Spaces:
Running
While using 405b llama model, I often get "Error while parcing tool calls, please retry", and can't continue, even if response was generated.
I've been playing with 405b model recently, and was using it with a big custom prompt (around 1k tokens). It works fine for some time, but inevitably "Error while parcing tool calls, please retry" happens, and I can't continue even if model gave it's response without visible problems. I don't use tools at all, and just talk to the model. Could you at least allow to continue conversation if it happens, please? And, interestingly, every next regeneration has the same problem, so it kind of softlocks you from continuing conversation completely.
The Llama 3.1 405B is a very big model. Mistral Large 2 is a distant dream. Lol
In general if that happens consider disabling all tools to skip the tool use step
To clarify, under "I don't use tools at all" I meant that they all are disabled. Sorry, shouldn't had phrased it like that.
Also, worth mentioning that problem seems to be temporary, because I've noticed that you can solve it by not using model for some time after getting error. Still, kinda annoying.