Much better than v5
#1
by sverinn - opened
This is much better than caiovicentino1/Qwopus3.5-9B-v3-HLWQ-Q5, good job. Let's hope that the quality measures up to the claim 👍
Tool calling doesn't work for me
It works with --tokenizer Qwen/Qwen3.5-9B but not optimal
thanks for the feedback! i will try to improve for sure
Dear Caiovicentino, the base model Human Eval is 87.8%, why do you say this is a matching quality.
BTW i used KV tq4, it is Human Eval base is about 67.68%.
Dear Caiovicentino, the base model Human Eval is 87.8%, why do you say this is a matching quality.
BTW i used KV tq4, it is Human Eval base is about 67.68%.
sorry, for any problems that you got, im traying my best to improve!