From 215a67e7af2d3d7a036f8edf23e192056645a0db Mon Sep 17 00:00:00 2001
From: choosewhatulike <1901722105@qq.com>
Date: Thu, 26 Jul 2018 22:45:25 +0800
Subject: [PATCH] fix optimizer

---
 fastNLP/action/optimizer.py |  5 +++++
 fastNLP/action/optimizor.py | 50 ---------------------------------------------
 2 files changed, 5 insertions(+), 50 deletions(-)
 create mode 100644 fastNLP/action/optimizer.py
 delete mode 100644 fastNLP/action/optimizor.py

diff --git a/fastNLP/action/optimizer.py b/fastNLP/action/optimizer.py
new file mode 100644
index 00000000..b493e3f0
--- /dev/null
+++ b/fastNLP/action/optimizer.py
@@ -0,0 +1,5 @@
+'''
+use optimizer from Pytorch
+'''
+
+from torch.optim import *
\ No newline at end of file
diff --git a/fastNLP/action/optimizor.py b/fastNLP/action/optimizor.py
deleted file mode 100644
index becdc499..00000000
--- a/fastNLP/action/optimizor.py
+++ /dev/null
@@ -1,50 +0,0 @@
-from torch import optim
-
-
-def get_torch_optimizer(params, alg_name='sgd', **args):
-    """
-    construct PyTorch optimizer by algorithm's name
-    optimizer's arguments can be specified, for different optimizer's arguments, please see PyTorch doc
-
-    usage:
-        optimizer = get_torch_optimizer(model.parameters(), 'SGD', lr=0.01)
-
-    """
-
-    name = alg_name.lower()
-    if name == 'adadelta':
-        return optim.Adadelta(params, **args)
-    elif name == 'adagrad':
-        return optim.Adagrad(params, **args)
-    elif name == 'adam':
-        return optim.Adam(params, **args)
-    elif name == 'adamax':
-        return optim.Adamax(params, **args)
-    elif name == 'asgd':
-        return optim.ASGD(params, **args)
-    elif name == 'lbfgs':
-        return optim.LBFGS(params, **args)
-    elif name == 'rmsprop':
-        return optim.RMSprop(params, **args)
-    elif name == 'rprop':
-        return optim.Rprop(params, **args)
-    elif name == 'sgd':
-        # SGD's parameter lr is required
-        if 'lr' not in args:
-            args['lr'] = 0.01
-        return optim.SGD(params, **args)
-    elif name == 'sparseadam':
-        return optim.SparseAdam(params, **args)
-    else:
-        raise TypeError('no such optimizer named {}'.format(alg_name))
-
-
-if __name__ == '__main__':
-    from torch.nn.modules import Linear
-
-    net = Linear(2, 5)
-
-    test1 = get_torch_optimizer(net.parameters(), 'adam', lr=1e-2, weight_decay=1e-3)
-    print(test1)
-    test2 = get_torch_optimizer(net.parameters(), 'SGD')
-    print(test2)