Merge branch 'fbopt_output_ma' into fbopt_setpoint_ma_output_ma

README.md

@@ -46,6 +46,23 @@ pip install -r requirements_notorch.txt
 conda install tensorboard
 ```
 
+#### Download PACS
+
+step 1:
+
+use the following script to download PACS to your local laptop and upload it to your cluster
+
+https://github.com/marrlab/DomainLab/blob/fbopt/data/script/download_pacs.py
+
+step 2:
+make a symbolic link following the example script in https://github.com/marrlab/DomainLab/blob/master/sh_pacs.sh
+
+where `mkdir -p data/pacs` is executed under the repository directory, 
+
+`ln -s /dir/to/yourdata/pacs/raw  ./data/pacs/PACS`
+will create a symbolic link under the repository directory
+
+
 #### Windows installation details
 
 To install DomainLab on Windows, please remove the `snakemake` dependency from the `requirements.txt` file.

domainlab/algos/trainers/args_fbopt.py

@@ -22,6 +22,10 @@ def add_args2parser_fbopt(parser):
 
     parser.add_argument('--coeff_ma', type=float, default=0.5,
                         help='exponential moving average')
+
+    parser.add_argument('--coeff_ma_output_state', type=float, default=0.5,
+                        help='setpoint output as state exponential moving average')
+
 
     parser.add_argument('--exp_shoulder_clip', type=float, default=10,
                         help='clip before exponential operation')

domainlab/algos/trainers/fbopt_setpoint_ada.py

@@ -4,6 +4,11 @@
 import torch
 from domainlab.utils.logger import Logger
 
+def list_add(list1, list2):
+    return [a + b for a, b in zip(list1, list2)]
+
+def list_multiply(list1, coeff):
+    return [ele * coeff for ele in list1]
 
 def is_less_list_any(list1, list2):
     """
@@ -36,9 +41,10 @@ def __init__(self, state=None, args=None):
                 state = DominateAnyComponent()
         self.transition_to(state)
         self.ma_epo_reg_loss = None
-        self.state_epo_reg_loss = None
         self.coeff_ma = 0.5  # FIXME
-        self.state_task_loss = None
+        self.state_task_loss = 0.0
+        self.state_epo_reg_loss = [0.0 for _ in range(10)] # FIXME
+        self.coeff_ma_output = args.coeff_ma_output_state
         # initial value will be set via trainer
         self.setpoint4R = None
         self.setpoint4ell = None
@@ -63,8 +69,10 @@ def observe(self, epo_reg_loss, epo_task_loss):
         read current epo_reg_loss continuously
         FIXME: setpoint should also be able to be eliviated
         """
-        self.state_epo_reg_loss = epo_reg_loss
-        self.state_task_loss = epo_task_loss
+        self.state_epo_reg_loss = [self.coeff_ma_output*a + ( 1-self.coeff_ma_output )*b if a != 0.0 else b for a, b in zip(self.state_epo_reg_loss, epo_reg_loss)]
+        if self.state_task_loss == 0.0:
+            self.state_task_loss = epo_task_loss
+        self.state_task_loss = self.coeff_ma_output * self.state_task_loss + (1-self.coeff_ma_output) * epo_task_loss
         if self.state_updater.update_setpoint():
             logger = Logger.get_logger(logger_name='main_out_logger', loglevel="INFO")
             self.update_setpoint_ma(self.state_epo_reg_loss)

run_fbopt_mnist.sh

@@ -3,4 +3,5 @@
 # although garbage collector has been explicitly called, sometimes there is still CUDA out of memory error
 # so it is better not to use GPU to do the pytest to ensure every time there is no CUDA out of memory error occuring
 # pytest -s tests/test_fbopt.py
-python main_out.py --te_d=1 --tr_d 0 3 --task=mnistcolor10 --bs=16 --aname=jigen --trainer=fbopt --nname=conv_bn_pool_2 --epos=2000 --es=50 --mu_init=0.00001
+
+python main_out.py --te_d=1 --tr_d 0 3 --task=mnistcolor10 --bs=16 --aname=jigen --trainer=fbopt --nname=conv_bn_pool_2 --epos=2000 --es=50 --mu_init=0.00001 --coeff_ma_output_state=0.5