| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
[Feature] add downcast to bf16 for fsdp2 and remove --bf16 Co-authored-by: 林明哲<linmingzhe3@huawei.com> # message auto-generated for no-merge-commit merge: !1914 merge fix1208 into master [Feature] add downcast to bf16 for fsdp2 and remove --bf16 Created-by: LinMingZhe Commit-by: 林明哲 Merged-by: ascend-robot Description: ## Motivation - Remove the redundant --bf16 flag to cut memory usage and boost performance; - Introduce weight-loading downcast to bf16, keeping computational precision identical to the previous --bf16 behavior. ## Modification - wan2.2 configs - vace configs - modify dcp patch and import to pretrain_sora.py - add --downcast-to-bf16 to args and introduce it in fsdp2.md ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1914 | 6 个月前 | |
| 10 个月前 | ||
[Bugfix]fix fpdt nonblocking data dependency bug Co-authored-by: Bian Zheng<bianzheng8@huawei.com> # message auto-generated for no-merge-commit merge: !1951 merge fix_fpdt_nonblocking into master [Bugfix]fix fpdt nonblocking data dependency bug Created-by: rewindz Commit-by: Bian Zheng Merged-by: ascend-robot Description: ## Motivation Fix the non-blocking data dependency bug in the FPDT (Fully Pipelined Distributed Transformer, aka ulysses offload) mechanism. In the current implementation, data transfer between CPU and NPU uses non-blocking mode, but data dependencies are not properly handled, which may lead to data inconsistency or calculation errors. ## Modification 1. Optimize data transfer mode : Change data transfer between CPU and NPU from non_blocking=False to non_blocking=True to improve data transfer efficiency. 2. Add stream synchronization mechanism : - When loading data to NPU, use offload_stream for asynchronous loading and wait for completion in compute_stream - Add compute_stream.synchronize() and general_offload_stream.synchronize() to ensure data transfer is complete before computation - Add torch_npu.npu.synchronize() before backpropagation to ensure all operations are complete 3. Refactor data loading logic : Encapsulate data loading operations in a stream context manager to clarify the asynchronous nature of data transfer. ## Self-test (Optional) The correctness of the modification has been verified through the following tests: - Functional test: Ensure the FPDT mechanism works normally in non-blocking mode - Performance test: Verify the speedup of step_time with 10~15% (baseline: blocking FPDT) - Consistency test: Ensure the calculation results are consistent with those before modification - Precision alignment: Precision has been aligned with PR #1886 ## BC-breaking (Optional) No compatibility issues. The modification only involves internal implementation details and does not affect external interfaces. The required cann/torch_npu version remains consistent with before modification. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1951 | 6 个月前 | |
【Bugfix】Correct the spelling errors in the code Co-authored-by: lu-jinfu1999<lujinfu1@h-partners.com> # message auto-generated for no-merge-commit merge: !1670 merge master into master 【Bugfix】Correct the spelling errors in the code Created-by: lu-jinfu1999 Commit-by: lu-jinfu1999 Merged-by: ascend-robot Description: ## Motivation Correct the spelling errors in the code. ## Modification Correct spelling errors in the document. Correct spelling errors in the code. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1670 | 7 个月前 | |
[bugfix]add ulysses-offload specific split function Co-authored-by: mazhuang<mazhuang21@huawei.com> # message auto-generated for no-merge-commit merge: !1886 merge fpdt-split-bugfix into master [bugfix]add ulysses-offload specific split function Created-by: mazhuang1234 Commit-by: mazhuang Merged-by: ascend-robot Description: ## Motivation [bugfix]add ulysses-offload specific split function ## Modification Please briefly describe what modification is made in this PR. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1886 | 6 个月前 | |
[mindspore] support qwen3vl finetune for megatron backend Co-authored-by: ffmh<fengminghao2@huawei.com> # message auto-generated for no-merge-commit merge: !1785 merge qwen3vl_megatron into master [mindspore] support qwen3vl finetune for megatron backend Created-by: ffmh Commit-by: ffmh Merged-by: ascend-robot Description: ## Motivation Please describe the motivation of this PR and the goal you want to achieve through this PR. 适配 qwen3vl 8B/30B megatron后端 ## Modification Please briefly describe what modification is made in this PR. 添加 layer_spec修改、text模块transformer block/layer添加deepstack相关适配 ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1785 | 7 个月前 | |
[Feature] Support Qwen3Omni Co-authored-by: x00824404<xuyaoyao824404@huawei.com> # message auto-generated for no-merge-commit merge: !1747 merge master into master [Feature] Support Qwen3Omni Created-by: x00824404 Commit-by: x00824404 Merged-by: ascend-robot Description: ## Motivation [Feature] Support Qwen3Omni ## Modification [Feature] Support Qwen3Omni Fix Qwen2.5Omni <e/bos> token ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1747 | 7 个月前 | |
[Feature]:The qwen2MoT of Bagel Model Co-authored-by: ningmengliu<liuhao438@huawei.com> # message auto-generated for no-merge-commit merge: !1656 merge master into master [Feature]:The qwen2MoT of Bagel Model Created-by: ningmenglh Commit-by: ningmengliu Merged-by: ascend-robot Description: ## Motivation support Bagel Model ## Modification Add the backbone model in the Bagel Model. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1656 | 8 个月前 | |
[bugfix]qwen3vl ulysses bugfix Co-authored-by: zs-derrick1<1434012475@qq.com> # message auto-generated for no-merge-commit merge: !1973 merge ring into master [bugfix]qwen3vl ulysses bugfix Created-by: zs-derrick1 Commit-by: zs-derrick1 Merged-by: ascend-robot Description: ## Motivation qwen3vl ulysses bugfix ## Modification qwen3vl ulysses bugfix ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1973 | 6 个月前 | |
[mindspore] support qwen3vl vision for megatron backend Co-authored-by: ffmh<fengminghao2@huawei.com> # message auto-generated for no-merge-commit merge: !1784 merge qwen3vl_vision into master [mindspore] support qwen3vl vision for megatron backend Created-by: ffmh Commit-by: ffmh Merged-by: ascend-robot Description: ## Motivation Please describe the motivation of this PR and the goal you want to achieve through this PR. 适配 qwen3vl vision 模块模型适配 ## Modification Please briefly describe what modification is made in this PR. 添加 Qwen3VLVisionModel，适配deepstack计算相关逻辑 ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1784 | 7 个月前 | |
| 11 个月前 | ||
| 11 个月前 | ||
| 1 年前 | ||
delete dist-train Co-authored-by: lu-jinfu1999<lujinfu1@h-partners.com> # message auto-generated for no-merge-commit merge: !1835 merge master into master [Modify] delete dist-train from master Created-by: lu-jinfu1999 Commit-by: lu-jinfu1999 Merged-by: ascend-robot Description: ## Motivation delete dist-train. ## Modification delete dist-train. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1835 | 7 个月前 | |
| 10 个月前 | ||
[Feature]:Bagel Model Co-authored-by: ningmengliu<liuhao438@huawei.com> # message auto-generated for no-merge-commit merge: !1621 merge master into master [Feature]:Bagel Model Created-by: ningmenglh Commit-by: ningmengliu Merged-by: ascend-robot Description: ## Motivation Submit Bagel model ## Modification Added the Bagel model backbone along with training interfaces, established framework APIs, and reused VAE and SigLIP components. Built the qwen2_mot model. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1621 | 8 个月前 | |
| 11 个月前 | ||
[mindspore] support qwen3vl finetune for megatron backend Co-authored-by: ffmh<fengminghao2@huawei.com> # message auto-generated for no-merge-commit merge: !1785 merge qwen3vl_megatron into master [mindspore] support qwen3vl finetune for megatron backend Created-by: ffmh Commit-by: ffmh Merged-by: ascend-robot Description: ## Motivation Please describe the motivation of this PR and the goal you want to achieve through this PR. 适配 qwen3vl 8B/30B megatron后端 ## Modification Please briefly describe what modification is made in this PR. 添加 layer_spec修改、text模块transformer block/layer添加deepstack相关适配 ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1785 | 7 个月前 | |
[Feature] support sparse-mode3 FA in videoalign Co-authored-by: chenpeizhe<chenpeizhe1@huawei.com> # message auto-generated for no-merge-commit merge: !1570 merge master into master [Feature] support sparse-mode3 FA in videoalign Created-by: chenpeizhe Commit-by: chenpeizhe Merged-by: ascend-robot Description: ## Motivation support sparse-mode3 FA in videoalign ## Modification support sparse-mode3 FA in videoalign ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1570 | 8 个月前 | |
[Bugfix] Layerzero saving weight bugfix Co-authored-by: zs-derrick1<1434012475@qq.com> # message auto-generated for no-merge-commit merge: !1919 merge wanbugfix into master [Bugfix] Layerzero saving weight bugfix Created-by: zs-derrick1 Commit-by: zs-derrick1 Merged-by: ascend-robot Description: ## Motivation Layerzero saving weight bugfix. **Fix for [#188](https://gitcode.com/Ascend/MindSpeed-MM/issues/188) ** ## Modification Layerzero saving weight bugfix. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1919 | 6 个月前 | |
[Feature] transformers model support aux_loss Co-authored-by: liyingxuan<liyingxuan3@huawei.com> # message auto-generated for no-merge-commit merge: !1965 merge aux_loss into master [Feature] transformers model support aux_loss Created-by: liyx616 Commit-by: liyingxuan Merged-by: ascend-robot Description: ## Motivation transformers model support aux_loss. ## Modification transformers_model支持aux_loss，配置方式： 在model.json中，加入下面的字段 json "loss_cfg": { "router_aux_loss_coef": 0.1 } 其中该系数对应的loss计算为：lm_loss + router_aux_loss_coef * aux_loss 如果不配置该字段，默认不开启aux loss计算，通过该字段开启aux loss计算需要transformers模型本身支持aux loss计算，否则不生效。 遗留问题： 当前per-token-loss通过缩放模型梯度实现，会导致auxloss也被错误的缩放了一个系数，所以有兼容性问题。 ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1965 | 6 个月前 | |
【Bugfix】Correct the spelling errors in the code Co-authored-by: lu-jinfu1999<lujinfu1@h-partners.com> # message auto-generated for no-merge-commit merge: !1670 merge master into master 【Bugfix】Correct the spelling errors in the code Created-by: lu-jinfu1999 Commit-by: lu-jinfu1999 Merged-by: ascend-robot Description: ## Motivation Correct the spelling errors in the code. ## Modification Correct spelling errors in the document. Correct spelling errors in the code. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1670 | 7 个月前 | |
[Modify] Refactor the fsdp2 model Co-authored-by: htwang<wanghaitao60@huawei.com> # message auto-generated for no-merge-commit merge: !1793 merge master into master [Modify] Refactor the fsdp2 model Created-by: htwang Commit-by: htwang Merged-by: ascend-robot Description: ## Motivation 1、修改modelzoo至modelhub 2、修改hf_src至transformers ## Modification 1、修改modelzoo至modelhub 2、修改hf_src至transformers ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/MindSpeed-MM!1793 | 7 个月前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 6 个月前 | ||
| 10 个月前 | ||
| 6 个月前 | ||
| 7 个月前 | ||
| 6 个月前 | ||
| 7 个月前 | ||
| 7 个月前 | ||
| 8 个月前 | ||
| 6 个月前 | ||
| 7 个月前 | ||
| 11 个月前 | ||
| 11 个月前 | ||
| 1 年前 | ||
| 7 个月前 | ||
| 10 个月前 | ||
| 8 个月前 | ||
| 11 个月前 | ||
| 7 个月前 | ||
| 8 个月前 | ||
| 6 个月前 | ||
| 6 个月前 | ||
| 7 个月前 | ||
| 7 个月前 |