Reimplement EOT weighting

Change red highlighting for debugging

Add MASK slider

Changed to bidirectional

Visualize noising

Update app.py

Fix MASK token noising

Remove unnecessary print statements - Add MASK noising

Changed model from tini_bi_m to tini_bi

Turn autocast back on

Try different monkey-patch

Monkey-patch (temporarily) for LoRA layers

Remove autocast

Change model

New masking implementation

Clamp logits BEFORE softmax

Set to bidirectional for debugging

Set to unidirectional for debugging

Remove single ]

Safe sampling

Deal with float values

Make attention mask float

Update llama_diffusion_model.py

Create safe fallback for models not yet initialized with masking_type

Overhaul code for appropriate masking for full model instead of just attention layers

Fix attention_weights referenced before assigned bug

Implement improved attention masking for bidirectional_masked

Fix noise_start to always start at 1.0

Remove erroneous indent

Fix pause at start

Fix initial noise display

Fix initial noise display

Fix initial noise display

Show initial noise

Fix input_ids instead of current_tokens for first noise iteration

Use updated settings for initial and clustered noise

Add noise start

Add comma

Add pause length

Update app.py

Update requirements.txt

Change LoRA size from 256 to 512, also back to bidirectional_masked

Changed back to bidirectional attention

Try out bidirectional_masked prediction

Show red highlighting before actual noising

Fix

Fix generation

Fix generation

Fix red higlighting

Fix red highlighting