Updates to use torchao's updated choose_qparams_affine and quantize/dequantize_affine

Summary: Updates to use torchao's updated choose_qparams_affine and quantize/dequantize_affine without the zero_point_domain arg

Differential Revision: D75228037

Author: jainapurva

backends/xnnpack/utils/quant_utils.py

@@ -52,23 +52,22 @@ def is_dynamic_qdq(node: torch.fx.Node) -> bool:
     if not (is_quant(node) or is_dequant(node)):
         return False
 
-    # check scales and zp are dynamically chosen
+    # check scales are dynamically chosen
     node_input_args = node.args
     if is_affine_qdq(node):
         node_input_args = extract_qdq_affine_op_args_for_decomposed_ops(node)
 
     scale = node_input_args[1]
-    zp = node_input_args[2]
-    if not (isinstance(scale, torch.fx.Node) and isinstance(zp, torch.fx.Node)):
+
+    if not (isinstance(scale, torch.fx.Node)):
         return False
 
-    if not (scale.target == operator.getitem and zp.target == operator.getitem):
+    if not (scale.target == operator.getitem):
         return False
 
     scale_choose_qparam = scale.all_input_nodes[0]
-    zp_choose_qparam = zp.all_input_nodes[0]
 
-    if not (is_qparam(scale_choose_qparam) and is_qparam(zp_choose_qparam)):
+    if not (is_qparam(scale_choose_qparam)):
         return False
 
     return True
@@ -222,9 +221,6 @@ def extract_qdq_affine_op_args_for_decomposed_ops(node: torch.fx.Node):
 
     # add target_dtype_node after quant_min/quant_max
     args.append(target_dtype)
-    # zero_point_domain
-    if len(node.args) > 7 and node.args[7] != "INT":
-        return None, None
 
     if is_per_channel_group(node):
         block_sizes = cast(list[int], node.args[1])

exir/passes/_quant_patterns_and_replacements.py

@@ -1017,7 +1017,6 @@ def embedding_byte_dtype_pattern(
             torch.int8,
             -128,
             127,
-            "INT",
             output_dtype,
         )
         return torch.ops.aten.embedding.default(dq, indices)
@@ -1062,7 +1061,6 @@ def embedding_2bit_dtype_pattern(
             torch.int8,
             -2,
             1,
-            "INT",
             output_dtype,
         )
         return torch.ops.aten.embedding.default(dq, indices)
@@ -1110,7 +1108,6 @@ def embedding_4bit_dtype_pattern(
             torch.int8,
             -8,
             7,
-            "INT",
             output_dtype,
         )
         return torch.ops.aten.embedding.default(dq, indices)