Add eval_loss and perplexity params to notify_aggregator signature

Merge pull request #1 from madch3m/madeleine/metrics

Add eval loss and perplexity metrics with publication quality plots

Refresh contributors

Update gradio frontend to allow interactive function in the UI

Updated the fed average implementation to be layer aware

Maintain history and information from gradients across sessions for each round

removing flash attention

fix the flash attention hanging

updating the meomry allocation and config for flash attention to work with a100

Updated and increase speed of hf downloads for model

added constants to notebook so it can easily be started

Updated the space so that it can maintian a consistent data between updates and they are persisted

Updating to adress the bottleneck in the poll for next roun

Fix aggregator client url

normalizing the url so that it does not get mismatched

fixed aggregator and submission issue

update the configuration so we can see dashboard visualization

fix(deps): pin Starlette <1 for Gradio 4.44 dashboard

chore(requirements): add upper bounds for runtime stability

feat(aggregator): optional ADMIN_SECRET and STATUS_READ_SECRET

fix(notebook): catch AggregatorMergeFailed in round_end_sync

test: guard rate-bucket reset; chore: add .dockerignore

ci: Docker build and pytest on push and PR

docs(README): operator runbook and hardening endpoints

feat(client): raise on merge_failed and add health_aggregator()

feat(aggregator): structured logging for security and merge events

feat(aggregator): validate /submit payload and round_num

feat(aggregator): rate limit POST /submit per client

feat(aggregator): add GET /health liveness probe

Update the application code as there was and issue with the space

fixed basemodel and pin

Merge hf/main into main (keep local notebook; HF had duplicate tail)

updated repo condition for first node

Updated the requirements to reduce size of gradio space

Missed the repository import

Fix import for repo

Fix import for repo

Added entry point for the hugging space to start the application

down graded huggingface version

Switch Space SDK from gradio to docker

Add Dockerfile to pin Python 3.10 for Gradio 4.44.1 compatibility

Remove gradio from requirements to fix HF Space build conflict

Fix Gradio version for Python 3.13 compatibility

Fix python version

Fix Python compatibility issue

Updated the dashboard to show progress in the space

updated readme to have better instructions for the aggregator

Initial repo changes for scaffolding