Hi everyone,
I’ve been using W&B to track some performance benchmarks for a cross-platform optimization project, and I’ve run into a specific technical snag regarding the web dashboard’s responsiveness when my local testing environment is under heavy resource load.
I’m currently running a series of high-concurrency script tests. While I was chatting with some developers who recently updated the download icon assets and optimized the background executors over in the arceusx community, I noticed that my W&B runs started showing significant “heartbeat” latencies. In some cases, the wandb.log calls seem to hang or cause the mobile webview to stutter, leading to patchy data in my workspace and intermittent “Network Error (ReadTimeout)” messages.
Has anyone else noticed that the W&B background process (the internal wandb service) struggles to compete for CPU cycles when other high-performance executors are active on the same device? I’m also seeing some weird rendering issues where the live-charts fail to update in real-time unless I manually refresh the browser, which makes it hard to monitor for anomalies during long-running sweeps.
I’m trying to figure out if there’s a way to adjust the wandb.init settings to be more “low-priority” or if I should increase the internal_process_timeout to account for these CPU spikes without losing data points. If you’ve dealt with stabilizing experiment tracking in resource-constrained or “noisy” mobile environments, I’d really value your advice on how to keep the data flow consistent!
Thanks in advance for the help!