Advertisement · 728 × 90

Posts by Dagfinn Parnas

Nice, Silkworm next?

1 year ago 0 0 1 0
Post image

My favorite books read in 2024

1 year ago 0 0 0 0

Yes, the error is on the ollama setup of the model as far as I can see. In some cases ollama issues can also have a root in llama.cpp

1 year ago 0 0 0 0
Preview
Add stop word <|endoftext|> to qwq models · Issue #7967 · ollama/ollama What is the issue? The qwq models currently go into an infinite loop. The reasons for this appears that the model outputs <|endoftext|> at the end of its response, but ollama does not handle this a...

Created bug report to ollama now
github.com/ollama/ollam...

1 year ago 2 0 1 0

Same thing I saw. Basically Ollama doesn't stop the llm when the model indicates it's done through the <|endoftext|> token.

Fixed for me through the custom model file link to above (which can be imported through ollama create qwk-fix-stop:latest -f qwq-fix-stop-modelfile.md
FROM qwq:latest)

1 year ago 1 0 2 0

Think the ollama modelfile for qwq is missing a stopword for <|endoftext|>

See of this helps

github.com/elsewhat/adv...

1 year ago 0 0 1 1
Advent of Code 2024

Testing out qwq local llm against adventofcode.com

The chain of thought reasoning of this 32b model is massively impressive.

See for example
github.com/elsewhat/adv...

Tested out day 1,2,3 and all were solved correctly on first attempt

1 year ago 1 0 0 0

Great planning for the fourth session of our bouvet internal architecture school.l Focuses on architecture in challenging deliveries through
1. Real life story telling
2. How to get back on track from delivery leads
3. Identify the signals
4. Incident mgmt and post mortems
5. Secure architecture

1 year ago 1 0 0 0
Advertisement
Preview
Wind and Truth (The Stormlight Archive, #5) The long-awaited explosive climax to the first arc of t…

Only two more days till the release of the final book in Brandon Sandersons Stormlight archives series
www.goodreads.com/book/show/20...

1 year ago 0 0 0 0
Preview
qwq QwQ is an experimental research model focused on advancing AI reasoning capabilities.

In awe of Qwq 32b model reasoning skills and chain of thoughts. Q4 runs fully in memory on my local 4090 gpu with great speed.

Plan to test some more on the advent of code tasks

ollama.com/library/qwq

1 year ago 0 0 0 0

Alt+F4 on x account.
Ahhh and now a fresh start

1 year ago 3 0 0 0