remove warning

gptqmodel/eora/eora.py

@@ -20,6 +20,7 @@
 
 from ..utils.logger import setup_logger
 from ..utils.rocm import IS_ROCM
+from ..utils.torch import TORCH_GTE_210
 
 log = setup_logger()
 
@@ -91,7 +92,7 @@ def eora_compute_lora(
     # save this later for SVD
     raw_scaling_diag_matrix = eigen_scaling_diag_matrix.to(device=device, dtype=torch.float64)
 
-    if IS_ROCM:
+    if IS_ROCM and not TORCH_GTE_210:
         # hip cannot resolve linalg ops
         original_backend = torch.backends.cuda.preferred_linalg_library()
         torch.backends.cuda.preferred_linalg_library(backend="magma")
@@ -131,7 +132,7 @@ def eora_compute_lora(
     del truc_s, truc_u, truc_v, truc_sigma, sqrtS
 
     # revert linalg backend
-    if IS_ROCM:
+    if IS_ROCM and not TORCH_GTE_210:
         torch.backends.cuda.preferred_linalg_library(original_backend)
 
     return A, B

gptqmodel/utils/linalg_warmup.py

@@ -8,11 +8,32 @@
 import threading
 
 import torch
+from .torch import TORCH_GTE_210
 
 
 _GLOBAL_WARMUP_LOCK = threading.Lock()
 
 
+def _get_cuda_preferred_linalg_library():
+    preferred = getattr(torch.backends.cuda, "preferred_linalg_library", None)
+    if preferred is None:
+        return None
+    if callable(preferred):
+        return preferred()
+    return preferred
+
+
+def _set_cuda_preferred_linalg_library(backend) -> bool:
+    preferred = getattr(torch.backends.cuda, "preferred_linalg_library", None)
+    if preferred is None:
+        return False
+    if callable(preferred):
+        preferred(backend=backend)
+        return True
+    setattr(torch.backends.cuda, "preferred_linalg_library", backend)
+    return True
+
+
 def _make_spd(size: int, device: torch.device, dtype: torch.dtype) -> torch.Tensor:
     """Generate a small symmetric positive definite matrix."""
     base = torch.randn((size, size), device=device, dtype=dtype)
@@ -64,19 +85,18 @@ def run_torch_linalg_warmup(device: torch.device) -> None:
             _run_qr(device, dtype)
 
         if device.type == "cuda" and hasattr(torch.backends, "cuda"):
-            preferred = getattr(torch.backends.cuda, "preferred_linalg_library", None)
-            if callable(preferred):
-                current = preferred()
+            current = _get_cuda_preferred_linalg_library()
+            if current is not None and not TORCH_GTE_210:
                 # Core warmup already ran using the currently preferred backend above.
                 # Some installations fall back to MAGMA when the primary solver is unavailable,
                 # so we pre-initialize MAGMA as well when it differs from the preferred backend.
                 if current and current != "magma":
                     with contextlib.suppress(Exception):
-                        torch.backends.cuda.preferred_linalg_library(backend="magma")
+                        _set_cuda_preferred_linalg_library("magma")
                         _run_cholesky_and_eigh(device, torch.float32)
                 if current:
                     with contextlib.suppress(Exception):
-                        torch.backends.cuda.preferred_linalg_library(backend=current)
+                        _set_cuda_preferred_linalg_library(current)
 
 
 __all__ = ["run_torch_linalg_warmup"]

tests/__init__.py

@@ -0,0 +1 @@
+"""Test package namespace used for local test helpers."""

tests/test_linalg.py

@@ -14,6 +14,7 @@
 import torch  # noqa: E402
 from logbar import LogBar  # noqa: E402
 from parameterized import parameterized  # noqa: E402
+from gptqmodel.utils.torch import TORCH_GTE_210
 
 
 log = LogBar.shared()
@@ -44,11 +45,15 @@ def test_linalg_eigh(self, dtype: torch.dtype, size: int):
         ]
     )
     def test_linalg_eigh_magma(self, dtype: torch.dtype, size: int):
-        # force `magma` backend for linalg
-        original_backend = torch.backends.cuda.preferred_linalg_library()
-        torch.backends.cuda.preferred_linalg_library(backend="magma")
+        # force `magma` backend for linalg when available and allowed
+        restore_backend = None
+        preferred_linalg_library = getattr(torch.backends.cuda, "preferred_linalg_library", None)
+        if not TORCH_GTE_210 and callable(preferred_linalg_library):
+            restore_backend = preferred_linalg_library()
+            preferred_linalg_library(backend="magma")
 
         matrix = torch.randn([size, size], device=ROCM, dtype=dtype)
         torch.linalg.eigh(matrix)
 
-        torch.backends.cuda.preferred_linalg_library(backend=original_backend)
+        if restore_backend is not None:
+            preferred_linalg_library(backend=restore_backend)