Local_modify

.gitignore

@@ -10,3 +10,7 @@ venv3/
 logs/
 results/
 .ipynb_checkpoints/
+
+kernels/int_quantization*
+kernels/build/
+kernels/dist/

inference/inference_sim.py

@@ -1,7 +1,14 @@
 import os, sys
-dir_path = os.path.dirname(os.path.realpath(__file__))
-root_dir = os.path.join(dir_path, os.path.pardir)
-sys.path.append(root_dir)
+
+# dir_path = os.path.dirname(os.path.realpath(__file__))
+# root_dir = os.path.join(dir_path, os.path.pardir)
+# sys.path.append(root_dir)
+
+current_dir = os.path.dirname(os.path.abspath(__file__))
+parent_dir = os.path.abspath(os.path.join(current_dir, os.pardir))
+sys.path.insert(0, parent_dir)
+
+
 import argparse
 import time
 import logging
@@ -127,8 +134,13 @@
 torch.manual_seed(12345)
 
 
+
 class InferenceModel:
     def __init__(self, ml_logger=None):
+
+
+        self.onnx_save = True
+
         self.ml_logger = ml_logger
         global args, best_prec1
 
@@ -229,6 +241,7 @@ def __init__(self, ml_logger=None):
             num_workers=args.workers, pin_memory=True)
 
     def run(self):
+
         if args.eval_precision:
             elog = EvalLog(['dtype', 'val_prec1', 'val_prec5'])
             print("\nFloat32 no quantization")
@@ -274,8 +287,8 @@ def run(self):
 
             return val_loss, val_prec1, val_prec5
 
-
 def validate(val_loader, model, criterion):
+    onnx_save = True
     batch_time = AverageMeter()
     losses = AverageMeter()
     top1 = AverageMeter()
@@ -304,6 +317,14 @@ def validate(val_loader, model, criterion):
                 QM().verbose = True
             input = input.to(args.device)
             target = target.to(args.device)
+            if i == 0 and onnx_save == True:
+                onnx_save = False
+                quantized_model_path = 'quantized_model.pth'
+                quantized_model_path_onnx = 'quantized_model.onnx'
+                torch.onnx.export(model, input, quantized_model_path_onnx)
+                torch.save(model.state_dict(), quantized_model_path)
+                print(f"Quantized model saved to {quantized_model_path}")
+
             if args.dump_dir is not None and i == 5:
                 with DM(args.dump_dir):
                     DM().set_tag('batch%d'%i)