| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
fix(pytorch): fix duplicate transports in mla and rope Co-authored-by: zhyebin01<zhangyebin@h-partners.com> # message auto-generated for no-merge-commit merge: !4309 merge bugfix2 into master fix(pytorch): fix duplicate transports in mla and rope Created-by: zhyebin01 Commit-by: zhyebin01 Merged-by: ascend-robot Description: ## What this PR does / why we need it? fix duplicate transports in mla and rope ## Does this PR introduce any user-facing change? no ## How was this patch tested? pipeline test passed See merge request: Ascend/MindSpeed-LLM!4309 | 4 个月前 | |
[pytorch][refact]delete orm code Co-authored-by: guihaowen666<guihaowen@huawei.com> | 9 个月前 | |
feat: deepseek-v4 support cp+tnd Co-authored-by: wanggangguo<wanggangguo@huawei.com> # message auto-generated for no-merge-commit merge: !4842 merge cp-master into master feat: deepseek-v4 support cp+tnd Created-by: isfrapples Commit-by: wanggangguo Merged-by: ascend-robot Description: ## What this PR does / why we need it? 1. Implements TND layout support for kvallgather CP in DeepSeek-v4 attention 2. Adds prefix KV mode with rank-order gather and per-batch rearrangement 3. Refactors kvallgather kernels into TND/BSND layout dispatchers 4. Fixes indexer loss scale capture to avoid multi-layer backward race ## Does this PR introduce any user-facing change? 1. Adds --ring-degree argument for ring context parallel configuration 2. Requires cmp_residual_k explicitly for TND layout in npu_lightning_indexer 3. DeepSeek4 now supports kvallgather_cp_algo with packed TND sequences 4. No breaking changes to existing BSND/CP paths ## How was this patch tested? 1. Verified step loss alignment between CP=1 and CP=2 with TND packed data 2. Confirmed prefix KV correctness via per-layer indexer loss comparison 3. Checked gradient propagation through indexer loss across CP ranks 4. Validated rotary position embedding alignment under continuous CP sharding fixes#1663 See merge request: Ascend/MindSpeed-LLM!4842 | 22 天前 | |
refactor: QLoRA feature Sunset Co-authored-by: cjy840282<chenjingyi9@huawei.com> # message auto-generated for no-merge-commit merge: !4663 merge qlora into master refactor: QLoRA feature Sunset Created-by: cjy840282 Commit-by: cjy840282 Merged-by: ascend-robot Description: ## What this PR does / why we need it? This PR is for removing qlora feature code, please refer associated issue for details. ## Does this PR introduce any user-facing change? Following internal discussions and the public announcement of the issue, we have decided to remove qlora feature. This will not affect users; they can continue to use qlora feature for training in the previous supported version. ## How was this patch tested? before: TBE ERROR in plog  after: no TBE ERROR in plog  See merge request: Ascend/MindSpeed-LLM!4663 | 1 个月前 | |
feat(torch): add GLM-4.5 scripts Co-authored-by: cjy840282<chenjingyi9@huawei.com> # message auto-generated for no-merge-commit merge: !4369 merge GLM-4.5-new into master feat(torch): add GLM-4.5 scripts Created-by: cjy840282 Commit-by: cjy840282 Merged-by: ascend-robot Description: ## What this PR does / why we need it? add GLM-4.5 scripts. ## Does this PR introduce any user-facing change? support GLM-4.5 lora finetune. ## How was this patch tested? vllm inference is normal. See merge request: Ascend/MindSpeed-LLM!4369 | 4 个月前 | |
[mindspore][patch]del useless patch Co-authored-by: ybwang19<1605891897@qq.com> # message auto-generated for no-merge-commit merge: !3973 merge master into master [mindspore][patch]del useless patch Created-by: ybwang19 Commit-by: ybwang19 Merged-by: ascend-robot Description: 删除部分失效patch See merge request: Ascend/MindSpeed-LLM!3973 | 7 个月前 | |
| 11 个月前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 4 个月前 | ||
| 9 个月前 | ||
| 22 天前 | ||
| 1 个月前 | ||
| 4 个月前 | ||
| 7 个月前 | ||
| 11 个月前 |