Very good model!

#1
by nikitayev - opened

With these parameters:
flashAttention = true
temperature = 1
topKSampling = 500
repeatPenalty = 1
minPSampling = 0.1
topPSampling = 0.95

in LM Studio - give equal results like DeepSeek R1 0528, when text analyzing.

Thank you! I appreciate the feedback. Feel free to share any responses you found noteworthy!

ZeroXClem changed discussion status to closed

Sign up or log in to comment