sgl-project/sglang
v0.5.23Curated ai tools / local-ai / runtime repository (sgl-project/sglang) tracked via conditional GitHub .atom polling.
SGLang is a high-performance serving framework for large language models and diffusion models, designed to maximize throughput and minimize latency through advanced Radix Attention, tensor parallelism, and custom kernel fusions. It addresses the high compute costs and memory bottlenecks of modern generative AI inference by optimizing KV caching and memory layouts across hardware accelerators.
Upgrade DeepSeek-V4 chat encoding and default SWA bounded replay. Update API configurations. Migration: Moderate.
Key Release Highlights (v0.5.23)
- DeepSeek-V4 and DeepSeek-V4.1 chat API requests now route through official
deepseek-recipeencoding. - Enabled DeepSeek-V4.1 decoder SWA bounded replay by default, reducing first-token latency.
- Removed the deprecated
tokenspeed_mlaattention backend and--speculative-skip-dp-mlp-syncflag. - Dropped unsafe
weights_only=Falsepickle loading support for legacy.ptmodel checkpoints.
Breaking Changes (Action Required)
- DeepSeek-V4/V4.1 chat requests now use
deepseek-recipe; customencoding_dsv4.pyendpoints are removed. - Numeric
reasoning_efforton V4 inputs returns a 400 error. - The
tokenspeed_mlaattention backend and--speculative-skip-dp-mlp-syncflag are completely removed. - Unsafe pickle fallback (
weights_only=False) for.ptcheckpoints is disabled.
uv pip install --prerelease=allow sglang==0.5.23Ecosystem Compatibility Matrix & Verified Graph Warnings
Zero-Noise Historical Release Timeline (1)
Format: [Core Change] + [Required Action] + [Migration Effort]Upgrade DeepSeek-V4 chat encoding and default SWA bounded replay. Update API configurations. Migration: Moderate.
Upgrade & Migration Q&A (sgl-project/sglang v0.5.23)
Direct Answers for Engineers & AI AgentsIs sgl-project/sglang v0.5.23 safe to upgrade in production?
sgl-project/sglang v0.5.23 is classified as BREAKING (MINOR release). Upgrade DeepSeek-V4 chat encoding and default SWA bounded replay. Update API configurations. Migration: Moderate.
What breaking changes or compatibility risks exist in sgl-project/sglang v0.5.23?
DeepSeek-V4/V4.1 chat requests now use deepseek-recipe; custom encoding_dsv4.py endpoints are removed. Numeric reasoning_effort on V4 inputs returns a 400 error. The tokenspeed_mla attention backend and --speculative-skip-dp-mlp-sync flag are completely removed. Unsafe pickle fallback (weights_only=False) for .pt checkpoints is disabled.
What is the bundle size impact of sglang v0.5.23?
Measured bundle size telemetry is currently unavailable for sglang v0.5.23; verify build output locally before deploying.
How do I upgrade sglang to v0.5.23?
Upgrade sglang to v0.5.23 by running "pip install sglang==0.5.23" and applying the migration steps: uv pip install --prerelease=allow sglang==0.5.23