v1.1.5

What's Changed

• fix not reading gpu in docker resource exporter by @swan-alexchen in #232
• add hugging face download cache directory mount if the host have it by @swan-alexchen in #233
• Fix/add taskuuid in requesting cp deploy image by @swan-alexchen in #234
• Expose metrics endpoints by @swan-alexchen in #235

Full Changelog: v1.1.4...v1.1.5

v1.1.4

This release (v1.1.4) introduces multiple improvements to deployment workflows, Kubernetes integration, and GPU resource handling. These updates make the system more robust, reduce deployment issues, and improve performance under various environments.

---

🚀 What's New & Fixed

• [Infra] Updated utils Docker image account to fix image download issue (#229)
• [Engine] Use new UBI task public key in sending tasks (#226)
• [Deploy] Fix issues on lagrange/cp/deploy job deployment API (#228)
• [Resource Exporter] Prevent adding --gpu flag if no GPU is available on the host to fix issue on CPU-only UBI workers (#222)
• [K8s] Added support for passing Kubernetes command and args to SSH pod (#227)
• [K8s] Increased shared memory (shm) size in Kubernetes mounts to fix vLLM parallelism errors (#223)

---

🛠 How to Upgrade

✅ Replace the computing-provider binary with the v1.1.4.

📌 Full Changelog: v1.1.3...v1.1.4

v1.1.3

This release(v1.1.3) is a recommended upgrade for all users. It introduces an important new feature that allows better control over CP machine resource utilization. By configuring a utilization threshold, the system can now intelligently decide when to stop assigning tasks based on GPU load, ensuring more stable performance under high demand.

🚀 What's New & Fixed

• [Config] Added GpuUtilizationRejectThreshold option

  You can now set a threshold in config.toml to control when the system should stop assigning new tasks to a GPU:

  GpuUtilizationRejectThreshold = 1  # When the GPU utilization reaches this value, no further tasks will be performed. For example, 0.5 means 50% utilization, while 1.0 means the GPU is fully utilized.

• Add: control over CP machine resource utilization (#220)

🔧 How to Upgrade

✅ Replace the computing-provider binary with the v1.1.3.

📌 Full Changelog: v1.1.2...v1.1.3

Mainnet-resource-multinode

This release(v1.1.2) focuses on improving stability and functionality, particularly for multi-node FCP deployments in Kubernetes environments. It resolves critical issues related to task creation failures and incorrect status updates while introducing a new /zk_task API. For FCP running a multi-node Kubenates cluster, upgrading is strongly recommended to ensure optimal performance and reliability.

🚀 What's New & Fixed

• Fixed: Issue where FCP failed to create a sampling task in Kubernetes with multi-node environments (#216)
• Fixed: Issue where rejected task statuses were not correctly updated (#217)
• Added: New /zk_task API for ECP running on Kubernetes (#218)
• Updated: Version bumped to v1.1.2 (#219)

🔧 How to Upgrade

✅ Replace the computing-provider binary with the v1.1.2.

📌 Full Changelog: v1.1.1...v1.1.2

Mainnet-resource-fix

This is a MANDATORY Upgrade for all Computing Providers (CPs). All CPs must complete this mandatory upgrade by 2:00 AM (UTC+0) on February 12, 2025. Failure to upgrade before this deadline will result in the inability to receive UBI tasks, directly affecting your UBI earnings.

We strongly urge all CPs to proceed with the upgrade as soon as possible.

Upgrade Details

1. Upgrade computing provider to v1.1.1

2. Upgrade resource-exporter to v12.0.0

   • Added collection of product_uuid and gpu_uuid information.
   • Introduced a resource information verification mechanism to prevent CPs from tampering with real resource data(#208)

3. FCP and ECP Enhancements

   • Support for deploying applications via deployment file string encoding format (#208)

4. Task Deployment Flexibility

   • ECP and FCP now support deploying a single task across multiple different types of GPUs(#208)

5. Configuration File Update

   • Update [HUB].OrchestratorPK to:
     0xd875bD44158208fD0FDD46729Aab6709f62C7821 (#209)

6. New Task Status Display

   • Tasks with the Rejected status will now be visible in the computing-provider list(#207)
   • Fix ubi task status(timeout) (#213)

How to Upgrade:

Please refer to the upgrade steps provided here.

Ensure your systems are updated promptly to avoid any disruptions in your UBI task flow and earnings.

Mainnet-resource

This is a MANDATORY Upgrade for all Computing Providers (CPs). All CPs must complete this mandatory upgrade by 2:00 AM (UTC+0) on February 12, 2025. Failure to upgrade before this deadline will result in the inability to receive UBI tasks, directly affecting your UBI earnings.

We strongly urge all CPs to proceed with the upgrade as soon as possible.

Upgrade Details

1. Upgrade computing provider to v1.1.0

2. Upgrade resource-exporter to v12.0.0

   • Added collection of product_uuid and gpu_uuid information.
   • Introduced a resource information verification mechanism to prevent CPs from tampering with real resource data(#208)

3. FCP and ECP Enhancements

   • Support for deploying applications via deployment file string encoding format (#208)

4. Task Deployment Flexibility

   • ECP and FCP now support deploying a single task across multiple different types of GPUs(#208)

5. Configuration File Update

   • Update [HUB].OrchestratorPK to:
     0xd875bD44158208fD0FDD46729Aab6709f62C7821 (#209)

6. New Task Status Display

   • Tasks with the Rejected status will now be visible in the computing-provider list(#207)

How to Upgrade:

Please refer to the upgrade steps provided here.

Ensure your systems are updated promptly to avoid any disruptions in your UBI task flow and earnings.

Mainnet-tasks

This is the optional update for the Swan Mainnet. This version includes task-related optimization updates, which makes it easier to view the status of all tasks and accurately calculate the success rate. The log of the application running is retained to review the status of the application

We highly encourage all users to upgrade to v1.0.2 for

Improvement & bugfix

• Fix the status of mining task in the task list (#201)
• Show the rejected tasks in the task list (#201)
• Retain the container log in the $CP_REPO/logs, and cp can control the cleanup period by configuring it at $CP_REPO/config.toml: (#201)

[API]
ClearLogDuration = 24  # Delete logs at intervals after the job is finished, the unit is hours

• Update resource-exporter to v11.3.2 How to Upgrade (#201)

Full Changelog: v1.0.1...v1.0.2

Mainnet-contractOwner

We are excited to release the first update following the TGE! This version introduces support for contract addresses as CP's owner, improving flexibility and compatibility. It also addresses key community feedback, enhancing UBI task success rates and optimizing high-end GPU utilization for better performance.

We highly encourage all users to upgrade to v1.0.1 for the latest features, improvements, and optimized performance.

Improvement && bugfix

• Support contract address as CP's owner (#195)
• Fix the timeout issue when receiving UBI tasks(#194)
• Sharding rental for high-end GPU (such as H100) that support NVlink technology(#194)
• Fix the nil error when generating the price configuration(#191)
• Upgrade resource-exporter version to v11.3.1How to Upgrade (#197)

Full Changelog: v1.0.0...v1.0.1

Mainnet-TGE

We are excited to announce the release of TGE Version 1.0.0, a critical update that introduces new features and improvements to enhance the security and functionality of the platform.

Upgrade Deadline

To maintain compatibility and ensure continued access to all functionalities, all CPs must upgrade to Version 1.0.0 by the specified deadline:
December 16, 2024, at 10:00 AM EST

Note: Failure to upgrade will result in restricted access to the platform and potential compliance issues with the updated protocol.

---

Key Updates

• Slash Mechanism Activation
  The slash mechanism is now enabled, providing a robust framework for penalizing misbehaviour and ensuring the integrity of the ecosystem. The details can be found here

• Mandatory Full Staking
  Full staking is required; CPs must fully deposit $SWAN to qualify for the UBI reward. The details can be found here

---

Improvements

• Improve the GPU allocates strategy by specifying indices(#181)

• Allow FCP to withdraw the collaterals from the Escrow balance.

  computing-provider collateral withdraw-request --fcp --owner <YOUR_OWNER_ADDRESS>  <AMOUNT>
  computing-provider collateral withdraw-confirm --fcp --owner <YOUR_OWNER_ADDRESS>  <AMOUNT>
  

• ECP can deploy inference service from the docker images and provide the logs; details are here (#181)

• Add ubi-0 cmd to view and withdraw ubi-0 collaterals after TGE(#181)

  • Check the collateral details of SWANU:

  computing-provider ubi-0 info 
  

  • Withdraw the collateral for the SWANU

  # For FCP
  computing-provider ubi-0 withdraw --fcp --owner <YOUR_OWNER_ADDRESS>  <AMOUNT>
  
  # For ECP
  computing-provider ubi-0 withdraw --ecp --owner <YOUR_OWNER_ADDRESS>  <AMOUNT>
  

• Update the orchestratorPK(#185)

  • CPs must update it in the config.toml

  [HUB]
  OrchestratorPk = "0xE2F887D4Ed3E4E5179C2f30c6Fdb2344cCbB21b6" 
  

• update contract address for TGE (#181, #184)

Upgrade Documentation

For a detailed guide on how to upgrade, please refer to the upgrade documentation: here

Swan Mainnet Network Information and Related Smart Contract Addresses

As part of this update, all relevant smart contract addresses across the network have been modified.

Item	Details
Currency Symbol	ETH
Chain ID	254
RPC URL	https://mainnet-rpc.swanchain.org
Swan Token Symbol	SWAN
SwanETH Bridge	https://bridge.swanchain.io
Block Explorer URL	https://mainnet-explorer.swanchain.io
Swan Provider Dashboard	https://provider.swanchan.io
SWAN Token Contract address	0xBb4eC1b56cB624863298740Fd264ef2f910d5564
Collateral Contract for FCP	0x1BE0F0C000ECDd6b8544b20694A3d118625cE005
CP Account Register Contract	0xbef1ec33e5Ac3a491fFeE1600e0b00eac97cb138
ZK Task Register Contract	0x4Cd9b2450c3e412a1F0F0aE78C79271337241E31
ECP Collateral Contract	0x557f306f917009cf83c32b8b32a79202e79948e5
Orchestrator Public Key[Updated]	0xE2F887D4Ed3E4E5179C2f30c6Fdb2344cCbB21b6
ZK engine Publicc Key	0xB5aeb540B4895cd024c1625E146684940A849ED9
Sequencer Contract	0xE0a88F1fA56D8a8D73E9bb9Fc03506439cb386Bd
Gas to Sequencer	Dynamic Pricing Strategy
Sequencer Blob Period	24 Hours

We appreciate your continued support and partnership in building a secure and efficient ecosystem. If you have any questions or encounter issues during the upgrade process, please contact us promptly.

Full Changelog: v0.7.1...v1.0.0

Mainnet-ECP-mining

This is the second version for FCP and ECP to join the Swan Mainnet Computing Provider UBI-0. This update includes many new mechanisms and practical features.

Improvements & Bug Fixes

• update taskTypes(add mining and inference, update fil-c2) of CP (#164)
• add the task management for the ECP's mining service, check it by computing-provider task list --type ecp (#165)
• fix update contract address in the ubi task list (#167)
• add sequencer's Dynamic Pricing Strategy (#168)
• add pricing="true" in the config.toml for cp to set the resource price(#167)
• fix the wrong label of cpu type in the cluster of FCP(#167)

Swan Mainnet Network Information

Item	Details
Currency Symbol	ETH
Chain ID	254
RPC URL	https://mainnet-rpc.swanchain.org/
Swan computing Units Symbol	SWANU
SwanETH Bridge	https://bridge.swanchain.io
Block Explorer URL	https://mainnet-explorer.swanchain.io
Swan Provider Dashboard	https://orchestrator.swanchain.io/provider-status
SWANU contract address	0x39cBBeaF88a91404618d45a16e0977Adab4d1Af1
Collateral Contract for FCP	0x643A820060c78D9122081aafF0C039b070937558
CP Account Register Contract	0xbef1ec33e5Ac3a491fFeE1600e0b00eac97cb138
ZK Task Register Contract	0x4Cd9b2450c3e412a1F0F0aE78C79271337241E31
ECP Collateral Contract	0x7A099B665e760BCCF3Aa0cBbb66C8da47F13C03d
ZK engine Publick Key	0xB5aeb540B4895cd024c1625E146684940A849ED9
Sequencer Contract	0xE0a88F1fA56D8a8D73E9bb9Fc03506439cb386Bd
Gas to Sequencer	Dynamic Pricing Strategy
Sequencer Blob Period	24 Hours

Full Changelog: v0.7.0...v0.7.1