TTNN Operation Calculator

A modern web application for testing and visualizing ttnn operations in real-time.

Features

• 🎯 150+ Operations: Support for Pointwise Unary, Binary, and Ternary operations
• 🔧 Flexible Inputs: Choose between tensor or scalar inputs with custom values
• 📊 Data Type Support: bfloat16, float32, int32, uint32, and more
• 📈 Result Visualization: View shape, dtype, and output values
• 🎨 Modern UI: Beautiful, responsive interface with real-time feedback
• 🔐 Secure Access: Password-protected with SHA-256 hashing
• ⚡ Optional Parameters: Support for 8 operations with configurable parameters
• 🔄 PyTorch Comparison: Real-time comparison with PyTorch results
• 🌐 Cloudflare Ready: Works seamlessly with Cloudflare Tunnel

Installation

1. Navigate to the project directory:

cd /home/aswin/ttnn-web-calculator

2. Activate the tt-metal Python environment:

cd /home/aswin/tt-metal
source python_env/bin/activate
cd /home/aswin/ttnn-web-calculator

3. Install Flask (if not already installed):

pip install Flask==3.0.0

Security

The application is password-protected for secure access:

Password: mcw.pass

• Uses SHA-256 password hashing
• Session-based authentication
• All routes protected (including API endpoints)
• Password is NOT stored in plain text

See SECURITY.md for detailed security information.

Usage

1. Start the web server:

python app.py

2. Open your browser and navigate to:

http://localhost:5000

3. Select an operation from the dropdown menu
4. Configure your inputs:
   • Choose between Tensor or Scalar
   • Set the value
   • Select the data type
5. Click "Calculate Result" to execute the operation
6. View the results:
   • TTNN operation result value
   • PyTorch equivalent result (for comparison)
   • Match indicator showing if results agree
   • Sample values from the output tensor

Supported Operations

Pointwise Unary (Single Input)

• Trigonometric: sin, cos, tan, asin, acos, atan, sinh, cosh, tanh
• Exponential: exp, exp2, expm1, log, log2, log10, log1p
• Activation: relu, sigmoid, gelu, silu, swish, tanh, softplus
• Other: abs, sqrt, square, rsqrt, reciprocal, neg, sign, floor, ceil, round

Pointwise Binary (Two Inputs)

• Arithmetic: add, subtract, multiply, div, pow, remainder
• Comparison: gt, lt, ge, le, eq, ne, maximum, minimum
• Logical: logical_and, logical_or, logical_xor
• Bitwise: bitwise_and, bitwise_or, bitwise_xor

Pointwise Ternary (Three Inputs)

• addcdiv, addcmul, mac, where, lerp

Architecture

• Backend: Flask with ttnn integration
• Frontend: HTML/CSS/JavaScript with modern responsive design
• Device Management: Automatic initialization and cleanup of ttnn device

Example

To test ttnn.add:

1. Select "ttnn.add" from the dropdown
2. Set Input 1: Type=Tensor, Value=5.0, dtype=bfloat16
3. Set Input 2: Type=Tensor, Value=3.0, dtype=bfloat16
4. Click "Calculate Result"
5. View the result: All values should be 8.0

Notes

• All tensor operations use a default shape of (2, 2, 32, 32) for compatibility
• Tensors are automatically moved to the device with TILE_LAYOUT
• The device is initialized on first use and persists between requests
• Results show the first 10 elements as samples