TaoFold v1.3.3

• Fix mass-setting exploit in miner returned proteins.
• New validators should weight copy on startup if they have no weights. This gives them a warm start

TaoFold v1.3.2

This includes:

1. Upgrading the bittensor version of Folding to 8.5.1 to support CR3

Taofold V1.3.1

Released a small hotfix for an exploit

TaoFold v1.3.0

We provide a major upgrade to the folding repo that is a first stepping stone towards a Global Job Pool. The major features include:

1. an rqlite distributed database (https://rqlite.io) for holding the job requests created by the validator
2. We use a centralized API from the new macrocosmos repo called Golgi-apparatus (https://github.com/macrocosm-os/golgi-apparatus) which is a FastAPI-based service for managing and interacting with job creation and deletion processes, backed by RQLite for distributed storage.
3. The API uses Epistula (https://epistula.sybil.com) for secure communications that are signed by the hotkey credentials of the validator/miner

TaoFold v1.2.3

Increase PingSynapse timeout because new Bittensor dendrite seems to be slower, and this was causing Timeout issues for all validators.

TaoFold v1.2.2

Increase Bittensor version... again

TaoFold v1.2.1

• Increase Bittensor version to 8.4.1
• Include better logging format from @loayei

TaoFold v1.2.0

This release includes:

1. Integration of new infra that allows for organic queries from an external endpoint. This is the first macrocosmos release to utilize atom, our SDK.
2. Utilizes poetry for installations
3. Uses Bittensor 8.3.1

TaoFold v1.1.1

Features:

• Remove async nature of wandb logging because this seems to have unintended issues with wandb getting stuck
• Add tenacity retry to try and set weights 3 times before moving on.

Bug Fixes:

• Move the initial sync of the validator to the outer loop to avoid deleting previously saved weights
• Remove the sync inside of the base validator to avoid having miner weights go to 0 on each restart.

Extras

• Increase block time for weight setting from 100 to 250 to help smooth out miner weights

TaoFold v1.1.0

Includes:

1. Async weight setting
2. Async job creation
3. Async job updating

This is to help address the issues of validator synchronous bottlenecks