From 3d23a56ed23aa701a056c3f12568c0b7867f40c9 Mon Sep 17 00:00:00 2001
From: William Falcon <waf2107@columbia.edu>
Date: Thu, 8 Aug 2019 10:59:16 -0400
Subject: [PATCH] make experiment param in trainer optional (#77)

* removed forced exp

* modified test to also run without exp
---
 pytorch_lightning/models/trainer.py | 35 ++++++++++++++++++-----------
 tests/test_models.py                |  6 -----
 2 files changed, 22 insertions(+), 19 deletions(-)

diff --git a/pytorch_lightning/models/trainer.py b/pytorch_lightning/models/trainer.py
index 605f2c78b6..e67a90cbd7 100644
--- a/pytorch_lightning/models/trainer.py
+++ b/pytorch_lightning/models/trainer.py
@@ -50,7 +50,7 @@ def reduce_distributed_output(output, nb_gpus):
 class Trainer(TrainerIO):
 
     def __init__(self,
-                 experiment,
+                 experiment=None,
                  early_stop_callback=None,
                  checkpoint_callback=None,
                  gradient_clip=0,
@@ -122,7 +122,9 @@ class Trainer(TrainerIO):
         self.on_gpu = gpus is not None and torch.cuda.is_available()
         self.progress_bar = progress_bar
         self.experiment = experiment
-        self.exp_save_path = experiment.get_data_path(experiment.name, experiment.version)
+        self.exp_save_path = None
+        if self.experiment is not None:
+            self.exp_save_path = experiment.get_data_path(experiment.name, experiment.version)
         self.cluster = cluster
         self.process_position = process_position
         self.current_gpu_name = current_gpu_name
@@ -312,13 +314,15 @@ class Trainer(TrainerIO):
 
     @property
     def __tng_tqdm_dic(self):
-        # ForkedPdb().set_trace()
         tqdm_dic = {
             'tng_loss': '{0:.3f}'.format(self.avg_loss),
-            'v_nb': '{}'.format(self.experiment.version),
             'epoch': '{}'.format(self.current_epoch),
             'batch_nb': '{}'.format(self.batch_nb),
         }
+
+        if self.experiment is not None:
+            tqdm_dic['v_nb'] = self.experiment.version
+
         tqdm_dic.update(self.tqdm_metrics)
 
         if self.on_gpu:
@@ -462,7 +466,8 @@ dataloader = Dataloader(dataset, sampler=dist_sampler)
         if self.use_ddp:
             # must copy only the meta of the exp so it survives pickle/unpickle
             #  when going to new process
-            self.experiment = self.experiment.get_meta_copy()
+            if self.experiment is not None:
+                self.experiment = self.experiment.get_meta_copy()
 
             if self.is_slurm_managing_tasks:
                 task = int(os.environ['SLURM_LOCALID'])
@@ -564,8 +569,9 @@ We recommend you switch to ddp if you want to use amp
 
         # recover original exp before went into process
         # init in write mode only on proc 0
-        self.experiment.debug = self.proc_rank > 0
-        self.experiment = self.experiment.get_non_ddp_exp()
+        if self.experiment is not None:
+            self.experiment.debug = self.proc_rank > 0
+            self.experiment = self.experiment.get_non_ddp_exp()
 
         # show progbar only on prog_rank 0
         self.prog_bar = self.prog_bar and self.node_rank == 0 and gpu_nb == 0
@@ -575,7 +581,8 @@ We recommend you switch to ddp if you want to use amp
         self.world_size = self.nb_gpu_nodes * len(self.data_parallel_device_ids)
 
         # let the exp know the rank to avoid overwriting logs
-        self.experiment.rank = self.proc_rank
+        if self.experiment is not None:
+            self.experiment.rank = self.proc_rank
 
         # set up server using proc 0's ip address
         # try to init for 20 times at max in case ports are taken
@@ -673,10 +680,12 @@ We recommend you switch to ddp if you want to use amp
 
         # give model convenience properties
         ref_model.trainer = self
-        ref_model.experiment = self.experiment
+
+        if self.experiment is not None:
+            ref_model.experiment = self.experiment
 
         # save exp to get started
-        if self.proc_rank == 0:
+        if self.proc_rank == 0 and self.experiment is not None:
             self.experiment.save()
 
         # track model now.
@@ -756,7 +765,7 @@ We recommend you switch to ddp if you want to use amp
 
                 # when batch should be saved
                 if (batch_nb + 1) % self.log_save_interval == 0 or early_stop_epoch:
-                    if self.proc_rank == 0:
+                    if self.proc_rank == 0 and self.experiment is not None:
                         self.experiment.save()
 
                 # when metrics should be logged
@@ -784,7 +793,7 @@ We recommend you switch to ddp if you want to use amp
                     # log metrics
                     scalar_metrics = self.__metrics_to_scalars(
                         metrics, blacklist=self.__log_vals_blacklist())
-                    if self.proc_rank == 0:
+                    if self.proc_rank == 0 and self.experiment is not None:
                         self.experiment.log(scalar_metrics, global_step=self.global_step)
                         self.experiment.save()
 
@@ -813,7 +822,7 @@ We recommend you switch to ddp if you want to use amp
                 if stop:
                     return
 
-    def __metrics_to_scalars(self, metrics, blacklist=[]):
+    def __metrics_to_scalars(self, metrics, blacklist=set()):
         new_metrics = {}
         for k, v in metrics.items():
             if type(v) is torch.Tensor:
diff --git a/tests/test_models.py b/tests/test_models.py
index c71d302392..cad55ff372 100644
--- a/tests/test_models.py
+++ b/tests/test_models.py
@@ -37,16 +37,10 @@ def test_simple_cpu():
     save_dir = init_save_dir()
 
     # exp file to get meta
-    test_exp_version = 10
-    exp = get_exp(False, version=test_exp_version)
-    exp.argparse(hparams)
-    exp.save()
-
     trainer_options = dict(
         max_nb_epochs=1,
         val_percent_check=0.1,
         train_percent_check=0.1,
-        experiment=exp,
     )
 
     # fit model