What is Azure thinking with this VM size NVadsA10 v5-serieslearn.microsoft.com1 points·by ssheng·há 2 anos·1 comments
ssheng·há 2 anos·discussTo get a full A10, one must also get 36 vCPU and 440 GB of memory. I must be missing something.
ssheng·há 2 anos·discussQuality loss with quantization is expected. It seems like with GPTQ the loss is within acceptable range based on the perplexity score shown.
ssheng·há 2 anos·discussCreative idea. Any data on how much it takes to load the LoRAs and how much latency it adds to the generation speed?
ssheng·há 2 anos·discussHello! 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.