]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
overview / pkg / ggml / sources / whisper.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: c5a5e65)
rpc : reuse compute graph buffers (llama/21299)
authorRadoslav Gerganov <redacted>
Fri, 3 Apr 2026 07:28:09 +0000 (10:28 +0300)
committerGeorgi Gerganov <redacted>
Thu, 30 Apr 2026 08:29:00 +0000 (11:29 +0300)
commit321f62823902b890bc1eb5594f937e853c6afc3b
treee3725abdcf34243cd307dcf8ed2b94af6a64296dtree
parentc5a5e6528ec6002cd1d84f7a11c42255f4550044commit | diff
rpc : reuse compute graph buffers (llama/21299)

Reuse the buffer for the ggml context which is used for creating the
compute graph on the server side. This partially addresses a memory leak
created by the CUDA backend due to using buffer addresses as cache
keys.

ref: #21265
ref: #20315
ggml/src/ggml-rpc/ggml-rpc.cpp diff | blob | history
Packaging of ggerganov/whisper.cpp