[SW-191317] Raise exception according to hqt config object

Change-Id: I06ba8fa912c811c88912987c11e5c12ef328348a
intel · Jul 3, 2024 · 768c2a4 · 768c2a4
1 parent 52a98f4
commit 768c2a4
Show file tree

Hide file tree

Showing 2 changed files with 7 additions and 5 deletions.
diff --git a/neural_compressor/torch/algorithms/fp8_quant/common.py b/neural_compressor/torch/algorithms/fp8_quant/common.py
@@ -24,7 +24,13 @@
 
 def save_calib_result(model):
     import habana_quantization_toolkit as hqt
-    hqt.finish_measurements(model)
+    if (hasattr(model, "__hqt_config__") and
+            isinstance(model.__hqt_config__, hqt._quant_common.quant_config.Fp8cfg)):
+        # TODO SW-184714 modify hqt notation to inc notation once code is ported
+        hqt.finish_measurements(model)
+    else:
+        raise NotImplementedError("Saving calibration results currently supported only in HPU.")
+
 
 
 def update_mode(config_path, measure_step=False, quant_step=False):

diff --git a/neural_compressor/torch/quantization/quantize.py b/neural_compressor/torch/quantization/quantize.py
@@ -220,9 +220,5 @@ def convert(
 
 
 def finalize_calibration(model):
-    if hasattr(model, "quant_config") and isinstance(model.quant_config, FP8Config): # FP8
         from neural_compressor.torch.algorithms.fp8_quant import save_calib_result
-
         save_calib_result(model)
-    else:
-        raise NotImplementedError("`finalize_calibration` only supports FP8 measurement now.")