Skip to content

Commit 33f2778

Browse files
bdmoore1bdmoore1
authored
Rearranged README to have a local section and DevCloud Section (oneapi-src#528)
1 parent d3795ad commit 33f2778

File tree

1 file changed

+62
-38
lines changed
  • AI-and-Analytics/Features-and-Functionality/IntelTensorFlow_PerformanceAnalysis

1 file changed

+62
-38
lines changed

‎AI-and-Analytics/Features-and-Functionality/IntelTensorFlow_PerformanceAnalysis/README.md

Lines changed: 62 additions & 38 deletions
Original file line numberDiff line numberDiff line change
@@ -34,10 +34,10 @@ Third party program Licenses can be found here: [third-party-programs.txt](https
3434

3535
## Build and Run the Sample
3636

37-
### Pre-requirement
37+
## Running Samples on the Intel® DevCloud
38+
If you are running this sample on the DevCloud, see [Running Samples on the Intel® DevCloud](#run-samples-on-devcloud)
3839

39-
> NOTE: No action is required if users use Intel DevCloud as their environment.
40-
Please refer to [Intel oneAPI DevCloud](https://intelsoftwaresites.secure.force.com/devcloud/oneapi) for Intel DevCloud.
40+
### Pre-requirements
4141

4242
1. **Intel AI Analytics Toolkit**
4343
You can refer to the oneAPI [main page](https://software.intel.com/en-us/oneapi) for toolkit installation,
@@ -47,39 +47,8 @@ Third party program Licenses can be found here: [third-party-programs.txt](https
4747
Users can install via PIP by `$pip install notebook`.
4848
Users can also refer to the [installation link](https://jupyter.org/install) for details.
4949

50-
#### **Conda Environment Creation**
50+
#### **Conda Environment Creation (Local Installation)**
5151

52-
##### **1. Intel oneAPI DevCloud**
53-
---
54-
###### **Stock TensorFlow**
55-
56-
1. Create conda env: `$conda create -n stock-tensorflow python matplotlib ipykernel psutil pandas gitpython`
57-
2. Activate the created conda env: `$source activate stock-tensorflow.`
58-
3. Install stock Tensorflow with a specific version: `(stock-tensorflow) $pip install tensorflow==2.3.0`
59-
4. Install extra needed package: `(stock-tensorflow) $pip install cxxfilt`
60-
5. Deactivate conda env: `(stock-tensorflow)$conda deactivate`
61-
6. Register the kernel to Jupyter NB: `$~/.conda/envs/stock-tensorflow/bin/python -m ipykernel install --user --name=stock-tensorflow`
62-
63-
> NOTE: Please change the python path if you have a different folder path for anaconda3.
64-
After profiling, users can remove the kernel from Jupyter NB with `$jupyter kernelspec uninstall stock-tensorflow.`
65-
66-
###### **Intel TensorFlow**
67-
68-
> NOTE: Intel-optimized Tensorflow is on DevCloud. However, users don't have access to install extra packages.
69-
Therefore, we need to clone Intel Tensorflow into the user's home directory for installing extra packages.
70-
71-
1. Source oneAPI environment variables: `$source /opt/intel/oneapi/setvars.sh`
72-
2. Create conda env: `$conda create --name intel-tensorflow --clone tensorflow`
73-
3. Activate the created conda env: `$source activate intel-tensorflow`
74-
4 Install the extra needed package: `(intel-tensorflow) $pip install cxxfilt matplotlib ipykernel psutil pandas gitpython`
75-
5. Deactivate conda env: `(intel-tensorflow)$conda deactivate`
76-
6. Register the kernel to Jupyter NB: `$~/.conda/envs/intel-tensorflow/bin/python -m ipykernel install --user --name=intel-tensorflow`
77-
78-
> NOTE: Please change the python path if you have a different folder path for anaconda3.
79-
After profiling, users can remove the kernel from Jupyter NB with `$jupyter kernelspec uninstall intel-tensorflow.`
80-
81-
##### **2. Linux with Intel oneAPI AI Analytics Toolkit**
82-
---
8352
###### **Stock TensorFlow**
8453

8554
1. Create conda env: `$conda create -n stock-tensorflow python matplotlib ipykernel psutil pandas gitpython`
@@ -106,7 +75,7 @@ Third party program Licenses can be found here: [third-party-programs.txt](https
10675
After profiling, users can remove the kernel from Jupyter NB with `$jupyter kernelspec uninstall intel-tensorflow.`
10776

10877

109-
### Running the Sample
78+
### Running the Sample (local installation)
11079

11180
1. Copy the Intel Model Zoo from your AI Analytics Toolkit installation path: `$cp -rf /opt/intel/oneapi/modelzoo/latest/models ~/`
11281
2. cd ~/models; git init; git add . ; git commit -m 'initial commit'
@@ -123,11 +92,66 @@ Third party program Licenses can be found here: [third-party-programs.txt](https
12392

12493
> NOTE: To compare stock and Intel-optimized TF results in the section "Analyze TF Timeline results among Stock and Intel Tensorflow," users need to run all cells before the comparison section with both stock-tensorflow and intel-tensorflow kernels.
12594
95+
### **Running Samples on the Intel&reg; DevCloud (Optional)<a name="run-samples-on-devcloud"></a>**
96+
97+
#### **Conda Environment Creation (DevCloud)**
98+
99+
##### **Stock TensorFlow**
100+
101+
1. Create conda env: `$conda create -n stock-tensorflow python matplotlib ipykernel psutil pandas gitpython`
102+
2. Activate the created conda env: `$source activate stock-tensorflow.`
103+
3. Install stock Tensorflow with a specific version: `(stock-tensorflow) $pip install tensorflow==2.3.0`
104+
4. Install extra needed package: `(stock-tensorflow) $pip install cxxfilt`
105+
5. Deactivate conda env: `(stock-tensorflow)$conda deactivate`
106+
6. Register the kernel to Jupyter NB: `$~/.conda/envs/stock-tensorflow/bin/python -m ipykernel install --user --name=stock-tensorflow`
107+
108+
> NOTE: Please change the python path if you have a different folder path for anaconda3.
109+
After profiling, users can remove the kernel from Jupyter NB with `$jupyter kernelspec uninstall stock-tensorflow.`
110+
111+
##### **Intel TensorFlow**
112+
113+
> NOTE: Intel-optimized Tensorflow is on DevCloud. However, users don't have access to install extra packages.
114+
Therefore, we need to clone Intel Tensorflow into the user's home directory for installing extra packages.
115+
116+
1. Source oneAPI environment variables: `$source /opt/intel/oneapi/setvars.sh`
117+
2. Create conda env: `$conda create --name intel-tensorflow --clone tensorflow`
118+
3. Activate the created conda env: `$source activate intel-tensorflow`
119+
4 Install the extra needed package: `(intel-tensorflow) $pip install cxxfilt matplotlib ipykernel psutil pandas gitpython`
120+
5. Deactivate conda env: `(intel-tensorflow)$conda deactivate`
121+
6. Register the kernel to Jupyter NB: `$~/.conda/envs/intel-tensorflow/bin/python -m ipykernel install --user --name=intel-tensorflow`
122+
123+
> NOTE: Please change the python path if you have a different folder path for anaconda3.
124+
After profiling, users can remove the kernel from Jupyter NB with `$jupyter kernelspec uninstall intel-tensorflow.`
125+
126+
### **Run in Interactive Mode**
127+
1. Copy the Intel Model Zoo from your AI Analytics Toolkit installation path: `$cp -rf /opt/intel/oneapi/modelzoo/latest/models ~/`
128+
2. Launch Jupyter notebook: `$jupyter notebook --ip=0.0.0.0`
129+
3. Follow the instructions to open the URL with the token in your browser
130+
4. Browse to the `models/docs/notebooks/perf_analysis` folder
131+
5. Click the `benchmark_perf_comparison.ipynb` or `benchmark_perf_timeline_analysis.ipynb` file
132+
6. Change your Jupyter notebook kernel to either "stock-tensorflow" or "intel-tensorflow" (highlighted in the diagram below)
133+
<br><img src="images/jupyter_kernels.png" width="300" height="300"><br>
134+
7. Run through every cell of the notebook one by one
135+
136+
> NOTE: To compare stock and Intel-optimized TF results in the section "Analyze TF Timeline results among Stock and Intel Tensorflow," users need to run all cells before the comparison section with both stock-tensorflow and intel-tensorflow kernels.
137+
138+
### **Request a Compute Node**
139+
In order to run on the DevCloud, you need to request a compute node using node properties such as: `gpu`, `xeon`, `fpga_compile`, `fpga_runtime` and others. For more information about the node properties, execute the `pbsnodes` command.
140+
This node information must be provided when submitting a job to run your sample in batch mode using the qsub command. When you see the qsub command in the Run section of the [Hello World instructions](https://devcloud.intel.com/oneapi/get_started/aiAnalyticsToolkitSamples/), change the command to fit the node you are using. Nodes which are in bold indicate they are compatible with this sample:
141+
142+
<!---Mark each compatible Node in BOLD-->
143+
| Node | Command |
144+
| ----------------- | ------------------------------------------------------- |
145+
| GPU | qsub -l nodes=1:gpu:ppn=2 -d . hello-world.sh |
146+
| CPU | qsub -l nodes=1:xeon:ppn=2 -d . hello-world.sh |
147+
| FPGA Compile Time | qsub -l nodes=1:fpga\_compile:ppn=2 -d . hello-world.sh |
148+
| FPGA Runtime | qsub -l nodes=1:fpga\_runtime:ppn=2 -d . hello-world.sh |
149+
126150
### Example of Output
127-
Users should be able to see some diagrams for performance comparison and analysis.
151+
Users should be able to see some diagrams for performance comparison and analysis.
128152
One example of performance comparison diagrams:
129153
<br><img src="images/perf_comparison.png" width="400" height="300"><br>
130154

131-
For performance analysis, users can also see pie charts for top hotspots of Tensorflow* operations among Stock and Intel Tensorflow.
155+
For performance analysis, users can also see pie charts for top hotspots of Tensorflow* operations among Stock and Intel Tensorflow.
132156
One example of performance analysis diagrams:
133157
<br><img src="images/compared_tf_op_duration_pie.png" width="900" height="400"><br>

0 commit comments

Comments
 (0)