v1.3.27

Install: pip install freesolo-flash==1.3.27

PyPI: https://pypi.org/project/freesolo-flash/1.3.27/

What's Changed

• Optimize DeepSeek serving and autoscaling; fix shared Qwen paths by @DavidBShan in #1508
• Merge pull request #1508 from freesolo-co/codex/deepinfer-latency by @DavidBShan in #1509
• Initialize Qwen tokenizer cache before admitting streams by @DavidBShan in #1510
• Initialize Qwen tokenizer cache before admitting streams (#1510) by @DavidBShan in #1511
• Scale DeepSeek with 64 spare streams across the pool by @DavidBShan in #1512
• Fix dev deployment verification for regional Modal URL by @DavidBShan in #1514
• Merge pull request #1512 from freesolo-co/codex/deepinfer-pool-headroom by @DavidBShan in #1513
• Bump release version after serving autoscaling changes (#1515) by @DavidBShan in #1516

Full Changelog: v1.3.26...v1.3.27