Added default NPI file

QEfficient/transformers/models/gemma3/modeling_gemma3.py

@@ -677,6 +677,17 @@ def forward(
         logits = logits.float()
         return logits, pixel_values, image_idx, outputs.past_key_values
 
+    def get_npi_file(self, model_name: str, **compiler_options):
+        if model_name == "google/gemma-3-4b-it":
+            compiler_options["node_precision_info"] = constants.DEFAULT_GEMMA3_4B_NODE_PRECISION_INFO
+        elif model_name == "google/gemma-3-27b-it":
+            compiler_options["node_precision_info"] = constants.DEFAULT_GEMMA3_27B_NODE_PRECISION_INFO
+        else:
+            raise ValueError(
+                f"For Model {self.pretrained_model_name_or_path} default NPI file is not supported/added. Please use one of the following: google/gemma-3-4b-it, google/gemma-3-27b-it"
+            )
+        return compiler_options
+
     def get_specializations(
         self,
         batch_size: int,

QEfficient/transformers/models/modeling_auto.py

@@ -1183,7 +1183,6 @@ def compile(
         compiler_options.pop("continuous_batching", None)
         compiler_options.pop("kv_cache_batch_size", None)
         compiler_options.pop("full_batch_size", None)
-
         if not skip_vision:
             self.vision_model._compile(
                 compile_dir=compile_dir,
@@ -1199,6 +1198,10 @@ def compile(
                 **compiler_options,
             )
 
+        # Custom NPI file options
+        if hasattr(self.model, "get_npi_file"):
+            compiler_options = self.model.get_npi_file(self.model.name_or_path)
+
         if not skip_lang:
             custom_io_lang = {}
             # Inputs
@@ -1212,7 +1215,6 @@ def compile(
             for output_name in output_names["lang"]:
                 if output_name.endswith("_RetainedState"):
                     custom_io_lang[output_name] = "float16" if "vision_embeds" in output_name else kv_cache_dtype
-
             self.lang_model._compile(
                 compile_dir=compile_dir,
                 compile_only=True,
@@ -1801,6 +1803,9 @@ def compile(
             **compiler_options,
         )
 
+        if hasattr(self.model, "get_npi_file"):
+            compiler_options = self.model.get_npi_file(self.model.name_or_path)
+
         custom_io = {}
         kv_cache_dtype = "mxint8" if mxint8_kv_cache else "float16"
         # inputs
@@ -1819,7 +1824,6 @@ def compile(
         compiler_options.pop("continuous_batching", None)
         compiler_options.pop("kv_cache_batch_size", None)
         compiler_options.pop("full_batch_size", None)
-
         self._compile(
             onnx_path=onnx_path,
             compile_dir=compile_dir,

QEfficient/utils/constants.py

@@ -7,6 +7,9 @@
 
 import os
 from dataclasses import dataclass
+from pathlib import Path
+
+from QEfficient.transformers.models import gemma3 as gemma3
 
 UTILS_DIR = os.path.dirname(os.path.abspath(__file__))
 QEFF_DIR = os.path.dirname(UTILS_DIR)
@@ -24,6 +27,23 @@
 ONNX_EXPORT_IMAGE_DEPTH = 3
 ONNX_EXPORT_CTX_LEN = 1024
 
+# Gemma3 Constant
+DEFAULT_GEMMA3_4B_NODE_PRECISION_INFO = str(
+    Path(__file__).resolve().parent.parent
+    / "transformers"
+    / "models"
+    / "gemma3"
+    / "configs"
+    / "fp32_nodes_gemma3_4b.yaml"
+)
+DEFAULT_GEMMA3_27B_NODE_PRECISION_INFO = str(
+    Path(__file__).resolve().parent.parent
+    / "transformers"
+    / "models"
+    / "gemma3"
+    / "configs"
+    / "fp32_nodes_gemma3_27b.yaml"
+)
 # Compiler defaults
 DEFAULT_AIC_NUM_CORES = 16
 DEFAULT_AIC_MXPF6_MATMUL = False

examples/image_text_to_text/models/gemma_vision/gemma3_example.py

@@ -44,7 +44,6 @@
         aic_enable_depth_first=True,
         skip_vision=True,
         mos=1,
-        node_precision_info="examples/gemma3_example/fp32_nodes_gemma3_4b.yaml",  # Change to fp32_nodes_gemma3_27b.yaml for 27B model
     )
 
     messages = [
@@ -80,7 +79,6 @@
         mxint8_kv_cache=False,
         aic_enable_depth_first=True,
         mos=1,
-        node_precision_info="examples/gemma3_example/fp32_nodes_gemma3_4b.yaml",  # Change to fp32_nodes_gemma3_27b.yaml for 27B model
     )
 
     ### IMAGE + TEXT ###

pyproject.toml

@@ -54,6 +54,9 @@ test = ["pytest","pytest-mock"]
 docs = ["Sphinx==7.1.2","sphinx-rtd-theme==2.0.0","myst-parser==3.0.1","sphinx-multiversion"]
 quality = ["black", "ruff", "hf_doc_builder@git+https://github.com/huggingface/doc-builder.git"]
 
+[tool.setuptools.package-data]
+"QEfficient.transformers.models.gemma3.configs" = ["*.yaml"]
+
 [build-system]
 requires = ["setuptools>=62.0.0"]
 build-backend = "setuptools.build_meta"