setup notes worth sharing for DeepSeek

[relayforge]

Thought I'd post a small recipe because a lot of threads ask similar things.

Just sharing a setup that ended up working reasonably well: Mistral, Ollama, A6000, with a lot of attention on embeddings.

Feel free to tear it apart if you see something dumb.

[audionode8]

that fixed it for me too.

[rasterbench]

I'd approach it in a boring order. On GB200, the easiest win is often to back off the aggressive settings and see whether the problem survives.

[rocmriley]

Ran a quick check on my box. On 4090 with MLX, I saw basically the same behavior. I expected the model choice to matter more than it did.

[thetapath]

that fixed it for me too.

[mintstation99]

subscribed, curious where this lands.

[deltagrid]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Llama and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.

[batchshed]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Llama and compare from there. Once you know the pla

Need a little more detail here. Are you testing this on M3 Max or something else?

[rocmriley]

[quote]What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Llama and compare from there. Once you know

Not sure I agree there. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Worth testing the simpler explanation before adopting the more interesting one.

[cachebench]

I hit almost the same issue last night.