Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Update ExternalSource framework examples #2482

Merged
merged 4 commits into from
Nov 19, 2020
Merged

Update ExternalSource framework examples #2482

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
935e36c
Fix build with Clang 6.
mzient Nov 19, 2020
254f361
Update PyTorch ExternalSource example
mzient Nov 18, 2020
61f57d7
Updated MXNet and PaddlePaddle examples.
mzient Nov 19, 2020
6f4b0a9
Restore ExecutionCount.
mzient Nov 19, 2020
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
4 changes: 2 additions & 2 deletions dali/pipeline/data/tensor_vector.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -436,8 +436,8 @@ void TensorVector<Backend>::update_view(int idx) {
}


template class TensorVector<CPUBackend>;
template class TensorVector<GPUBackend>;
template class DLL_PUBLIC TensorVector<CPUBackend>;
template class DLL_PUBLIC TensorVector<GPUBackend>;
Comment on lines +439 to +440
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Without it debug build with Clang 6 didn't link properly.

template void TensorVector<CPUBackend>::Copy<CPUBackend>(const TensorVector<CPUBackend>&, cudaStream_t); // NOLINT
template void TensorVector<CPUBackend>::Copy<GPUBackend>(const TensorVector<GPUBackend>&, cudaStream_t); // NOLINT
template void TensorVector<GPUBackend>::Copy<CPUBackend>(const TensorVector<CPUBackend>&, cudaStream_t); // NOLINT
Expand Down
77 changes: 23 additions & 54 deletions docs/examples/frameworks/mxnet/mxnet-external_input.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@
},
{
"cell_type": "code",
"execution_count": 1,
"execution_count": 5,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

?

"metadata": {},
"outputs": [],
"source": [
Expand All @@ -24,8 +24,9 @@
"import numpy as np\n",
"from random import shuffle\n",
"from nvidia.dali.pipeline import Pipeline\n",
"import nvidia.dali.ops as ops \n",
"import nvidia.dali.types as types\n",
"import nvidia.dali as dali\n",
"import nvidia.dali.fn as fn\n",
"import mxnet\n",
"\n",
"batch_size = 3\n",
"epochs = 3"
Expand All @@ -40,7 +41,7 @@
},
{
"cell_type": "code",
"execution_count": 2,
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
Expand Down Expand Up @@ -72,17 +73,15 @@
"\n",
" for _ in range(self.batch_size):\n",
" jpeg_filename, label = self.files[self.i].split(' ')\n",
" f = open(self.images_dir + jpeg_filename, 'rb')\n",
" batch.append(np.frombuffer(f.read(), dtype = np.uint8))\n",
" labels.append(np.array([label], dtype = np.uint8))\n",
" batch.append(np.fromfile(self.images_dir + jpeg_filename, dtype = np.uint8)) # we can use numpy\n",
" labels.append(mxnet.ndarray.array([int(label)], dtype = 'uint8')) # or MXNet native arrays\n",
" self.i = (self.i + 1) % self.n\n",
" return (batch, labels)\n",
"\n",
" @property\n",
" def size(self,):\n",
" def __len__(self):\n",
" return self.data_set_len\n",
"\n",
" next = __next__\n"
" next = __next__"
]
},
{
Expand All @@ -91,46 +90,24 @@
"source": [
"### Defining the pipeline\n",
"\n",
"Now the pipeline itself will be defined. First of all, a framework iterator will be used so we need to make sure that images and the output of the pipeline are uniforms in size, so resize operator is used. Also, `iter_setup` will raise the StopIteration exception when the AdvancedExternalInputIterator run of data. Worth notice is that iterator needs to be recreated so next time `iter_setup` is called it has ready data to consume."
"Now let's define our pipeline. We need an instance of ``Pipeline`` class and some operators which will define the processing graph. Our external source provides 2 outpus which we can conveniently unpack by specifying ``num_outputs=2`` in the external source operator."
]
},
{
"cell_type": "code",
"execution_count": 3,
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"class ExternalSourcePipeline(Pipeline):\n",
" def __init__(self, batch_size, num_threads, device_id, external_data):\n",
" super(ExternalSourcePipeline, self).__init__(batch_size,\n",
" num_threads,\n",
" device_id,\n",
" seed=12)\n",
" self.input = ops.ExternalSource()\n",
" self.input_label = ops.ExternalSource()\n",
" self.decode = ops.ImageDecoder(device = \"mixed\", output_type = types.RGB)\n",
" self.res = ops.Resize(device=\"gpu\", resize_x=240, resize_y=240)\n",
" self.cast = ops.Cast(device = \"gpu\",\n",
" dtype = types.UINT8)\n",
" self.external_data = external_data\n",
" self.iterator = iter(self.external_data)\n",
"\n",
" def define_graph(self):\n",
" self.jpegs = self.input()\n",
" self.labels = self.input_label()\n",
" images = self.decode(self.jpegs)\n",
" images = self.res(images)\n",
" output = self.cast(images)\n",
" return (output, self.labels)\n",
"\n",
" def iter_setup(self):\n",
" try:\n",
" (images, labels) = self.iterator.next()\n",
" self.feed_input(self.jpegs, images)\n",
" self.feed_input(self.labels, labels)\n",
" except StopIteration:\n",
" self.iterator = iter(self.external_data)\n",
" raise StopIteration"
"def ExternalSourcePipeline(batch_size, num_threads, device_id, external_data):\n",
" pipe = Pipeline(batch_size, num_threads, device_id)\n",
" with pipe:\n",
" jpegs, labels = fn.external_source(source=external_data, num_outputs=2)\n",
" images = fn.image_decoder(jpegs, device=\"mixed\")\n",
" images = fn.resize(images, resize_x=240, resize_y=240)\n",
" output = fn.cast(images, dtype=dali.types.UINT8)\n",
" pipe.set_outputs(output, labels)\n",
" return pipe"
]
},
{
Expand All @@ -148,17 +125,9 @@
},
{
"cell_type": "code",
"execution_count": 4,
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/usr/local/lib/python3.6/dist-packages/nvidia/dali/plugin/base_iterator.py:124: Warning: Please set `reader_name` and don't set last_batch_padded and size manually whenever possible. This may lead, in some situations, to miss some samples or return duplicated ones. Check the Sharding section of the documentation for more details.\n",
" _iterator_deprecation_warning()\n"
]
},
{
"name": "stdout",
"output_type": "stream",
Expand Down Expand Up @@ -194,7 +163,7 @@
"eii = ExternalInputIterator(batch_size, 0, 1)\n",
"pipe = ExternalSourcePipeline(batch_size=batch_size, num_threads=2, device_id = 0,\n",
" external_data = eii)\n",
"pii = MXNetIterator(pipe, size=eii.size, last_batch_padded=True, last_batch_policy=LastBatchPolicy.PARTIAL)\n",
"pii = MXNetIterator(pipe, size=len(eii), last_batch_padded=True, last_batch_policy=LastBatchPolicy.PARTIAL)\n",
"\n",
"for e in range(epochs):\n",
" for i, data in enumerate(pii):\n",
Expand All @@ -219,7 +188,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.6.9"
"version": "3.6.10"
}
},
"nbformat": 4,
Expand Down
70 changes: 23 additions & 47 deletions docs/examples/frameworks/paddle/paddle-external_input.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,8 +24,8 @@
"import numpy as np\n",
"from random import shuffle\n",
"from nvidia.dali.pipeline import Pipeline\n",
"import nvidia.dali.ops as ops \n",
"import nvidia.dali.types as types\n",
"import nvidia.dali as dali\n",
"import nvidia.dali.fn as fn\n",
"\n",
"batch_size = 3\n",
"epochs = 3"
Expand Down Expand Up @@ -72,17 +72,15 @@
"\n",
" for _ in range(self.batch_size):\n",
" jpeg_filename, label = self.files[self.i].split(' ')\n",
" f = open(self.images_dir + jpeg_filename, 'rb')\n",
" batch.append(np.frombuffer(f.read(), dtype = np.uint8))\n",
" labels.append(np.array([label], dtype = np.uint8))\n",
" batch.append(np.fromfile(self.images_dir + jpeg_filename, dtype = np.uint8))\n",
" labels.append(np.uint8([int(label)]))\n",
" self.i = (self.i + 1) % self.n\n",
" return (batch, labels)\n",
"\n",
" @property\n",
" def size(self,):\n",
" def __len__(self):\n",
" return self.data_set_len\n",
"\n",
" next = __next__\n"
" next = __next__"
]
},
{
Expand All @@ -91,7 +89,7 @@
"source": [
"### Defining the pipeline\n",
"\n",
"Now the pipeline itself will be defined. First of all, a framework iterator will be used so we need to make sure that images and the output of the pipeline are uniforms in size, so resize operator is used. Also, `iter_setup` will raise the StopIteration exception when the AdvancedExternalInputIterator run of data. Worth notice is that iterator needs to be recreated so next time `iter_setup` is called it has ready data to consume."
"Now let's define our pipeline. We need an instance of ``Pipeline`` class and some operators which will define the processing graph. Our external source provides 2 outpus which we can conveniently unpack by specifying ``num_outputs=2`` in the external source operator."
]
},
{
Expand All @@ -100,37 +98,15 @@
"metadata": {},
"outputs": [],
"source": [
"class ExternalSourcePipeline(Pipeline):\n",
" def __init__(self, batch_size, num_threads, device_id, external_data):\n",
" super(ExternalSourcePipeline, self).__init__(batch_size,\n",
" num_threads,\n",
" device_id,\n",
" seed=12)\n",
" self.input = ops.ExternalSource()\n",
" self.input_label = ops.ExternalSource()\n",
" self.decode = ops.ImageDecoder(device = \"mixed\", output_type = types.RGB)\n",
" self.res = ops.Resize(device=\"gpu\", resize_x=240, resize_y=240)\n",
" self.cast = ops.Cast(device = \"gpu\",\n",
" dtype = types.UINT8)\n",
" self.external_data = external_data\n",
" self.iterator = iter(self.external_data)\n",
"\n",
" def define_graph(self):\n",
" self.jpegs = self.input()\n",
" self.labels = self.input_label()\n",
" images = self.decode(self.jpegs)\n",
" images = self.res(images)\n",
" output = self.cast(images)\n",
" return (output, self.labels)\n",
"\n",
" def iter_setup(self):\n",
" try:\n",
" (images, labels) = self.iterator.next()\n",
" self.feed_input(self.jpegs, images)\n",
" self.feed_input(self.labels, labels)\n",
" except StopIteration:\n",
" self.iterator = iter(self.external_data)\n",
" raise StopIteration"
"def ExternalSourcePipeline(batch_size, num_threads, device_id, external_data):\n",
" pipe = Pipeline(batch_size, num_threads, device_id)\n",
" with pipe:\n",
" jpegs, labels = fn.external_source(source=external_data, num_outputs=2)\n",
" images = fn.image_decoder(jpegs, device=\"mixed\")\n",
" images = fn.resize(images, resize_x=240, resize_y=240)\n",
" output = fn.cast(images, dtype=dali.types.UINT8)\n",
" pipe.set_outputs(output, labels)\n",
" return pipe"
]
},
{
Expand All @@ -148,7 +124,7 @@
},
{
"cell_type": "code",
"execution_count": 4,
"execution_count": 5,
"metadata": {},
"outputs": [
{
Expand Down Expand Up @@ -186,7 +162,7 @@
"eii = ExternalInputIterator(batch_size, 0, 1)\n",
"pipe = ExternalSourcePipeline(batch_size=batch_size, num_threads=2, device_id = 0,\n",
" external_data = eii)\n",
"pii = PaddleIterator(pipe, size=eii.size, last_batch_padded=True, last_batch_policy=LastBatchPolicy.PARTIAL)\n",
"pii = PaddleIterator(pipe, size=len(eii), last_batch_padded=True, last_batch_policy=LastBatchPolicy.PARTIAL)\n",
"\n",
"for e in range(epochs):\n",
" for i, data in enumerate(pii):\n",
Expand All @@ -197,21 +173,21 @@
],
"metadata": {
"kernelspec": {
"display_name": "Python 2",
"display_name": "Python 3",
"language": "python",
"name": "python2"
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 2
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython2",
"version": "2.7.15+"
"pygments_lexer": "ipython3",
"version": "3.6.10"
}
},
"nbformat": 4,
Expand Down
Loading