HackerTrans
TopNewTrendsCommentsPastAskShowJobs

ssheng

no profile record

Submissions

What is Azure thinking with this VM size NVadsA10 v5-series

learn.microsoft.com
1 points·by ssheng·2 anni fa·1 comments

comments

ssheng
·2 anni fa·discuss
To get a full A10, one must also get 36 vCPU and 440 GB of memory. I must be missing something.
ssheng
·2 anni fa·discuss
Quality loss with quantization is expected. It seems like with GPTQ the loss is within acceptable range based on the perplexity score shown.
ssheng
·2 anni fa·discuss
How does Exllama rank among these? Heard good things about it.
ssheng
·2 anni fa·discuss
Creative idea. Any data on how much it takes to load the LoRAs and how much latency it adds to the generation speed?
ssheng
·2 anni fa·discuss
Hello! We're the authors of this blog post. Please let us know if there are other models and inference backends you'd like us to benchmark next.