From 41a2cd34d7ed35399e1bc49ae2c4e53a4a65a0b0 Mon Sep 17 00:00:00 2001 From: 0oshowero0 Date: Fri, 11 Sep 2026 22:03:16 +0800 Subject: [PATCH 1/2] update Signed-off-by: 0oshowero0 --- README.md | 1 + 1 file changed, 1 insertion(+) diff --git a/README.md b/README.md index 0a04641e..cd0df328 100644 --- a/README.md +++ b/README.md @@ -32,6 +32,7 @@ TransferQueue offers **fine-grained, sub-sample-level** data management and **lo

🔄 Updates

- **September 8, 2026**: 🔥 TransferQueue has been adopted in [Meshy](https://github.com/OpenBMB/Meshy). Meshy models each role as a dedicated service, and TransferQueue serves as the data flow manager that connects each role. + - **August 14, 2026**: 🔥 TransferQueue has been adopted in [Dressage]https://github.com/Accio-Lab/Dressage. By integrating TQ, peak memory usage of the data plane on the master node is reduced by 91% in a 32-node GLM-5.2 setup. Refer to the [report](https://github.com/Accio-Lab/Dressage/blob/main/docs/trajectory-storage-refactor-en.md) for details. - **June 18, 2026**: 🔥 TransferQueue has been adopted in [ROLL](https://github.com/alibaba/ROLL/pull/463). This integration introduces a [`RemoteBatch`](https://github.com/alibaba/ROLL/blob/main/docs_roll/docs/User%20Guides/Advanced%20Features/remote_batch_transfer.md) abstraction, enabling seamless compatibility with existing `DataProto` design. - **June 9, 2026**: 🔥 TransferQueue has been adopted in [UniRL](https://github.com/Tencent-Hunyuan/UniRL), a unified RL framework for multimodal models developed by Tencent Hunyuan. - **April 15, 2026**: 🔥 TransferQueue has been adopted in [Relax](https://github.com/redai-infra/Relax)! By leveraging the `StreamingDataLoader` abstraction, it schedules training data across the cluster at micro-batch granularity, reducing synchronization barriers in a single-controller setup. From b6dc35322d36b44ce5d4dd659abd1bf1d6e18ace Mon Sep 17 00:00:00 2001 From: 0oshowero0 Date: Fri, 11 Sep 2026 22:04:32 +0800 Subject: [PATCH 2/2] update Signed-off-by: 0oshowero0 --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index cd0df328..3c08a6ad 100644 --- a/README.md +++ b/README.md @@ -32,7 +32,7 @@ TransferQueue offers **fine-grained, sub-sample-level** data management and **lo

🔄 Updates

- **September 8, 2026**: 🔥 TransferQueue has been adopted in [Meshy](https://github.com/OpenBMB/Meshy). Meshy models each role as a dedicated service, and TransferQueue serves as the data flow manager that connects each role. - - **August 14, 2026**: 🔥 TransferQueue has been adopted in [Dressage]https://github.com/Accio-Lab/Dressage. By integrating TQ, peak memory usage of the data plane on the master node is reduced by 91% in a 32-node GLM-5.2 setup. Refer to the [report](https://github.com/Accio-Lab/Dressage/blob/main/docs/trajectory-storage-refactor-en.md) for details. + - **August 14, 2026**: 🔥 TransferQueue has been adopted in [Dressage](https://github.com/Accio-Lab/Dressage). By integrating TQ, peak memory usage of the data plane on the master node is reduced by 91% in a 32-node GLM-5.2 setup. Refer to the [report](https://github.com/Accio-Lab/Dressage/blob/main/docs/trajectory-storage-refactor-en.md) for details. - **June 18, 2026**: 🔥 TransferQueue has been adopted in [ROLL](https://github.com/alibaba/ROLL/pull/463). This integration introduces a [`RemoteBatch`](https://github.com/alibaba/ROLL/blob/main/docs_roll/docs/User%20Guides/Advanced%20Features/remote_batch_transfer.md) abstraction, enabling seamless compatibility with existing `DataProto` design. - **June 9, 2026**: 🔥 TransferQueue has been adopted in [UniRL](https://github.com/Tencent-Hunyuan/UniRL), a unified RL framework for multimodal models developed by Tencent Hunyuan. - **April 15, 2026**: 🔥 TransferQueue has been adopted in [Relax](https://github.com/redai-infra/Relax)! By leveraging the `StreamingDataLoader` abstraction, it schedules training data across the cluster at micro-batch granularity, reducing synchronization barriers in a single-controller setup.