MasterJH5574
Follow
Highlights
- Pro
Pinned
1,411 contributions in the last year
Activity overview
Contribution activity
April 2023
Created 41 commits in 3 repositories
Created 1 repository
- MasterJH5574/web-llm Python
Created a pull request in apache/tvm that received 7 comments
[Unity] BlockBuilder assigning unique tensor names in call_te
This PR changes the naming behavior of call_te. With this PR, all tensors created by call_te will be named in alphabetical order (A, B, C...). When…
+24
−1
•
7
comments
Opened 38 other pull requests in 3 repositories
apache/tvm
15
merged
1
closed
- [MetaSchedule] PostProc not rewriting unroll for purely spatial block
-
[Unity]
enable_warningoption for LegalizeOps and MSApplyDatabase - [Unity][VM] LibComparator using dtype from input
- [Unity][VM] Converting tuple arg to Python tuple
- [Unity][MetaSchedule] BlockCollector focusing on current func
- [TIR][Schedule] Method returning the function being worked on
- [TIR][Schedule] Method returning the function being worked on
- [Fix][TIR][Analysis] Reduction block checking alloc_buffers
- [Fix][TIR] UnifyThreadBinding creating unit loop with annotation
- [Unity] DefaultGPUSchedule working for targets other than CUDA
- [Target] Add Apple M1 GPU tag with 256-thread restriction
- [Runtime] Runtime module property mask for Metal and Vulkan
- [Unity] Memory planning with TIR var upper bound
- [Unity][Analysis] Get symbolic TIR vars from struct info
- [Unity] LiftTransformParams with symbolic shape robustness
- [Unity] Torch-like NN module enhancement
mlc-ai/relax
14
merged
- [Op] Avoid indices in TIR matmul being 0
- [Cherry-Pick][Unity][VM] LibComparator using dtype from input
- [Unity] Update specific builtins for LM
- [UX] Suppress warnings passes
- [Cherry-Pick][Fix][TIR][Analysis] Reduction block checking alloc_buffers (apache/tvm#14589)
- [Cherry-Pick][Fix][TIR] UnifyThreadBinding creating unit loop with annotation (apache/tvm#14588)
- [Fix][TIR][Schedule] Reverse-Compute-Inline for dyn case
- [HotFix] Rename BufferReplacer to avoid linking issue
- [Cherry-Pick][Runtime] Runtime module property mask for Metal and Vulkan (apache/tvm#14524)
- [Cherry-Pick][Unity] DefaultGPUSchedule working for targets other than CUDA
- [Cherry-Pick][Target] Add Apple M1 GPU tag with 256-thread restriction
- [Cherry-Pick][ARITH] Enhance CanProve to handle symbolic bound (#14523)
- [Cherry-Pick][ARITH] Enhance CanonicalSimplify to Simplify ProdDiv
- [TIR] Blockize keeping T.init in inner block when outer block does no…
mlc-ai/web-llm
7
merged
1
closed
- Minor patch on the annotation and indents of TIR func dispatch
- Quantization with optional transposition and fusion with matmul
- Inline embedding lookup into dequantize
- NN GeMM dispatch
- NN GeMM dispatch
- Introducing per-function profiling
- Add transposition option to quantization pass
- Match decode + NT-GeMV + [ewise] pattern
Reviewed 18 pull requests in 3 repositories
apache/tvm
14 pull requests
- [Unity] Improve and reduces possible memory leak RPC debug
-
[Unity]
enable_warningoption for LegalizeOps and MSApplyDatabase - [Unity][CODEGEN] Fix metal codegen when with only single working dim
- [Unity][UX] Symbolic Variables Used in Multiple Functions
- [TensorIR] Fix ComputeAt with perfect symbolic bound
-
[Unity][TOPI] Symbolic shape support for
collapse_sum - [ARITH] Enhance CanonicalSimplify to Simplify ProdDiv
- [Unity][Pass] Enhance Dynamic-aware FuseOps
- [Unity][Op] Negative Log Likelihood Loss
- [Unity] Support Simple Dynamic-Shape-Aware in FuseTIR
- [Unity] Memory planning with TIR var upper bound
-
[Unity][Op] Extend
relax.op.takeop to match behavior withtopi.take. - [Unity][Analysis] Get symbolic TIR vars from struct info
- [Unity][Op] vm.alloc_tensor infer struct info
mlc-ai/web-llm
2 pull requests
mlc-ai/relax
2 pull requests
56
contributions
in private repositories
Apr 4 – Apr 18