[MTHREADS] adapt some config to mt backend

src/flag_gems/runtime/backend/_mthreads/ops/arange.py

@@ -13,7 +13,7 @@
 from flag_gems.utils import triton_lang_extension as tle
 
 logger = logging.getLogger(
-    f'flag_gems.runtime.backend._mthreads.ops.{__name__.split(".")[-1]}'
+    f"flag_gems.runtime.backend._mthreads.ops.{__name__.split('.')[-1]}"
 )
 
 device_ = runtime.device
@@ -140,6 +140,19 @@ def arange_start(
     else:
         device = torch.device(device)
 
+    # Handle int64 dtype with float parameters - convert to int
+    if dtype is torch.int64:
+        if (
+            isinstance(start, float)
+            or isinstance(end, float)
+            or isinstance(step, float)
+        ):
+            start = int(start) if isinstance(start, float) else start
+            end = int(end) if isinstance(end, float) else end
+            step = int(step) if isinstance(step, float) else step
+            if step == 0:
+                raise RuntimeError("step must be nonzero")
+
     is_float_dtype = torch.is_floating_point(torch.tensor(0, dtype=dtype))
     use_int64 = dtype == torch.int64
     size = _compute_size(start, end, step, is_float_dtype)

src/flag_gems/runtime/backend/_mthreads/ops/repeat.py

@@ -381,7 +381,7 @@ def repeat(inp: torch.Tensor, sizes) -> torch.Tensor:
         assert (
             sizes_shape[i] >= 0
         ), f"the number of repetitions per dimension out of range (expected to >= 0) but got {sizes_shape[i]}"
-        if sizes_shape[i] == 0:
+        if in0_shape[i] * sizes_shape[i] == 0:
             is_empty = True
         out_shape.append(in0_shape[i] * sizes_shape[i])
 

src/flag_gems/runtime/backend/_mthreads/ops/repeat_interleave.py

@@ -12,7 +12,7 @@
 from flag_gems.utils.tensor_wrapper import StridedBuffer
 
 logger = logging.getLogger(
-    f'flag_gems.runtime.backend._mthreads.ops.{__name__.split(".")[-1]}'
+    f"flag_gems.runtime.backend._mthreads.ops.{__name__.split('.')[-1]}"
 )
 
 
@@ -437,6 +437,9 @@ def fused_repeat_interleave_dim0(inp, repeats, dim):
 def repeat_interleave_self_tensor(inp, repeats, dim=None, *, output_size=None):
     logger.debug("GEMS_MTHREADS REPEAT_INTERLEAVE_SELF_TENSOR")
 
+    if repeats.numel() == 0:
+        return inp.clone()
+
     if dim is None:
         inp = inp.flatten()
         dim = 0

src/flag_gems/runtime/backend/_mthreads/tune_configs.yaml

@@ -301,6 +301,26 @@ cross_entropy_loss_sum_and_scale:
       - 256
       - 1024
 
+conj_physical:
+  - META:
+      BLOCK_SIZE: 64
+    num_warps: 8
+  - META:
+      BLOCK_SIZE: 128
+    num_warps: 8
+  - META:
+      BLOCK_SIZE: 256
+    num_warps: 8
+  - META:
+      BLOCK_SIZE: 512
+    num_warps: 8
+  - META:
+      BLOCK_SIZE: 1024
+    num_warps: 8
+  - META:
+      BLOCK_SIZE: 2048
+    num_warps: 8
+
 upsample_nearest2d:
   - gen: true
     param_map: