0
代码
代码
Issues
Pull Requests
讨论
Wiki
分析
0
ascend-robotascend-robotstyle: pre-commit autofix cleancode (base check)
37ccda36创建于 13 小时前历史提交
文件最后提交记录最后更新时间
configuration_qwen3_asr.py
Add upstream Qwen3-ASR model source Co-authored-by: ye_qm<yeqiangmao@huawei.com> # message auto-generated for no-merge-commit merge: !2569 merge qwen3-asr-upstream-source-pr into master Add upstream Qwen3-ASR model source Created-by: ye_qm Commit-by: ye_qm Merged-by: ascend-robot Description: ## What this PR does / why we need it? This PR introduces the upstream Qwen3-ASR HuggingFace model source files into MindSpeed-MM as the foundation for later FSDP2 adaptation. Background: Qwen3-ASR is an audio speech recognition model with a thinker architecture, including an audio tower and a text decoder. To support Qwen3-ASR in MindSpeed-MM's pure FSDP2 backend, we first need to import the original model configuration, modeling, and processor implementation as a clean upstream baseline. Main changes: - Add configuration_qwen3_asr.py from the upstream Qwen3-ASR implementation. - Add modeling_qwen3_asr.py from the upstream Qwen3-ASR implementation. - Add processing_qwen3_asr.py from the upstream Qwen3-ASR implementation. This PR intentionally keeps the changes close to the upstream source and does not include MindSpeed-MM-specific dataset, collator, launcher, or training configuration changes. The MindSpeed-MM adaptation logic will be submitted in a follow-up PR. ## Does this PR introduce any user-facing change? No direct user-facing training workflow change is introduced by this PR alone. This PR only adds the upstream Qwen3-ASR model source files under the MindSpeed-MM model directory. It does not register the model in the MindSpeed-MM plugin system and does not add dataset preprocessing, collator support, examples, or training scripts. User-facing Qwen3-ASR usage will be enabled by the follow-up MindSpeed-MM FSDP2 adaptation PR. Related files: - mindspeed_mm/fsdp/models/qwen3_asr/configuration_qwen3_asr.py - mindspeed_mm/fsdp/models/qwen3_asr/modeling_qwen3_asr.py - mindspeed_mm/fsdp/models/qwen3_asr/processing_qwen3_asr.py ## How was this patch tested? Please explain how to verify the correctness and effectiveness of this feature, as well as its usage constraints and limitations. See merge request: Ascend/MindSpeed-MM!25697 天前
modeling_qwen3_asr.py
style: pre-commit autofix cleancode (base check) Co-authored-by: liyingxuan<liyingxuan3@huawei.com> # message auto-generated for no-merge-commit merge: !2616 merge master into master style: pre-commit autofix cleancode (base check) Created-by: liyx616 Commit-by: liyingxuan Merged-by: ascend-robot Description: ## What this PR does / why we need it? Please describe the background and detailed changes of the PR. If it is a bugfix, please attach the related issue. ## Does this PR introduce any user-facing change? Please describe whether the PR will result in any user-facing usage changes. If there is related documentation, please specify its path. ## How was this patch tested? Please explain how to verify the correctness and effectiveness of this feature, as well as its usage constraints and limitations. See merge request: Ascend/MindSpeed-MM!261613 小时前
processing_qwen3_asr.py
Add upstream Qwen3-ASR model source Co-authored-by: ye_qm<yeqiangmao@huawei.com> # message auto-generated for no-merge-commit merge: !2569 merge qwen3-asr-upstream-source-pr into master Add upstream Qwen3-ASR model source Created-by: ye_qm Commit-by: ye_qm Merged-by: ascend-robot Description: ## What this PR does / why we need it? This PR introduces the upstream Qwen3-ASR HuggingFace model source files into MindSpeed-MM as the foundation for later FSDP2 adaptation. Background: Qwen3-ASR is an audio speech recognition model with a thinker architecture, including an audio tower and a text decoder. To support Qwen3-ASR in MindSpeed-MM's pure FSDP2 backend, we first need to import the original model configuration, modeling, and processor implementation as a clean upstream baseline. Main changes: - Add configuration_qwen3_asr.py from the upstream Qwen3-ASR implementation. - Add modeling_qwen3_asr.py from the upstream Qwen3-ASR implementation. - Add processing_qwen3_asr.py from the upstream Qwen3-ASR implementation. This PR intentionally keeps the changes close to the upstream source and does not include MindSpeed-MM-specific dataset, collator, launcher, or training configuration changes. The MindSpeed-MM adaptation logic will be submitted in a follow-up PR. ## Does this PR introduce any user-facing change? No direct user-facing training workflow change is introduced by this PR alone. This PR only adds the upstream Qwen3-ASR model source files under the MindSpeed-MM model directory. It does not register the model in the MindSpeed-MM plugin system and does not add dataset preprocessing, collator support, examples, or training scripts. User-facing Qwen3-ASR usage will be enabled by the follow-up MindSpeed-MM FSDP2 adaptation PR. Related files: - mindspeed_mm/fsdp/models/qwen3_asr/configuration_qwen3_asr.py - mindspeed_mm/fsdp/models/qwen3_asr/modeling_qwen3_asr.py - mindspeed_mm/fsdp/models/qwen3_asr/processing_qwen3_asr.py ## How was this patch tested? Please explain how to verify the correctness and effectiveness of this feature, as well as its usage constraints and limitations. See merge request: Ascend/MindSpeed-MM!25697 天前