Adding a 4th card causes size estimation for same model with same settings to be different: 3 Cards: | NAME |ID |SIZE|PROCESSOR | UNTIL | --- | --- | ---...

Tested v0.9.0 native install in Ubuntu 24.04 with qwen3:30b-a3b on Tesla M40: old-engine: 18.15 token/s new-engine: 18.65 token/s

I copy that. Same error. Same effects. Trying to connect Gluster 9.6 on Ubuntu focal to Gluster 11.1 on Ubuntu noble Volumes replicate and distributed-replicate.

DON'T upgrade your cluster to 11.1 trying to join a new peer! I upgraded 9.6 -> 10.5 flawlessley, but when upgrading the first node 10.5 -> 11.1 it is not...

> DON'T upgrade your cluster to 11.1 trying to join a new peer! > > I upgraded 9.6 -> 10.5 flawlessley, but when upgrading the first node 10.5 -> 11.1...

Finally you are able to expand the cluster with a new 11.1 node. Peer probe is now successful.