bitsandbytes: simplify 8bit dequantization (#35068)

src/transformers/integrations/bitsandbytes.py

@@ -363,13 +363,14 @@ def dequantize_bnb_weight(weight: "torch.nn.Parameter", dtype: "torch.dtype", st
     if state.SCB is None:
         state.SCB = weight.SCB
 
-    im = torch.eye(weight.data.shape[-1]).contiguous().half().to(weight.device)
-    im, imt, SCim, SCimt, coo_tensorim = bnb.functional.double_quant(im)
-    im, Sim = bnb.functional.transform(im, "col32")
-    if state.CxB is None:
-        state.CxB, state.SB = bnb.functional.transform(weight.data, to_order=state.formatB)
-    out32, Sout32 = bnb.functional.igemmlt(im, state.CxB, Sim, state.SB)
-    return bnb.functional.mm_dequant(out32, Sout32, SCim, state.SCB, bias=None).t().to(dtype)
+    if hasattr(bnb.functional, "int8_vectorwise_dequant"):
+        # Use bitsandbytes API if available (requires v0.45.0+)
+        dequantized = bnb.functional.int8_vectorwise_dequant(weight.data, state.SCB)
+    else:
+        # Multiply by (scale/127) to dequantize.
+        dequantized = weight.data * state.SCB.view(-1, 1) * 7.874015718698502e-3
+
+    return dequantized.to(dtype)
 
 
 def _create_accelerate_new_hook(old_hook):