10-17-2026, 08:29 PM
Did a small comparison this afternoon. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.
learning in public
Inference
›
Drivers / Frameworks / Toolkits
›
PyTorch / JAX / Other Frameworks
framework overhead in real deployments: early impressions
framework overhead in real deployments: early impressions
|
framework overhead in real deployments: early impressions
|
|
« Next Oldest | Next Newest »
|
| Messages In This Thread |
|
framework overhead in real deployments: early impressions - by llm_wren - 10-14-2026, 10:57 PM
Re: framework overhead in real deployments: early impressions - by rocmrunner40 - 10-16-2026, 11:04 PM
Re: framework overhead in real deployments: early impressions - by latentpeak22 - 10-17-2026, 11:22 AM
Re: framework overhead in real deployments: early impressions - by forgemixer - 10-17-2026, 07:36 PM
Re: framework overhead in real deployments: early impressions - by torchsupply99 - 10-17-2026, 08:29 PM
Re: framework overhead in real deployments: early impressions - by kernelmike - 10-17-2026, 09:30 PM
Re: framework overhead in real deployments: early impressions - by griffinloom66 - 10-17-2026, 09:56 PM
|
Users browsing this thread: 1 Guest(s)