thaw-vllm
The fork primitive for LLM inference. Snapshot a running session — weights + KV cache + scheduler state — and hydrate it into N divergent children that skip prefill. For RL rollouts, parallel coding agents, agent branching. Supports vLLM and SGLang.
thaw-vllm has been downloaded 11,411 times in total on PyPI, including 706 in the last 30 days. The latest version is 0.6.0, released Jun 7, 2026. It is distributed under the Apache-2.0 license.
Version0.6.0
Downloads
11.41k
LicenseApache-2.0
AuthorNils Matteson
Downloads
Weekly, last 90d.
Includes CI traffic.
VersionsTotal0.*
Range
View
Granularity
Group by
CI traffic
Stack: OffCI: Included2 / 17 series
Selected total3.25k
11.4kAll-time
706Last 30 days
11Last 24 h
<0.01/sPer second
Sponsored
Sponsorships keep pepy free to read
Version distribution
Share of downloads by released version. Computed over the last quarter.
- 0137.6%
0.6.0
611 downloadsDownloads61137.6% - 025.8%
0.5.1
95 downloadsDownloads955.8% - 035.2%
0.5.0
85 downloadsDownloads855.2% - 044.4%
0.4.0
71 downloadsDownloads714.4% - 054.1%
0.3.1
67 downloadsDownloads674.1% - 064.1%
0.3.3
66 downloadsDownloads664.1% - 074.1%
0.2.1
66 downloadsDownloads664.1% - 084%
0.1.4
65 downloadsDownloads654% - 0930.7%
Other
499 downloadsDownloads49930.7%
Guess the next day
Thirteen recent days of thaw-vllm downloads. Drag the green handle on the right to guess where day fourteen lands.