Hey guys how is the audio, i fixed the static issue from last video. Is it the volume ok now? Also how you guys like my edits now. we come far guys from a week ago to now. We got cool backgrouund and zoom now. What you guys think? LFG we gonna go harder than everyother ai channel. Thx for yall support!!
Did you try it in a fresh project or was there lots of context. I know groq free plan has rate limits not sure if that is the case here. I’m on developer plan with higher limits. Try simpler prompt like Hello world :) with no files
@AGENTWORKFLOW I created a new conversation and said "hello". Apparently, my custom memory prompt and initial prompt add up to 14k instead of the 6k limit. Most of my requests average around 20k. That's 50 requests before reaching 1m tokens. I'm guessing as optimization goes up in GROQ, the rates will also go up.
yes! In fact it was last video ua-cam.com/video/K4mL426Sb1Y/v-deo.html I should do a better job of highlighting my older video. 😂😂 skip to 2:45 to see run deepseek-r1-distill-1.5b. I tried that once because it can easily be run on smartphones. But of course the best performance once are 32b & 70b if have the hardware. Play around with the quant to get the best one. GL lmk how it goes :)
deepseek-r1-distill-llama-70b. That is the name of the model listed on groq and on deepseek and hf etc. It’s distilled version of course. Thanks I have updated the title so its not to be confused :)
![Roo Cline Vs Code + DeepSeek R1 + Copilot Sonnet FREE AI IDE [Better than Windsurf/ Cursor in 5min]](/img/n.gif)
Hey guys how is the audio, i fixed the static issue from last video. Is it the volume ok now?
Also how you guys like my edits now. we come far guys from a week ago to now. We got cool backgrouund and zoom now. What you guys think? LFG we gonna go harder than everyother ai channel. Thx for yall support!!
use ai to enhance your voice :)
@@archiee1337😮😅. Should I do a video on Kokoro TTS ?
@@AGENTWORKFLOW yessss. and next: audio (via mic) to audio
I always get request too large for distilled with your exact same prompt.
Did you try it in a fresh project or was there lots of context. I know groq free plan has rate limits not sure if that is the case here. I’m on developer plan with higher limits. Try simpler prompt like
Hello world :) with no files
@AGENTWORKFLOW I created a new conversation and said "hello". Apparently, my custom memory prompt and initial prompt add up to 14k instead of the 6k limit. Most of my requests average around 20k. That's 50 requests before reaching 1m tokens. I'm guessing as optimization goes up in GROQ, the rates will also go up.
Thanks great
The distill version could be run locally, right? Have you tried that?
yes! In fact it was last video ua-cam.com/video/K4mL426Sb1Y/v-deo.html
I should do a better job of highlighting my older video. 😂😂
skip to 2:45 to see run deepseek-r1-distill-1.5b. I tried that once because it can easily be run on smartphones. But of course the best performance once are 32b & 70b if have the hardware. Play around with the quant to get the best one.
GL lmk how it goes :)
70b is not R1
deepseek-r1-distill-llama-70b. That is the name of the model listed on groq and on deepseek and hf etc. It’s distilled version of course. Thanks I have updated the title so its not to be confused :)