[resubmit] Gemlite fix (#1435)

* [resubmit] Gemlite fix

Summary:
Resubmitting #1432 since it has some rebase issues and
we want to merge the fix asap

Test Plan:
see #1432

Reviewers:

Subscribers:

Tasks:

Tags:

* ruff

test/integration/test_integration.py

@@ -958,11 +958,20 @@ def test_gemlite_layout(self, device, dtype):
                         self._test_lin_weight_subclass_api_impl(
                             api,
                             device,
-                            15, 
+                            15,
                             test_shape=test_shape,
                             test_dtype=dtype,
                         )
 
+        # test that shapes with non divisible by 128 shapes aren't causing errors
+        self._test_lin_weight_subclass_api_impl(
+            lambda mod: quantize_(mod, gemlite_uintx_weight_only(None, 4, 32)),
+            device,
+            15,
+            test_shape=[1, 1025, 513],
+            test_dtype=dtype,
+        )
+
 
     @parameterized.expand(COMMON_DEVICE_DTYPE)
     @unittest.skipIf(not TORCH_VERSION_AT_LEAST_2_3, "int4 requires torch nightly.")

torchao/dtypes/uintx/gemlite_layout.py

@@ -1,3 +1,4 @@
+import warnings
 from dataclasses import dataclass
 from typing import Dict, Optional, Tuple
 
@@ -76,6 +77,15 @@ def apply_gemlite_quant(
     out_features, in_features = weight.shape
     group_size = in_features if group_size is None else group_size
 
+    if in_features % 128 != 0 and out_features % 128 != 0:
+        warnings.simplefilter("once", UserWarning)
+        warnings.warn(
+            "Gemlite only works for layers with in_features or out_features divisible by 128, "
+            + "some layers have been skipped",
+            UserWarning,
+        )
+        return weight
+
     quant_kwargs = get_gemlite_quant_kwargs(bit_width, group_size)
 
     layout = GemlitePackedLayout(
@@ -173,6 +183,11 @@ def from_plain(
             exhaustive=False,
             use_cuda_graph=False,
         )
+        if _layout.group_size is None and _layout.bit_width == 4:
+            from gemlite.core import GEMLITE_ACC_DTYPE
+            from gemlite.dtypes import DType
+
+            GEMLITE_ACC_DTYPE[DType.FP16] = DType.FP32
 
         out_features, in_features = int_data.shape
         input_dtype, output_dtype = DType.FP16, DType.FP16