added Cudnn roles

.gitignore

@@ -5,3 +5,4 @@ main
 server/docker/__pycache__
 p2p-rendering-computation
 config.json
+*.tgz

.idea/vcs.xml

@@ -2,5 +2,7 @@
 <project version="4">
   <component name="VcsDirectoryMappings">
     <mapping directory="$PROJECT_DIR$" vcs="Git" />
+    <mapping directory="$PROJECT_DIR$/sample-run/tensorflow/roles/ansible-role-cudnn" vcs="Git" />
+    <mapping directory="$PROJECT_DIR$/sample-run/tensorflow/roles/cudnn" vcs="Git" />
   </component>
 </project>

sample-run/tensorflow/hosts

@@ -6,7 +6,7 @@ test:
   hosts:
     test1:
       ansible_host: 0.0.0.0     # Replace with your remote IP
-      ansible_port: 44003          # Replace with your remote SSH port
+      ansible_port: 38775          # Replace with your remote SSH port
       ansible_user: master       # Replace wtih your username
       ansible_ssh_pass: password
       ansible_sudo_pass: password

sample-run/tensorflow/packages.yml

@@ -26,8 +26,11 @@
     #   vars:
     #     users: "{{ secrets.users }}"
 
-    - role: cuda # CUDA Driver Setup
-      tags: [cuda]
+#    - role: cuda # CUDA Driver Setup
+#      tags: [cuda]
+
+    - role: cudnn
+      tags: [cudnn]
 
     # - role: nvidia # NVIDIA-Docker Runtime Setup
     #   tags: [nvidia]

sample-run/tensorflow/roles/cuda/README.md

@@ -1,23 +0,0 @@
-# The `cuda` Role
-
-The `cuda` role installs the Nvidia CUDA GPU repository and drivers.
-
-This role may be executed independently by running:
-
-```bash
-$ ansible-playbook packages.yml --tags "cuda"
-```
-
-On execution, this role installs the following directly to the remote machine's operating system:
-
-### Apt Repositories (Debian/Ubuntu)
-
-- Nvidia CUDA
-
-### Apt GPG Signing Keys (Debian/Ubuntu)
-
-- cudatools <cudatools@nvidia.com>
-
-### Operating System Packages (Debian/Ubuntu)
-
-- cuda v10.x

sample-run/tensorflow/roles/cuda/handlers/main.yml

@@ -1,5 +0,0 @@
----
-# CUDA Driver Setup
-- name: reboot the machine
-  reboot:
-  listen: "reboot the machine"

sample-run/tensorflow/roles/cuda/tasks/debian.yml

@@ -1,22 +0,0 @@
----
-# Cuda Driver Package Installation
-- name: copy the CUDA repo onto the machine
-  copy:
-    src: "{{ cuda.apt.repo }}"
-    dest: "{{ general.working_dir }}/{{ cuda.apt.repo }}"
-
-- name: get the CUDA signing key
-  apt_key:
-    state: present
-    url: "{{ cuda.apt.signing_key_url }}"
-
-- name: add the CUDA repo
-  apt:
-    deb: "{{ general.working_dir }}/{{ cuda.apt.repo }}"
-    state: present
-
-- name: install the CUDA drivers
-  apt:
-    name: "{{ cuda.apt.package }}"
-    update_cache: yes
-  notify: "reboot the machine"

sample-run/tensorflow/roles/cuda/tasks/main.yml

@@ -1,3 +0,0 @@
----
-- import_tasks: debian.yml
-  when: ansible_facts['os_family']|lower == 'debian' # debian, ubuntu

sample-run/tensorflow/roles/cuda/vars/main.yml

@@ -1,10 +0,0 @@
----
-cuda:
-  apt:
-    package: cuda=11.* # known good with 10.0.130-1
-    repo: cuda-11-3_11.3.1-1_amd64.deb # Network install
-    dependency: cuda-runtime-11-3_11.3.1-1_amd64.deb
-    signing_key_url: https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/7fa2af80.pub
-
-general:
-  working_dir: /tmp

sample-run/tensorflow/roles/cudnn/.gitignore

@@ -0,0 +1 @@
+files/

sample-run/tensorflow/roles/cudnn/.travis.yml

@@ -0,0 +1,29 @@
+---
+language: python
+python: "2.7"
+
+# Use the new container infrastructure
+sudo: false
+
+# Install ansible
+addons:
+  apt:
+    packages:
+    - python-pip
+
+install:
+  # Install ansible
+  - pip install ansible
+
+  # Check ansible version
+  - ansible --version
+
+  # Create ansible.cfg with correct roles_path
+  - printf '[defaults]\nroles_path=../' >ansible.cfg
+
+script:
+  # Basic role syntax check
+  - ansible-playbook tests/test.yml -i tests/inventory --syntax-check
+
+notifications:
+  webhooks: https://galaxy.ansible.com/api/v1/notifications/

sample-run/tensorflow/roles/cudnn/README.md

@@ -0,0 +1,43 @@
+Ansible Role: cuDNN
+=========
+
+Install [cuDNN](https://developer.nvidia.com/cudnn) on Linux System.
+
+Requirements
+------------
+
+- [CUDA Toolkit](https://developer.nvidia.com/cuda-toolkit) installed.
+- Download [cuDNN](https://developer.nvidia.com/cudnn) to [`files`](files) folder with your NVIDIA Developer account.
+
+Role Variables
+--------------
+
+`cuda_version`: The version of CUDA Toolkit installed. (default: 8.0)
+
+`cudnn_version`: The version cuDNN. (default: v5.1)
+
+Dependencies
+------------
+
+None.
+
+Example Playbook
+----------------
+
+```yaml
+- hosts: gpu
+  become: yes
+
+  roles:
+    - { role: cudnn, cuda_version: 8.0, cudnn_version: v5.1}
+```
+
+License
+-------
+
+BSD
+
+Author Information
+------------------
+
+This role was created by [Daniel D](https://github.com/djx339).

sample-run/tensorflow/roles/cudnn/defaults/main.yml

@@ -0,0 +1,6 @@
+---
+# defaults file for cudnn
+cuda_version: 8.0
+cuda_version_str: "v{{ cuda_version }}"
+
+cudnn_version: v5.1

sample-run/tensorflow/roles/cudnn/handlers/main.yml

@@ -0,0 +1,2 @@
+---
+# handlers file for cudnn

sample-run/tensorflow/roles/cudnn/meta/main.yml

@@ -0,0 +1,26 @@
+galaxy_info:
+  author: Daniel D
+  description: An engineer
+  company: 
+
+  license: BSD
+
+  min_ansible_version: 1.2
+
+  platforms:
+    - name: Ubuntu
+      versions:
+        - all
+        - trusty
+        - xenial
+    - name: CentOS
+      versions:
+        - all
+    
+
+  galaxy_tags:
+    - cudnn
+    - nvidia
+    - cuda
+
+dependencies: []

sample-run/tensorflow/roles/cudnn/tasks/main.yml

@@ -0,0 +1,15 @@
+---
+# tasks file for cudnn
+- name: Create installer dir
+  file: "path={{ cudnn_installer_dir }} state=directory mode=0755"
+
+- name: Uncompressing cudnn lib
+  unarchive: "src={{ cudnn_installer.local_file }} dest={{ cudnn_installer_dir }}"
+  register: __cudnn_result
+
+- name: Install cudnn
+  shell:  cp -rf {{ item.src }} {{ item.dest }}
+  with_items:
+    - { src: "{{ cudnn_installer_dir }}/cuda/lib64/*", dest: "/usr/local/cuda/lib64/" }
+    - { src: "{{ cudnn_installer_dir }}/cuda/include/cudnn.h", dest: "/usr/local/cuda/include/" }
+  when: __cudnn_result|changed

sample-run/tensorflow/roles/cudnn/tests/inventory

@@ -0,0 +1 @@
+localhost

sample-run/tensorflow/roles/cudnn/tests/test.yml

@@ -0,0 +1,5 @@
+---
+- hosts: localhost
+  remote_user: root
+  roles:
+    - cudnn

sample-run/tensorflow/roles/cudnn/vars/main.yml

@@ -0,0 +1,31 @@
+---
+# vars file for cudnn
+cudnn_installer_map:
+  v11.5:
+    v8.2.0.53:
+      url: https://developer.nvidia.com/rdp/assets/cudnn-11.3-linux-x64-v8.2.53.tgz
+      local_file: cudnn-8.0-linux-x64-v5.1.tgz
+  v8.0:
+    v6.0:
+      url: https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v6/prod/8.0_20170427/cudnn-8.0-linux-x64-v6.0-tgz
+      local_file: cudnn-8.0-linux-x64-v6.0.tgz
+    v5.1:
+      url: https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v5.1/prod_20161129/8.0/cudnn-8.0-linux-x64-v5.1-tgz
+      local_file: cudnn-8.0-linux-x64-v5.1.tgz
+    v5.0:
+      url: https://developer.nvidia.com/rdp/assets/cudnn-8.0-linux-x64-v5.0-ga-tgz
+      local_file: cudnn-8.0-linux-x64-v5.0-ga.tgz
+  v7.5:
+    v6.0:
+      url: https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v6/prod/7.5_20170427/cudnn-7.5-linux-x64-v6.0-tgz
+      local_file: cudnn-7.5-linux-x64-v6.0.tgz
+    v5.1:
+      url: https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v5.1/prod_20161129/7.5/cudnn-7.5-linux-x64-v5.1-tgz
+      local_file: cudnn-7.5-linux-x64-v5.1.tgz
+    v5.0:
+      url: https://developer.nvidia.com/rdp/assets/cudnn-7.5-linux-x64-v5.0-ga.tgz
+      local_file: cudnn-7.5-linux-x64-v5.0-ga.tgz
+
+
+cudnn_installer: "{{ cudnn_installer_map[cuda_version_str][cudnn_version] }}"
+cudnn_installer_dir: "/tmp/cudnn_installer"