RequestHunt is a community-driven platform that collects and curates feature requests from X (Twitter), Reddit, and GitHub. We help product teams discover what features real users are asking for, making it easier to validate product ideas and prioritize feature development based on actual user demand.

How does RequestHunt collect feature requests?

RequestHunt uses AI-powered extraction to automatically collect feature requests from social media platforms including X, Reddit, and GitHub. The platform monitors discussions, issues, and posts to identify genuine user feature requests, extracting key information like the requested feature, product mentioned, and user sentiment.

Who uses RequestHunt?

RequestHunt is used by product managers, founders, developers, and product teams who want to discover what features real users are requesting. It's particularly valuable for startups looking to validate ideas, established companies prioritizing their roadmap, and developers building products that solve real user problems.

Is RequestHunt free to use?

Yes, browsing and searching feature requests on RequestHunt is completely free. We also offer API access for developers who want to integrate feature request data into their own applications. Check our pricing page for API plan details.

How often is RequestHunt data updated?

RequestHunt continuously collects feature requests from Reddit, X (Twitter), and GitHub. You can also trigger on-demand scraping via the API or web interface to get the freshest data for any topic.

[vLLM] Optimize on-device sampling performance on Tenstorrent hardware

Optimize on-device sampling performance in vLLM on Tenstorrent hardware. Non-greedy device sampling is currently ~2x slower than CPU sampling at batch=1. Investigation will start with IR analysis and Tracy profiling to identify the bottlenecks.

Original Source

@@kmabeeTT | 0 pts | 6d ago

Tracking issue for optimizing on-device sampling performance in vLLM on Tenstorrent hardware. Individual issues will be opened as specific areas are investigated. Non-greedy device sampling is currently ~2x slower than CPU sampling at batch=1: | Configuration | OPT-125M tok/s | Llama-3.1-8B tok/s | |---|---|---| | Greedy device | 11.61 | 9.82 | | Non-greedy device | 5.94 | 4.02 | | Non-greedy CPU | 11.00 | 7.89 | Investigation will start with IR analysis and Tracy profiling to identify the dominant ops in the non-greedy sampling graph. Branch: `kmabee/vllm_perf_debug`

Discussion

Loading comments...