logs: catch rocm errors (#12888)

This will help bubble up more crash errors
This commit is contained in:
Daniel Hiltgen
2025-10-31 09:54:25 -07:00
committed by GitHub
parent 7dd4862a89
commit 83537993d7

View File

@@ -23,6 +23,7 @@ func NewStatusWriter(out *os.File) *StatusWriter {
var errorPrefixes = []string{ var errorPrefixes = []string{
"error:", "error:",
"CUDA error", "CUDA error",
"ROCm error",
"cudaMalloc failed", "cudaMalloc failed",
"\"ERR\"", "\"ERR\"",
"error loading model", "error loading model",