Coursehub Transformers Lab

Model Hyperparameters

Transformer Model

Temperature 0.70

Nucleus Sampling (Top-P) 0.90

Max Output Tokens 80

Autoregressive Prompt Input

🧠 Attention Matrix Weight Graph Q * Kᵀ / √d_k

TOKEN DECODER STREAM

            UTF-8 Logits
          

System initialized. Awaiting prompt request...

Inference Latency 0.00 ms

Tokens Decoded 0

Generation Rate 0.0 tps

Transformer Attention Layer Trace

[00:00:00] Console idle. Ready for autoregressive mapping.

Network Architecture

Neural Network Type

Training Hyperparameters

Learning Rate (η) 0.100

Batch Size 32

Steps Per Click 50

Newton's Laws Corpus Training Console

Dataset: Newton's Laws of Motion (~1,000 characters). The model is training to predict the next character based on a context window (1 char for Bigram, 3 chars for MLP).

Epoch Step

0

🧠 Active Synapse Weight Map Loss: --

Loss Convergence History

Next Character Probabilities

Context Input Seed

Type in the seed to show predictions.

Secure Access Required