Fix the synthetic data generation pipeline

Use FUS logs (not uploaded to repo) to compare length difference and edit distance distributions in FUS and in our dataset (resulting charts are not included).

Added some description to the README.md

Latest version of the code; config updated to JetBrains-Research

Generate charts for the presentation & diploma;some refactoring; add (commented) Student's t-test

Remove the histograms

Add distribution charts; add more detailed statistics; compute multi-reference TER as mean of TERs for each reference to improve the performance

Add the aggregated correlations table

Visualizer bugs fixed; added normalized editdist

Fix the bug

Fix the visualization

Update requirements.txt

Update requirements.txt

Revert "Update requirements.txt"

Update requirements.txt

Pretty-print all the correlations in the visualization app

Add noref gpt-eval to the pipeline

Add edit distance and edit time metrics; add GPT-based metric

Generate a dataset for the labeling app

Update the parameters

Add checkpoints

Compute & compare metrics

Prompts & params adjusted

Default value for "start_to_end"

Start-to-end generation

Fix the bug

Default value for "end_to_start" for manual df

Display the end-to-start field

Keep the session column

Add returns

Fix the dfs in visualization

Fix the statistics in visualization

Fix the visualization

- New version of the end->start synthetics samples generation

Fix

# of deletions rel to initial message length

Add datasets comparison

Force the light theme

Full dataset generation

Add some examples to the synthetic ds generation prompt

Synthetic dataset generation for the first 5 samples; visualization fixed

Synthetic dataset visualization

Synthetic dataset generation

Spaces config changed

Spaces config changed

Add requirements.txt

Add HF spaces config

Create visualize app

WIP on annotated diffs generation

WIP on annotated diffs generation