Fix typo for 16a4w_block quantization

rohansjoshi · web-flow · commit 63b047b6ecb4 · 2025-06-20T15:01:42.000-07:00
Differential Revision: D76949897 Pull Request resolved: pytorch#11821
diff --git a/examples/qualcomm/oss_scripts/llama/llama.py b/examples/qualcomm/oss_scripts/llama/llama.py
@@ -398,7 +398,7 @@ def quantize(self, quant_dtype, args, tokenizer, custom_annotations=()):
                 self.llama_graph_module, self.inputs, strict=True
             ).module()
 
-            if QuantDtype == QuantDtype.use_16a4w_block:
+            if quant_dtype == QuantDtype.use_16a4w_block:
                 conv_nodes = [
                     n for n in fx_graph_module.graph.nodes if "conv" in n.name
                 ]

Original file line number	Diff line number	Diff line change
`@@ -398,7 +398,7 @@ def quantize(self, quant_dtype, args, tokenizer, custom_annotations=()):`
`398`	`398`	`self.llama_graph_module, self.inputs, strict=True`
`399`	`399`	`).module()`
`400`	`400`
`401`		`- if QuantDtype == QuantDtype.use_16a4w_block:`
	`401`	`+ if quant_dtype == QuantDtype.use_16a4w_block:`
`402`	`402`	`conv_nodes = [`
`403`	`403`	`n for n in fx_graph_module.graph.nodes if "conv" in n.name`
`404`	`404`	`]`