You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The latest docker pull for anythingllm on a CPU with AVX (No AVX2) seems to result in a core dump when a search is performed after which the container shuts down. Running the same built image on a CPU with AVX2 support doesn't exhibit the same behaviour.
I'm not sure which part of the stack is core dumping but the container logs shows.
[NativeEmbedder] Initialized
[NativeEmbedder] Initialized
[Event Logged] - update_embedding_engine
/usr/local/bin/docker-entrypoint.sh: line 7: 107 Illegal instruction (core dumped) node /app/server/index.js
If something is being compiled with the AVX2 flag enabled, would it be possible to provide a workaround for this on older CPUs with AVX support only?
Cheers.
Are there known steps to reproduce?
No response
The text was updated successfully, but these errors were encountered:
In fact, you just confirmed that hypothesis. The issue is that the underlying libraries do not support running ONNX on AVX arch CPUs. If you swap to another embedder model then you will not have this issue as you will not attempt to run anything via the ONNX runtime, which apparently does not support AVX?
How are you running AnythingLLM?
Docker (local)
What happened?
Hi there,
The latest docker pull for anythingllm on a CPU with AVX (No AVX2) seems to result in a core dump when a search is performed after which the container shuts down. Running the same built image on a CPU with AVX2 support doesn't exhibit the same behaviour.
I'm not sure which part of the stack is core dumping but the container logs shows.
[NativeEmbedder] Initialized
[NativeEmbedder] Initialized
[Event Logged] - update_embedding_engine
/usr/local/bin/docker-entrypoint.sh: line 7: 107 Illegal instruction (core dumped) node /app/server/index.js
If something is being compiled with the AVX2 flag enabled, would it be possible to provide a workaround for this on older CPUs with AVX support only?
Cheers.
Are there known steps to reproduce?
No response
The text was updated successfully, but these errors were encountered: