<@516729262556971018> gpt-oss on ollama on Intel A...
# show-and-tell
j
@bitter-ability-37232 gpt-oss on ollama on Intel A770 GPU:
Copy code
❯ ollama run gpt-oss:20b --verbose
>>> Test
Thinking...
The user just says "Test". They likely want us to respond, maybe to test the conversation. We can respond with something like "Hello! How can I help you today?" Or "Test acknowledged." The instruction says: "You are
ChatGPT..." So likely just respond. We'll respond politely.
...done thinking.

Hello! It looks like you're running a quick test. How can I assist you today?

total duration:       3.947586075s
load duration:        166.244667ms
prompt eval count:    68 token(s)
prompt eval duration: 48.591066ms
prompt eval rate:     1399.43 tokens/s
eval count:           88 token(s)
eval duration:        3.694221308s
eval rate:            23.82 tokens/s
(I actually don't know what a good token per second is, but it was totally acceptable with how quickly it responded for me.)