Feature: increase context length

Refactor diffusion model functions and update Gradio interface for improved user interaction

Update app.py

Update requirements.txt

Update requirements.txt

Update requirements.txt

Update README.md

Add semi-autoregressive generation

add eos_bias

Revert back to 8B

Changed to 3B

Only noise until max_it-1

Change standard settings and revert to 8B

Change to 3B

Noisify without remasking

Improve white space display

Use autocast to float16

Fix loading of model and tokenizer

Add model config files

Improve code structure

Correct name of model

Change prompt to Llama format

Change model to 8B

Revert back to large LoRA layers

Revert to old model temporarily

Update LoRa layer size

Update model to Llama 3.2 3B-Instruct based, changed prompt format

Fix interface bug

Add top_p and top_k sliders

Fix time_elapsed issue

Change highlighting

Fix top-k/top-p

Fix clamping and introduce top-k and top-p filtering

Fix confidence guided noising call

Fix confidence guided noising

Improve confidence guided noising and show number of tokens generated

Change original user question cosmetic

Change dummy-input question

Changed initialization to MASK tokens instead of EOS tokens

Fix model repo name

Remove random noising

Change back to User-Assistant conversation

Change back to normal model

Change to mask model

Remove placeholder

Enable direct prompting

Remove answer_start check

Set answer_start to 0

Remove conditional generation

Remove BidirectionalLlamaAttentionLayer input