Search results

[Bug] Megatron MoE LoRA with merge_lora=False fails the weight sync on Qwen3.6-35B MoE

Status: Open.

#1852 In NovaSky-AI/SkyRL;

· casper-hansen opened on Jun 30, 2026

SkyRL Q3 2026 Roadmap

Status: Open.

#1846 In NovaSky-AI/SkyRL;

· erictang000 opened on Jun 29, 2026

[algorithm] Add adaptive entropy methods (ADAPO, REPO-R, MAI-Thinking)

Status: Open.

#1845 In NovaSky-AI/SkyRL;

· erictang000 opened on Jun 29, 2026

[RFC] Stateful dataloader and custom samplers for SFTTrainer

Status: Open.

#1839 In NovaSky-AI/SkyRL;

· SumanthRH opened on Jun 26, 2026

Add delete checkpoint endpoint to prevent Tinker checkpoint disk exhaustion

good first issueGood for newcomersGood for newcomers

Status: Open.

#1834 In NovaSky-AI/SkyRL;

· j316chuck opened on Jun 26, 2026

[tinker] cascading distributed barrier timeouts with multi-tenant LoRA

Status: Open.

#1827 In NovaSky-AI/SkyRL;

· nikhilbarhate99 opened on Jun 23, 2026

Fix reduction for entropy and other metrics in policy workers

Status: Open.

#1822 In NovaSky-AI/SkyRL;

· SumanthRH opened on Jun 22, 2026

adv_estimator (function param) vs advantage_estimator (config field) naming mismatch causes confusing startup crash

Status: Open.

#1792 In NovaSky-AI/SkyRL;

· KTanmay1 opened on Jun 16, 2026

[LoRA] Investigate LoRA training inference mismatch for Qwen3.6 MoE and GLM-4.7-Flash

Status: Open.

#1776 In NovaSky-AI/SkyRL;

· CharlieFRuan opened on Jun 10, 2026

[fsdp] monitor FSDP training stability regression after #1679

Status: Open.

#1767 In NovaSky-AI/SkyRL;

· erictang000 opened on Jun 9, 2026

Example for training on SWE (agentic software-engineering) tasks?

Status: Open.

#1757 In NovaSky-AI/SkyRL;

· dipta007 opened on Jun 7, 2026

[feature] Add ECHO patch to SkyRL

Status: Open.

#1715 In NovaSky-AI/SkyRL;

· erictang000 opened on May 28, 2026