From fd7a7301b0977b0747b89a37286e16a6c4f86bb7 Mon Sep 17 00:00:00 2001
From: RageLtMan <rageltman [at] sempervictus>
Date: Thu, 5 Jun 2025 01:14:42 -0400
Subject: [PATCH] Update cudarc crate to 0.16.4

CUDA/cublas/etc interface library is getting somewhat long in the
tooth compared to current CUDA releases. This is likely why 12.9
fails to build with the appropriate cotainer and may contain fix
for the current matmul/cublas/drop stack traces breaking paged
attention in mistralrs.

Testing:
  Local build - didn't blow up, haven't gotten to mistral tests
---
 Cargo.toml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Cargo.toml b/Cargo.toml
index 7f8245b92e..0d1cc58c51 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -47,7 +47,7 @@ candle-onnx = { path = "./candle-onnx", version = "0.8.0" }
 candle-transformers = { path = "./candle-transformers", version = "0.8.0" }
 clap = { version = "4.2.4", features = ["derive"] }
 criterion = { version = "0.5.1", default-features=false }
-cudarc = { version = "0.13.3", features = ["std", "cublas", "cublaslt", "curand", "driver", "nvrtc", "f16", "cuda-version-from-build-system", "dynamic-linking"], default-features=false }
+cudarc = { version = "0.16.4", features = ["std", "cublas", "cublaslt", "curand", "driver", "nvrtc", "f16", "cuda-version-from-build-system", "dynamic-linking"], default-features=false }
 fancy-regex = "0.13.0"
 gemm = { version = "0.17.0", features = ["wasm-simd128-enable"] }
 hf-hub = { version = "0.3.3", package = "candle-hf-hub" }