main

ascend-robotfix: support cp_para as dict and fix is_eod_reset condition

9c4bdd7e创建于 10 天前历史提交

文件	最后提交记录	最后更新时间
attention	fix: support cp_para as dict and fix is_eod_reset condition Co-authored-by: Liz<lizhi166@huawei.com> # message auto-generated for no-merge-commit merge: !89 merge master into main fix: support cp_para as dict and fix is_eod_reset condition Created-by: Liz_ Commit-by: Liz Merged-by: ascend-robot Description: fix: support cp_para as dict and fix is_eod_reset condition + precommit修改 本PR涉及的修改范围包含，cp_para 支持字典和对象取值、is_eod_reset赋值逻辑修复，涉及代码10行左右，其余700+都是precommit格式整改。 针对precommit格式的修改，AI测试： codex修改并提交代码：  claude code + ds v4 pro view代码：  验证通过 See merge request: Ascend/TransformerEngineNPU!89	10 天前
module	refactor(quantized_tensor): replace QuantizedTensorView with tensor tuples Co-authored-by: Muu<koimuu@163.com>	11 天前
ops	refactor(quantized_tensor): replace QuantizedTensorView with tensor tuples Co-authored-by: Muu<koimuu@163.com>	11 天前
optimizers	feat: add FP8 support and fix optimizer state handling Co-authored-by: gcw_61YBRfIt<chuzhenxing@huawei.com>	18 天前
quantization	feat: enhance fp8_blockwise Co-authored-by: junhang<wangjunhang7@huawei.com> # message auto-generated for no-merge-commit merge: !88 merge feat_fp_blockwise into main feat(fp8_blockwise): support fp8_blockwise_grouped_quant Created-by: goodflower9 Commit-by: junhang Merged-by: ascend-robot Description: feat: enhance fp8_blockwise 测试结果见 https://wiki.huawei.com/domains/177655/wiki/373156/WIKI2026061211450503 See merge request: Ascend/TransformerEngineNPU!88	11 天前
tensor	feat: enhance fp8_blockwise Co-authored-by: junhang<wangjunhang7@huawei.com> # message auto-generated for no-merge-commit merge: !88 merge feat_fp_blockwise into main feat(fp8_blockwise): support fp8_blockwise_grouped_quant Created-by: goodflower9 Commit-by: junhang Merged-by: ascend-robot Description: feat: enhance fp8_blockwise 测试结果见 https://wiki.huawei.com/domains/177655/wiki/373156/WIKI2026061211450503 See merge request: Ascend/TransformerEngineNPU!88	11 天前
triton	feat:adaptor moe Co-authored-by: dao_qian<tanao4@huawei.com>	1 个月前
__init__.py	refactor(distributed): Refactor distributed tools and simplify linear parameters Co-authored-by: Muu<koimuu@163.com>	1 个月前
constants.py	feat(FP4): Support W4A4-MXFP4 Co-authored-by: mingzhenwang<wangmingzhen4@huawei.com>	14 天前
cpu_offload.py	tenpu cpu_offload Co-authored-by: clc2025<chenlucong@huawei.com>	1 个月前
distributed.py	refactor(distributed): Refactor distributed tools and simplify linear parameters Co-authored-by: Muu<koimuu@163.com>	1 个月前
fp8.py	feat(FP4): Support W4A4-MXFP4 Co-authored-by: mingzhenwang<wangmingzhen4@huawei.com>	14 天前
jit.py	refactor(fp8): fill fp8 vacancy Co-authored-by: Muu<koimuu@163.com>	2 个月前
permutation.py	permutation相关接口 lazy import triton Co-authored-by: dao_qian<tanao4@huawei.com>	1 个月前
quantized_tensor.py	refactor(quantized_tensor): replace QuantizedTensorView with tensor tuples Co-authored-by: Muu<koimuu@163.com>	11 天前
router.py	adapt aux_loss Co-authored-by: dao_qian<tanao4@huawei.com>	1 个月前
torch_version.py	refactor(fp8): fill fp8 vacancy Co-authored-by: Muu<koimuu@163.com>	2 个月前
utils.py	refactor(grouped_linear,gemm,fp8): overhaul grouped matmul with unified dispatch, NPU version check, and FP8 quantization cleanup Co-authored-by: Muu<koimuu@163.com>	16 天前