Jumping between the frames, it looks like the 32b model beats Llama 3.1 70b handily, which is amazing for a model we should be able to run on consumer gpus
the modelfile of qwen 2.5 instruct is broken as of now (ollama version 0.3.11). you have to replace the template with a simpler one, so that the system prompt works.
Thanks codeking
always waiting for your new video
And me
King, you are working hard, thanks for your work!
Need some auto code suggestion videos with these videos
Jumping between the frames, it looks like the 32b model beats Llama 3.1 70b handily, which is amazing for a model we should be able to run on consumer gpus
King will always like your comment😂
yeah we all are very lucky liked by the King directly, usually commoner like you won't get that much affections.
the modelfile of qwen 2.5 instruct is broken as of now (ollama version 0.3.11). you have to replace the template with a simpler one, so that the system prompt works.
Amazing stuff!
This 14B model have same issue for Russian language as Llama 3.1 8B/70B, sometimes it writes on chinese instead
best ai channel. make discord please, which would be cool !!
if only those not censored they can talk each other in CoT agentic to reach o1 like reasoning....
which is better?? yi coder 9b or qwen2.5 coder 7b??
can we put on cursor or pearAI
Can we put in Claude Dev. That would be ..............................................................really cool
Carmella Drives
it's pretty cool
Inshallah looking forward to seeing the real life tests. Maybe you can also compare the closed sourced version with Claude?
Hi. Your videos are so detailed and insightful. I wanted to see if you'd be open to doing a collaboration together.
Sponsorship?
Good!
Alibaba💪💪