Prompt Sequence & Tokens
Presets:
Vocabulary Size
32 Tokens
Embedding Dim
16-d
Attention Heads
4 Heads
Layers
2 Blocks
Self-Attention Score Heatmap
Head:
Sampling Hyperparameters
Temperature (T)
0.70
Top-K Filter
3
Sampling Strategy
Top-K Weighted
Next-Token Logit & Probability Distribution
Active Last Context Token:
"next"
Predicted Next Token:
"token"
Max Softmax Prob:
0.00%