Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

remove warning in amp tutorial #15135

Merged
merged 1 commit into from
Jun 3, 2019
Merged

remove warning in amp tutorial #15135

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 2 additions & 11 deletions docs/tutorials/amp/amp_tutorial.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,10 +92,9 @@ train_data = SyntheticDataLoader(data_shape, batch_size)
def get_network():
# SSD with RN50 backbone
net_name = 'ssd_512_resnet50_v1_coco'
net = get_model(net_name, pretrained_base=True, norm_layer=gluon.nn.BatchNorm)
async_net = net
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
warnings.simplefilter("ignore")
net = get_model(net_name, pretrained_base=True, norm_layer=gluon.nn.BatchNorm)
net.initialize()
net.collect_params().reset_ctx(ctx)

Expand All @@ -112,9 +111,6 @@ net = get_network()
net.hybridize(static_alloc=True, static_shape=True)
```

/mxnet/code/python/mxnet/gluon/block.py:1138: UserWarning: Cannot decide type for the following arguments. Consider providing them as input:
data: None
input_sym_arg_type = in_param.infer_type()[0]


Next, we need to create a Gluon Trainer.
Expand Down Expand Up @@ -192,11 +188,6 @@ net = get_network()
net.hybridize(static_alloc=True, static_shape=True)
```

/mxnet/code/python/mxnet/gluon/block.py:1138: UserWarning: Cannot decide type for the following arguments. Consider providing them as input:
data: None
input_sym_arg_type = in_param.infer_type()[0]


For some models that may be enough to start training in mixed precision, but the full FP16 recipe recommends using dynamic loss scaling to guard against over- and underflows of FP16 values. Therefore, as a next step, we create a trainer and initialize it with support for AMP's dynamic loss scaling. Currently, support for dynamic loss scaling is limited to trainers created with `update_on_kvstore=False` option, and so we add it to our trainer initialization.


Expand Down