Create Modelgarden

Custom ML Model Registry

The Custom ML Model Registry allows you to manage and deploy custom machine learning models, providing a centralized repository for your models. This feature streamlines the process of organizing, versioning, and deploying models.

Steps to Access: 1. Click on "Models." 2. This action will open a page displaying available models. 3. There is an option to search for models on this page. 4. On the right side of this page, there is a button labeled "Create New." 5. Beside the "Create New" button, there is an option called "Playground."

Creating a New Model:

Fine-Tuning:

To create a new fine-tuned model, follow these steps:

Click on "Create New"
This will open a dropdown with two options: "Fine Tune" and "Register Model."
Select "Fine Tune"
This will open the initial form for creating a fine-tune.

Step 1: Create Model

Form Fields
Create Model: This is the title of the form.
Model Name: Input field to enter the name of the model.
- Example: "Enter model name"
Base Model: Dropdown to select a base model.
- Example: "8de59b50-5993-433d-9b30-f51b91da52e2"
Next Button
Click "Next" to proceed to the next step.

Step 2: Configure Fine-Tuning

Form Fields
Create Model: This is the title of the form.
Model Name: Display field showing the entered model name.
- Example: "test"
Model ID: Display field showing the selected base model ID.
- Example: "8de59b50-5993-433d-9b30-f51b91da52e2"
Select the Finetuning: Dropdown to select the finetuning type.
Enter Epoch Count: Input field to enter the epoch count.
- Example: "3"
Learning Rate: Input field to enter the learning rate.
- Example: "0.001"
Batch Size: Input field to enter the batch size.
- Example: "1"
Max Sequence Length: Input field to enter the max sequence length.
- Example: "200"
Warmup Steps: Input field to enter the warmup steps.
- Example: "3"
Gradient Accumulation Steps: Input field to enter the gradient accumulation steps.
- Example: "3"
Next Button
Click "Next" to proceed to the final step.

Step 3: Upload Data

Form Fields
Create Model: This is the title of the form.
Model Name: Display field showing the entered model name.
- Example: "test"
Model ID: Display field showing the selected base model ID.
- Example: "8de59b50-5993-433d-9b30-f51b91da52e2"
Upload File: Drag & Drop files here or click to browse.
Maximum file size is unlimited
Create Button
Click "Create" to finalize the fine-tune.

Registering a New Model:

To register a new model, follow these steps:

Click on "Create New"
This will open a dropdown with two options: "Fine Tune" and "Register Model."
Select "Register Model"
This will open the initial form for registering a model.

Step 1: Register Model

Form Fields
Register Model: This is the title of the form.
Model Name: Input field to enter the name of the model.
- Example: "Enter model name"
Select Container: Dropdown to select the container type.
- Example: "prebuild"
Category: Dropdown to select the category.
- Options: Natural Language Processing, Computer Vision, Tabular
Tag: Input field to enter tags.
Description: Text area to enter the model description.
- Example: "Enter model description"
Next Button
Click "Next" to proceed to the next step.

Step 2: Upload Files

Form Fields
Register Model: This is the title of the form.
Model Name: Display field showing the entered model name.
- Example: "Test"
Select Framework: Dropdown to select the framework.
Upload Models or Supporting Files: Drag & Drop files here or click to browse.
- Maximum file size is unlimited (First file must be a zip file).
Upload Inference Script: Drag & Drop files here or click to browse.
- Maximum file size is unlimited (Second file must be a Python file).
Upload Environment Type Files: Drag & Drop files here or click to browse.
- Maximum file size is unlimited (Third file must be a text file).
Next Button
Click "Next" to proceed to the final step.

Step 3: Compute Requirements

Form Fields
Register Model: This is the title of the form.
Model Name: Display field showing the entered model name.
- Example: "Test"
Compute Required
How Much Memory Do You Need?: Dropdown to select the memory requirement.
How Much CPU Do You Need?: Dropdown to select the CPU requirement.
How Much GPU Do You Need?: Dropdown to select the GPU requirement.
Create Button
Click "Create" to finalize the model registration.

Notes: - Ensure all required fields are completed before proceeding to the next step. - Customizing fine-tuning parameters and model registration allows for more effective model training and deployment based on specific needs.

7. Model Inference

Model Inference allows you to deploy and use your trained models to make predictions and generate outputs based on new data. This feature is essential for applying your models to real-world scenarios and deriving value from them.

Steps to Access: 1. Click on "Model Inference" from the home page. 2. Select the model you wish to use for inference.

Using Model Inference: 1. Input Data: Enter the data you want the model to process. 2. Run Inference: Click "Run" to get predictions or outputs from the model. 3. View Results: The results will be displayed on the screen.

Potential Applications: - Using predictive models to forecast trends and outcomes. - Generating recommendations based on user data. - Automating decision-making processes in business workflows. - Conducting data analysis and deriving insights from large datasets.

Notes: - Ensure the model is properly trained and fine-tuned for accurate predictions. - Monitor the performance of the model and adjust parameters as needed for optimal results.

8. LLM Comparison Tool

The LLM Comparison Tool allows users to compare the performance of different language models by inputting questions and evaluating their responses. This feature helps in selecting the most appropriate model for your needs.

Steps to Access: 1. Click on "Playground." 2. This action will open a page with two options: - Model Compare - Model API Testing

Model Compare: The Model Compare feature allows users to compare different models by asking questions and evaluating the responses.

Interface: - On the left side, there is an option called "Enter text here" where users can input questions. - On the right side, there are various options to select:

Models
Search: Input field to search for models.
Available Models: List of available models to choose from.
- gpt-4o
- Claude-3-chat
- gpt-4-turbo
- gpt-4
- Claude-3
- Gemini
- gpt-3.5-turbo
Compare Controls
Maximum Length: Input field to set the maximum length of the response.
- Example: "200"
Temperature: Input field to set the temperature.
- Example: "1"
Top P: Input field to set the Top P value.
Reset and Apply Buttons:
- Reset: Resets the settings to default.
- Apply: Applies the selected settings.

Steps to Compare Models: 1. Enter Text Here: Input your question in the text area on the left. 2. Select Models and Settings: Choose the models from the list on the right. 3. Apply Settings: Click "Apply" to generate responses from the selected models. 4. View Responses: Compare the responses from the two models. 5. Ask Another Question: Input another question to generate new answers.

Model API Testing: The Model API Testing feature allows users to test models in real-time with specific input data.

Interface: - There are two main sections: - Input Data to Test Realtime - Test Result

Steps to Test Model API: 1. Input Data to Test Realtime: Enter your input data in the provided text area. 2. Select Options: - Category: Select the category of the model. - Model: Select the model to test. - Reset and Apply Buttons: - Reset: Resets the settings to default. - Apply: Applies the selected settings. 3. Run Test: Click "Test" to execute the test with the provided input data. 4. View Test Result: The test result will be displayed in the Test Result section.

Notes: - Ensure that all required fields are completed before running tests or applying settings. - The comparison and testing tools allow for effective evaluation and utilization of different models based on specific needs.