Fix #1588 - torch compatability for <=2.4 (#1590)

bitsandbytes/_ops.py

@@ -19,7 +19,7 @@ else:
 # Higher level op: int8 matmul + dequant + bias
 torch.library.define(
     "bitsandbytes::int8_scaled_mm",
-    "(Tensor A, Tensor B, Tensor row_stats, Tensor col_stats, Tensor? bias=None, ScalarType dtype=float16) -> Tensor",
+    "(Tensor A, Tensor B, Tensor row_stats, Tensor col_stats, Tensor? bias=None, ScalarType? dtype=None) -> Tensor",
 )
 
 
@@ -30,10 +30,10 @@ def _(
     row_stats: torch.Tensor,
     col_stats: torch.Tensor,
     bias: Optional[torch.Tensor] = None,
-    dtype=torch.float16,
+    dtype: Optional[torch.dtype] = None,
 ) -> torch.Tensor:
     shapeC = (*A.shape[:-1], B.shape[0])
-    return torch.empty(shapeC, device=A.device, dtype=dtype)
+    return torch.empty(shapeC, device=A.device, dtype=dtype or torch.float16)
 
 
 torch.library.define(
@@ -98,7 +98,7 @@ def _(A: torch.Tensor, stats: torch.Tensor) -> torch.Tensor:
 
 
 # Default PyTorch-native implementation
-@register_kernel("bitsandbytes::int8_vectorwise_dequant", None)
+@register_kernel("bitsandbytes::int8_vectorwise_dequant", "default")
 def _(A: torch.Tensor, stats: torch.Tensor):
     # To dequantize we divide by 127, or multiply by the reciprocal.
     return A * stats.view(-1, 1) * 7.874015718698502e-3
@@ -106,7 +106,7 @@ def _(A: torch.Tensor, stats: torch.Tensor):
 
 torch.library.define(
     "bitsandbytes::int8_mm_dequant",
-    "(Tensor A, Tensor row_stats, Tensor col_stats, ScalarType dtype=float16, Tensor? bias=None) -> Tensor",
+    "(Tensor A, Tensor row_stats, Tensor col_stats, ScalarType? dtype=None, Tensor? bias=None) -> Tensor",
 )
 
 
@@ -115,11 +115,11 @@ def _(
     A: torch.Tensor,
     row_stats: torch.Tensor,
     col_stats: torch.Tensor,
-    dtype=torch.float16,
+    dtype: Optional[torch.dtype] = None,
     bias: Optional[torch.Tensor] = None,
 ) -> torch.Tensor:
     torch._check(A.dtype == torch.int32, lambda: "A must be int32")
-    return torch.empty_like(A, dtype=dtype)
+    return torch.empty_like(A, dtype=dtype or torch.float16)
 
 
 torch.library.define(

bitsandbytes/backends/cpu/ops.py

@@ -28,7 +28,7 @@ def _(
     A: torch.Tensor,
     row_stats: torch.Tensor,
     col_stats: torch.Tensor,
-    dtype=torch.float16,
+    dtype: Optional[torch.dtype] = None,
     bias: Optional[torch.Tensor] = None,
 ) -> torch.Tensor:
     torch._check(A.dtype == torch.int32, lambda: f"A must be int32, got {A.dtype}")
@@ -43,7 +43,7 @@ def _(
     if bias is not None:
         out += bias
 
-    return out.to(dtype)
+    return out.to(dtype or torch.float16)
 
 
 @register_kernel("bitsandbytes::quantize_blockwise", "cpu")

bitsandbytes/backends/cuda/ops.py

@@ -90,7 +90,7 @@ def _(
     A: torch.Tensor,
     row_stats: torch.Tensor,
     col_stats: torch.Tensor,
-    dtype=torch.float16,
+    dtype: Optional[torch.dtype] = None,
     bias: Optional[torch.Tensor] = None,
 ) -> torch.Tensor:
     torch._check(A.dtype == torch.int32, lambda: f"A must be int32, got {A.dtype}")
@@ -121,7 +121,7 @@ def _(
     if bias is not None and bias.dtype != torch.float16:
         out.add_(bias)
 
-    return out.to(dtype)
+    return out.to(dtype or torch.float16)
 
 
 @register_kernel("bitsandbytes::int8_vectorwise_quant", "cuda")

bitsandbytes/backends/default/ops.py

@@ -5,26 +5,31 @@ import torch
 from ..._ops import register_kernel
 
 
-@register_kernel("bitsandbytes::int8_scaled_mm", None)
+@register_kernel("bitsandbytes::int8_scaled_mm", "default")
 def _(
     A: torch.Tensor,
     B: torch.Tensor,
     row_stats: torch.Tensor,
     col_stats: torch.Tensor,
     bias: Optional[torch.Tensor] = None,
-    dtype=torch.float16,
+    dtype: Optional[torch.dtype] = None,
 ) -> torch.Tensor:
     out_i32 = torch.ops.bitsandbytes.int8_linear_matmul.default(A, B)
-    out = torch.ops.bitsandbytes.int8_mm_dequant.default(out_i32, row_stats, col_stats, dtype=dtype, bias=bias)
-    return out
+    return torch.ops.bitsandbytes.int8_mm_dequant.default(
+        out_i32,
+        row_stats,
+        col_stats,
+        dtype=dtype or torch.float16,
+        bias=bias,
+    )
 
 
-@register_kernel("bitsandbytes::int8_linear_matmul", None)
+@register_kernel("bitsandbytes::int8_linear_matmul", "default")
 def _(A: torch.Tensor, B: torch.Tensor):
     return _int8_linear_matmul_impl(A, B)
 
 
-@register_kernel("bitsandbytes::int8_linear_matmul.out", None)
+@register_kernel("bitsandbytes::int8_linear_matmul.out", "default")
 def _(A: torch.Tensor, B: torch.Tensor, out: torch.Tensor):
     torch._check(out.dtype == torch.int32)
     _int8_linear_matmul_impl(A, B, out)