ACoolFIsh
  • Joined on 2026-03-12
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-04 21:27:22 +08:00
497e723754 [quant_per_channel_cast_fused] perf: 优化 FP32 Plain 与小规模 Rescale
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-04 17:05:49 +08:00
34907bfc6d [quant_per_channel_cast_fused] perf: 优化小规模 Rescale kernel
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-04 14:56:45 +08:00
8869622c27 [quant_per_channel_cast_fused] test: 补强正确性矩阵与性能报告
7edb919d41 [quant_per_channel_cast_fused] feat: 实现四种融合逐通道 FP8 转换算子
4efe72d12e [PR A] 确认规范,同意合并
Compare 3 commits »
ACoolFIsh created pull request ccf-ai-infra/TileOPs-Metax#29 2026-08-04 01:22:06 +08:00
[quant_per_channel_cast_fused] feat: 新增 spec-only Manifest
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-04 01:18:46 +08:00
81a39fda24 [quant_per_channel_cast_fused] feat: 完善 4 variant(shape_rules/roofline/workloads 按队友 review 改进)
76af4399a8 [quant_per_channel_cast_fused] feat: 补充 gather + rescale+expand variant(完整 4 种输入场景)
3519c7186b [quant_per_channel_cast_fused] feat: 补充 gather/expand variant(pos_to_token 场景)
6406409732 [quant_per_channel_cast_fused] feat: 补充 rescale variant + 统一 roofline 整数除法
ba2214600e [quant_per_channel_cast_fused] docs: 补充 scale 计算 semantics(YAML 注释)
Compare 14 commits »
ACoolFIsh closed pull request ccf-ai-infra/TileOPs-Metax#28 2026-08-04 01:13:12 +08:00
[quant_per_channel_cast_fused] feat: 新增 spec-only Manifest
ACoolFIsh pushed to summer-camp-2026 at ACoolFIsh/TileOPs-Metax 2026-08-04 01:09:06 +08:00
feb8825a0d [PR A] 确认规范,同意合并
027cd098cc [PR A] 确认规范,同意合并
a86d92431d merge latest summer-camp-2026 into feat/moe_topk_gate
b2e8c38957 feat(moe): add spec-only topk gate manifest
5eb069274c fix: update shape rules and add new workload for QuantSwiGLUFwdChannelCastTransposeOp
Compare 8 commits »
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-03 23:39:04 +08:00
24ab070002 [quant_per_channel_cast_fused] feat: 完善 4 variant(shape_rules/roofline/workloads 按队友 review 改进)
ACoolFIsh created pull request ccf-ai-infra/TileOPs-Metax#28 2026-08-03 23:00:35 +08:00
[quant_per_channel_cast_fused] feat: 新增 spec-only Manifest
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-03 22:42:51 +08:00
762ea18975 [quant_per_channel_cast_fused] feat: 补充 gather + rescale+expand variant(完整 4 种输入场景)
ACoolFIsh closed pull request ccf-ai-infra/TileOPs-Metax#27 2026-08-03 22:35:17 +08:00
[quant_per_channel_cast_fused] feat: 新增 spec-only Manifest
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-03 22:33:15 +08:00
beafd4d1d9 [quant_per_channel_cast_fused] feat: 补充 gather/expand variant(pos_to_token 场景)
ACoolFIsh created pull request ccf-ai-infra/TileOPs-Metax#27 2026-08-03 22:24:59 +08:00
[quant_per_channel_cast_fused] feat: 新增 spec-only Manifest
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-03 21:58:26 +08:00
b173ab27f3 [quant_per_channel_cast_fused] feat: 补充 rescale variant + 统一 roofline 整数除法
9b2a32694a [quant_per_channel_cast_fused] docs: 补充 scale 计算 semantics(YAML 注释)
2a39bbf587 [quant_per_channel_cast_fused] feat: 新增 spec-only Manifest
802e24b54d [PR A] 确认规范,同意合并
277a2a7748 Merge remote-tracking branch 'upstream/summer-camp-2026' into summer-camp-2026
Compare 10 commits »
ACoolFIsh pushed to summer-camp-2026 at ACoolFIsh/TileOPs-Metax 2026-08-03 21:56:07 +08:00
802e24b54d [PR A] 确认规范,同意合并
277a2a7748 Merge remote-tracking branch 'upstream/summer-camp-2026' into summer-camp-2026
fc4470c89c [PR A] 确认规范,同意合并
166b09e12a [PR A] 确认规范,同意合并
955f6dc889 [moe_get_fused_mapping] feat: 新增 spec-only Manifest
Compare 7 commits »
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-03 21:41:48 +08:00
4c6dd26d0a [quant_per_channel_cast_fused] feat: 补充 rescale variant + 统一 roofline 整数除法
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-03 16:56:03 +08:00
d8bb357318 [quant_per_channel_cast_fused] docs: 补充 scale 计算 semantics(YAML 注释)
ACoolFIsh created pull request ccf-ai-infra/TileOPs-Metax#18 2026-08-03 16:40:34 +08:00
[quant_per_channel_cast_fused] feat: 新增 spec-only Manifest
ACoolFIsh created branch feat/quant-per-channel-cast-fused in ACoolFIsh/TileOPs-Metax 2026-08-03 16:11:55 +08:00
ACoolFIsh pushed to feat/quant-per-channel-cast-fused at ACoolFIsh/TileOPs-Metax 2026-08-03 16:11:55 +08:00
823cdf8ab9 [quant_per_channel_cast_fused] feat: 新增 spec-only Manifest