feat: support for cross-device row split (#1735)

This commit is contained in:
Piotr Wilkin (ilintar)
2026-07-04 10:31:18 +02:00
committed by GitHub
parent 7bcd189639
commit 68f3d6df9f
12 changed files with 322 additions and 21 deletions

View File

@@ -227,6 +227,7 @@ typedef struct {
bool eager_load; // Load all params into the params backend at model-load time instead of lazily on first use
const char* backend;
const char* params_backend;
const char* split_mode; // weight distribution for multi-device modules: layer (default) or row, or per-module assignments e.g. "diffusion=row"
const char* rpc_servers;
} sd_ctx_params_t;