llama-fit-params: keep explicit --ctx-size 0 (#19070)

This commit is contained in:
Johannes Gäßler 2026-01-24 22:13:08 +01:00 committed by GitHub
parent 4e5b83b226
commit e9fd8dcab4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 12 additions and 3 deletions

View file

@ -489,6 +489,7 @@ extern "C" {
// - returns true if the parameters could be successfully modified to fit device memory
// - this function is NOT thread safe because it modifies the global llama logger state
// - only parameters that have the same value as in llama_default_model_params are modified
// with the exception of the context size which is modified if and only if equal to 0
LLAMA_API enum llama_params_fit_status llama_params_fit(
const char * path_model,
struct llama_model_params * mparams,