Skip to content

Latest commit

 

History

History
338 lines (248 loc) · 9.75 KB

README.md

File metadata and controls

338 lines (248 loc) · 9.75 KB
Raw

Jaseci Quick Start for WSL using Ubuntu - Not Official Docs

jaseci-quick-start

Install Jaseci and dependencies

To run commands for Jaseci we need a terminal that accepts bash arguments.

  • The official Jaseci docs recommend using the Ubuntu terminal that comes as the default with WSL.
  • To install WSL, first check if WSL is installed by running the following command in Windows powershell terminal,
wsl -l -v
  • If nothing is returned like this,
  NAME          STATE           VERSION
* kali-linux    Stopped         2
  Ubuntu        Running         2
  • Run this in Windows powershell terminal to install WSL,
wsl --install
  • Next restart your computer and open Ubuntu terminal
  • Once opened again, make sure Ubuntu is updated, and feel free to do this regularly,
sudo apt update && sudo apt upgrade

Add a new user to Ubuntu if you haven't already

sudo adduser your-username-here
  • It will prompt you to enter a password, and you will not be able to see the password as you enter it, but it is there!
  • Once you have created your new user home, you can login using the login command,
login your-username-here
  • Grant root permissions for this new user, use visudo to edit the sudoers file
root@YourComputerName:~# visudo
  • Use down arrow key to scroll to the following section:
# User privilege specification
root    ALL=(ALL:ALL) ALL
  • Add your user name to this list below root, like this:
# User privilege specification
root    ALL=(ALL:ALL) ALL
YourUsernameHere! ALL=(ALL:ALL) ALL
  • Press Ctrl x on your kewboard to exit, then press y to save, and click Enter to finish.

Verify the permission change

Use su followed by YourUsernameHere! to switch to the new user account:

root@YourComputerName:~# su - YourUsernameHere!
YourUsernameHere!@YourComputerName:~$
  • Use sudo -i to verify that the user account can elevate permissions. At the prompt, enter the new user’s password:
YourUsernameHere!@YourComputerName:~$ sudo -i
[sudo] password for YourUsernameHere!:
root@YourComputerName:~#
  • Use whoami to verify that you are currently the root user:
root@YourComputerName:~# whoami
root

Install Jaseci

  • Install Jaseci dependencies
apt-get install python3.10-dev python3-pip git g++ build-essential pkg-config cmake
  • Install and upgrade pip to the latest
pip install --upgrade pip
  • Install Jaseci
pip install jaseci
  • Test the Jaseci install, to ensure our installation is working run,
jsctl -m

Quick Start

  • Clone this repo into your new project folder!

  • Create this example file: training_data.json in the root of your project, (feel free to change this content to fit your business needs)

{
    "pricing categories": [
        "What are the main pricing options for Amazon EC2 instances?",
        "can I know about ec2 prices?",
        "can i buy ec2 immediately",
        "I would like to know ec2 pricing",
        "is it possible that i can findout the value of ec2 instances",
        "I want to notice methods of ec2 values?",
        "I'd want to study about ec2 pricing",
        "Can i understand the ec2 values?",
        "I need to see the cost of ec2 instance classes",
        "by chance can i take sense the amount of instances?"
    ],
    "ebs storage": [
        "what is EBS storage?",
        "what do you mean by EBS storage?",
        "can you elaborate about ebs",
        "i heard about ebs, what is it?",
        "explain to me regards elastic block-level storage?",
        "i want to know about ebs volume?",
        "tell me, what really ebs is?",
        "main storage to add to ecs instances",
        "how should I know about ebs from you guys",
        "can you describe what ebs is?"
    ]
}
  • If you do change this file make sure to change this content in the dialogue.jac file here:
// change the pricing_state, name, and response
pricing_state = spawn node::dialogue_state(
    name = "ec2_price_method",
    response = "EC2 offers five pricing options: On-Demand, Savings Plans, Reserved Instances, Spot price, and Dedicated Hosts pricing. Other costs include egress data transfers, premium support, and block storage costs."
);
// change the ebs_state, name, and response
ebs_state = spawn node::dialogue_state(
    name = "ebs_storage",
    response = "An Amazon EBS volume is a durable, block-level storage device that you can attach to your instances."
);
  • Create .venv environment,
virtualenv venv
  • Rename venv file to .venv
  • Activate .venv
source .venv/bin/activate
  • Start the Redis server
sudo service redis-server restart
  • Start Jaseci runtime
jsctl -m
  • Create graph
jaseci > graph create -set_active true
  • Compile the project
jaseci > sentinel register main.jac
  • Create the alias list
jaseci > alias list
  • Create the main.dot file
jaseci > graph get -mode dot -o main.dot
  • Exit jsctl
jaseci > exit
  • Create the main.pdf file
dot -Tpdf main.dot -o main.pdf

Training the bi-encoder model with the training data

  • Train the bi-encoder model with the training data using the following commands:

  • Starting the Redis server

sudo service redis-server restart
  • Enter Jaseci runtime in memory:
jsctl -m
actions load module jac_nlp.bi_enc
  • Run the following command,
jac run bi_enc.jac -walk train -ctx '{"train_file": "training_data.json"}'
You should see something like this:
jaseci > jac run bi_enc.jac -walk train -ctx '{"train_file": "training_data.json"}'
2023-05-10 14:05:27,231 - WARNING - rt_warn: bi_enc.jac:bi_enc.jac - line 2, col 4 - rule can_stmt - Attempting auto-load for bi_enc.train
2023-05-10 14:05:27.381448: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-05-10 14:05:27.432466: I tensorflow/tsl/cuda/cudart_stub.cc:28] Could not find cuda drivers on your machine, GPU will not be used.
2023-05-10 14:05:27.432895: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-05-10 14:05:28.223594: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
shared model created
Using device for training ->  cpu
Saving shared model to : modeloutput
shared model created
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00,  6.62it/s]
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00,  7.14it/s]

            Epoch : 1
            loss : 0.14623326311508814
            LR : 0.0002307692307692308
  • The training epoch is set in the bi_enc.jac file in the walker train,
walker train {
    has train_file;
    has num_train_epochs = 100, from_scratch = true, model_name = "";
    root {
        spawn here ++> node::bi_enc;
        take --> node::bi_enc;
    }
    bi_enc: here::train;
}
  • I set this to 100.
Epoch: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 100/100 [01:20<00:00,  1.25batch/s]
{
  "success": true,
  "report": [],
  "final_node": "urn:uuid:faed4ef8-45b8-4fd1-b18c-5f2a467f8b66",
  "yielded": false
}
  • Infer the model
jac run bi_enc.jac -walk infer -ctx '{"labels": ["pricing categories", "ebs storage"]}'
  • Save the model
jac run bi_enc.jac -walk save_model -ctx '{"model_path": "dialogue_intent_model"}'
  • Load the model
jac run bi_enc.jac -walk load_model -ctx '{"model_path": "dialogue_intent_model"}'
  • Running the Chatbot Example
jac run main.jac