Issues
is:issue state:open
is:issue state:open
Search results
MTP weights can not export from DCP to HF safetensors with
megatron exportandswift exportbugSomething isn't workingSomething isn't workingStatus: Open.#9504 In modelscope/ms-swift;loss = 0 when sft qwen3.5-9B
bugSomething isn't workingSomething isn't workingStatus: Open.#9499 In modelscope/ms-swift;SFT 训练过程中 GPU 显存波动异常
bugSomething isn't workingSomething isn't workingStatus: Open.#9495 In modelscope/ms-swift;ms-swift 4.2.3镜像中缺少tilelang安装包,手动安装出现libcudart_stub.so: undefined symbol: cudaDeviceReset
bugSomething isn't workingSomething isn't workingStatus: Open.#9494 In modelscope/ms-swift;DeepSeek v4系列的cp和tp什么时候可以支持
enhancementNew feature or requestNew feature or requestStatus: Open.#9490 In modelscope/ms-swift;qwen3.5 0.8B,如何关闭输出<think>\n\n</think>\n\n
questionFurther information is requestedFurther information is requestedStatus: Open.#9485 In modelscope/ms-swift;Atlas A3设备网络调度,不支持背靠背直连hccs api接口,在走灵衢的双机正常拉起,背靠背npu直连的双机,同样的脚本无法拉起
bugSomething isn't workingSomething isn't workingStatus: Open.#9481 In modelscope/ms-swift;Kimi-K2.6训练的多模态(image)支持
enhancementNew feature or requestNew feature or requestStatus: Open.#9469 In modelscope/ms-swift;有无支持训练Qwen-VLA的计划?
questionFurther information is requestedFurther information is requestedStatus: Open.#9467 In modelscope/ms-swift;OPD无法用于Qwen3.5 MoE/Qwen3.6 MoE模型
bugSomething isn't workingSomething isn't workingStatus: Open.#9466 In modelscope/ms-swift;grpo多轮训练中,没有设置dynamic_sample,但是报错Padding free mode is not supported for dynamic sample
bugSomething isn't workingSomething isn't workingStatus: Open.#9454 In modelscope/ms-swift;GLM5.1 MoE + PP 训练卡在 Train 0/100:batch_p2p_comm=True 但实际触发 unbatched P2P send/recv lazy NCCL communicator init
bugSomething isn't workingSomething isn't workingStatus: Open.#9451 In modelscope/ms-swift;