Upgrade autoround 0.10.2 (#2410)

Signed-off-by: yiliu30 <yi4.liu@intel.com>

SUMMARY:
"please provide a brief summary"


TEST PLAN:
"please outline how the changes were tested"

cc @hshen14 @thuang6 @chensuyue

---------

Signed-off-by: yiliu30 <yi4.liu@intel.com>
Co-authored-by: HDCharles <39544797+HDCharles@users.noreply.github.com>

Author: yiliu30

setup.py

@@ -129,10 +129,9 @@ def localversion_func(version: ScmVersion) -> str:
         ),
         ("datasets>=4.0.0,<=4.6.0" if BUILD_TYPE == "release" else "datasets>=4.0.0"),
         (
-            # auto-round 0.9.1 cannot work with accelerate <1.10.0
-            "auto-round>=0.9.6,<=0.10.2"
+            "auto-round>=0.10.2,<=0.10.2"
             if BUILD_TYPE == "release"
-            else "auto-round>=0.9.6"
+            else "auto-round>=0.10.2"
         ),
         (
             "accelerate>=1.6.0,<=1.12.0"

src/llmcompressor/modifiers/autoround/base.py

@@ -262,7 +262,7 @@ def apply_autoround(self, state, subgraph):
 
         # Build kwargs for AutoRound initialization
         ar_quant_scheme = self._mapping_config_to_autoround()
-        fp_layers = self.get_unquantized_layer_names(decoding_layer)
+        ignore_layers = self.get_unquantized_layer_names(decoding_layer)
         kwargs = {
             "tokenizer": "",  # A placeholder
             "scheme": ar_quant_scheme,
@@ -271,7 +271,7 @@ def apply_autoround(self, state, subgraph):
             "enable_torch_compile": self.enable_torch_compile,
             "batch_size": self.batch_size,
             "device_map": self.device_ids,
-            "fp_layers": ",".join(fp_layers) if fp_layers else "",
+            "ignore_layers": ",".join(ignore_layers) if ignore_layers else "",
         }
 
         llmc_registered_qparams = self._preprocess_qparams(decoding_layer)