Commited code from 2024

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2024 TastyPancakes. 
+Copyright (c) 2024 Eduardo Cueto-Mendoza.
 
 Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
 

README.md

@@ -1,3 +1,7 @@
 # Energy efficiency comparison
 
 This experiment compares a Frequentist CNN model against a Bayesian CNN model
+
+## Example run command
+
+python run_service.py -f1 -s -e && sleep 60 && python run_service.py -f2 -s -e && sleep 60 && python run_service.py -f3 -s -e && sleep 60 && python run_service.py -f4 -s -e && sleep 60 && python run_service.py -f5 -s -e && sleep 60 & python run_service.py -f6 -s -e && sleep 60 && python run_service.py -f7 -s -e && sleep 60 && python run_service.py -b1 -s -e && sleep 60 && python run_service.py -b2 -s -e && sleep 60 && python run_service.py -b3 -s -e && sleep 60 && python run_service.py -b4 -s -e && sleep 60 && python run_service.py -b5 -s -e && sleep 60 && python run_service.py -b6 -s -e && sleep 60 && python run_service.py -b7 -s -e && sleep 60

arguments.py

@@ -22,6 +22,6 @@ def makeArguments(arguments: ArgumentParser) -> dict:
                           help="Save model")
     all_args.add_argument('--net_type', default='lenet', type=str,
                           help='model = [lenet/AlexNet/3Conv3FC]')
-    all_args.add_argument('--dataset', default='CIFAR10', type=str,
+    all_args.add_argument('--dataset', default='MNIST', type=str,
                           help='dataset = [MNIST/CIFAR10/CIFAR100]')
     return vars(all_args.parse_args())

main_bayesian.py

@@ -4,7 +4,7 @@ import os
 import data
 import utils
 import torch
-import pickle
+# import pickle
 import metrics
 import numpy as np
 from datetime import datetime
@@ -15,12 +15,41 @@ from models.BayesianModels.BayesianAlexNet import BBBAlexNet
 from models.BayesianModels.Bayesian3Conv3FC import BBB3Conv3FC
 from stopping_crit import earlyStopping, energyBound, accuracyBound
 
-with (open("configuration.pkl", "rb")) as file:
+# with (open("configuration.pkl", "rb")) as file:
-    while True:
+#     while True:
-        try:
+#         try:
-            cfg = pickle.load(file)
+#             cfg = pickle.load(file)
-        except EOFError:
+#         except EOFError:
-            break
+#             break
+
+cfg = {
+        "model": {"net_type": "lenet", "type": "bayes", "size": 1,
+                  "layer_type": "lrt", "activation_type": "softplus",
+                  "priors": {
+                      'prior_mu': 0,
+                      'prior_sigma': 0.1,
+                      'posterior_mu_initial': (0, 0.1),  # (mean,std) normal_
+                      'posterior_rho_initial': (-5, 0.1),  # (mean,std) normal_
+                      },
+                  "n_epochs": 100,
+                  "sens": 1e-9,
+                  "energy_thrs": 100000,
+                  "acc_thrs": 0.99,
+                  "lr": 0.001,
+                  "num_workers": 4,
+                  "valid_size": 0.2,
+                  "batch_size": 256,
+                  "train_ens": 1,
+                  "valid_ens": 1,
+                  "beta_type": 0.1,  # 'Blundell','Standard',etc.
+                  # Use float for const value
+                  },
+        #"data": "CIFAR10",
+        "data": "MNIST",
+        "stopping_crit": 1,
+        "save": 1,
+        "pickle_path": None,
+}
 
 
 # CUDA settings
@@ -126,8 +155,7 @@ def run(dataset, net_type):
                    activation_type).to(device)
 
     ckpt_dir = f'checkpoints/{dataset}/bayesian'
-    ckpt_name = f'checkpoints/{dataset}/bayesian/model_{net_type}_{layer_type}\
+    ckpt_name = f'checkpoints/{dataset}/bayesian/model_{net_type}_{layer_type}_{activation_type}_{cfg["model"]["size"]}'
-            _{activation_type}_{cfg["model"]["size"]}.pt'
 
     if not os.path.exists(ckpt_dir):
         os.makedirs(ckpt_dir, exist_ok=True)
@@ -178,18 +206,23 @@ def run(dataset, net_type):
                 break
         elif stp == 4:
             print('Using accuracy bound')
-            if accuracyBound(train_acc, cfg.acc_thrs) == 1:
+            if accuracyBound(train_acc, cfg["model"]["acc_thrs"]) == 1:
                 break
         else:
             print('Training for {} epochs'.format(cfg["model"]["n_epochs"]))
 
         if sav == 1:
             # save model when finished
-            if epoch == cfg.n_epochs-1:
+            # if epoch == cfg["model"]["n_epochs"]-1:
-                torch.save(net.state_dict(), ckpt_name)
+            torch.save({
+                'epoch': epoch,
+                'model_state_dict': net.state_dict(),
+                'optimizer_state_dict': optimizer.state_dict(),
+                'loss': train_loss,
+                }, ckpt_name + '_epoch_{}.pt'.format(epoch))
 
-    with open("bayes_exp_data_"+str(cfg["model"]["size"])+".pkl", 'wb') as f:
+    # with open("bayes_exp_data_"+str(cfg["model"]["size"])+".pkl", 'wb') as f:
-        pickle.dump(train_data, f)
+    #     pickle.dump(train_data, f)
 
 
 if __name__ == '__main__':

main_frequentist.py

@@ -2,7 +2,7 @@ from __future__ import print_function
 import os
 import data
 import torch
-import pickle
+# import pickle
 import metrics
 import numpy as np
 import torch.nn as nn
@@ -13,12 +13,41 @@ from models.NonBayesianModels.AlexNet import AlexNet
 from stopping_crit import earlyStopping, energyBound, accuracyBound
 from models.NonBayesianModels.ThreeConvThreeFC import ThreeConvThreeFC
 
-with (open("configuration.pkl", "rb")) as file:
+# with (open("configuration.pkl", "rb")) as file:
-    while True:
+#     while True:
-        try:
+#         try:
-            cfg = pickle.load(file)
+#             cfg = pickle.load(file)
-        except EOFError:
+#         except EOFError:
-            break
+#             break
+
+cfg = {
+        "model": {"net_type": "lenet", "type": "freq", "size": 1,
+                  "layer_type": "lrt", "activation_type": "softplus",
+                  "priors": {
+                      'prior_mu': 0,
+                      'prior_sigma': 0.1,
+                      'posterior_mu_initial': (0, 0.1),  # (mean,std) normal_
+                      'posterior_rho_initial': (-5, 0.1),  # (mean,std) normal_
+                      },
+                  "n_epochs": 100,
+                  "sens": 1e-9,
+                  "energy_thrs": 100000,
+                  "acc_thrs": 0.99,
+                  "lr": 0.001,
+                  "num_workers": 4,
+                  "valid_size": 0.2,
+                  "batch_size": 256,
+                  "train_ens": 1,
+                  "valid_ens": 1,
+                  "beta_type": 0.1,  # 'Blundell','Standard',etc.
+                  # Use float for const value
+                  },
+        #"data": "CIFAR10",
+        "data": "MNIST",
+        "stopping_crit": 1,
+        "save": 1,
+        "pickle_path": None,
+}
 
 # CUDA settings
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
@@ -80,8 +109,7 @@ def run(dataset, net_type):
     net = getModel(net_type, inputs, outputs).to(device)
 
     ckpt_dir = f'checkpoints/{dataset}/frequentist'
-    ckpt_name = f'checkpoints/{dataset}/frequentist/model\
+    ckpt_name = f'checkpoints/{dataset}/frequentist/model_{net_type}_{cfg["model"]["size"]}'
-            _{net_type}_{cfg["model"]["size"]}.pt'
 
     if not os.path.exists(ckpt_dir):
         os.makedirs(ckpt_dir, exist_ok=True)
@@ -132,11 +160,17 @@ def run(dataset, net_type):
 
         if sav == 1:
             # save model when finished
-            if epoch == n_epochs:
+            # if epoch == n_epochs:
-                torch.save(net.state_dict(), ckpt_name)
+            #     torch.save(net.state_dict(), ckpt_name)
+            torch.save({
+                'epoch': epoch,
+                'model_state_dict': net.state_dict(),
+                'optimizer_state_dict': optimizer.state_dict(),
+                'loss': train_loss,
+                }, ckpt_name + '_epoch_{}.pt'.format(epoch))
 
-    with open("freq_exp_data_"+str(cfg["model"]["size"])+".pkl", 'wb') as f:
+#    with open("freq_exp_data_"+str(cfg["model"]["size"])+".pkl", 'wb') as f:
-        pickle.dump(train_data, f)
+#        pickle.dump(train_data, f)
 
 
 if __name__ == '__main__':

run_service.py

@@ -24,7 +24,7 @@ cfg = {
                       },
                   "n_epochs": 100,
                   "sens": 1e-9,
-                  "energy_thrs": 10000,
+                  "energy_thrs": 100000,
                   "acc_thrs": 0.99,
                   "lr": 0.001,
                   "num_workers": 4,