yzh119
Follow
Pinned
1,255 contributions in the last year
Activity overview
Contributed to
uwsampl/SparseTIR,
mlc-ai/mlc-llm,
uwsampl/sparsetir-artifact
and 41 other
repositories
Contribution activity
July 2023
Created 20 commits in 6 repositories
Created a pull request in apache/tvm that received 17 comments
[CI] Add ml_dypes dependency for all docker images
The PR #15183 adds ml_dtypes as a dependency of TVM, however, some of the CI docker images do not have ml_dtypes installed, the PR fixes the issue. c…
+2
−1
•
17
comments
Opened 17 other pull requests in 3 repositories
mlc-ai/mlc-llm
7
merged
- [Doc] Clarification on the memory requirement of Llama2-70b-chat model
- [Doc] Announce Llama 2 70B CLI support
- [Doc] Add instructions on installation of AMD drivers
- Add OpenCL+LLVM to compile targets
-
[Refactor] Make
mlc_llma package - [Model] Add YuLan, and update documentation on supported model architectures
- [Bugfix] Make GPTJ compatible with new codebase
mlc-ai/package
1
open
2
merged
3
closed
Reviewed 31 pull requests in 4 repositories
mlc-ai/mlc-llm
16 pull requests
- Add Python build api
- Llama-2: Support Grouped Query Attention
- QKV and MLP matmul combination for LLaMA
-
Introducing
q3f16_1group quantization scheme - QKV and MLP matmul combination for LLaMA
-
[Refactor] Make
mlc_llma package - [Param Manager] suppport auto-gptq checkpoints from param manager.
- [BugFix] PlaceInPrompt behavior
- Add completion endpoint to server and provide QA example
- Adds model category from hf config file
- Supporting GPTBigCode with ParamManager
- Add support for WizardLM
- Obtain Model Type from Config
- gradio polish and minigpt cli
- Add support for Guanaco
- Support GPT-NeoX with ParamManager and embedding separation
apache/tvm
12 pull requests
- [CI] Add ml_dypes dependency for all docker images
- [Unity][Dlight] Fix DecodeGeMV rule for spatial-inner with grouping
- [Unity][Dlight] Rule matmul avoiding blockIdx.z
- [TIR] ThreadAllreduce warp-level primitive support with multi-warp
- [Fix][TIR] LowerThreadAllreduce with correct thread mask
- [Runtime] Flush L2 cache in time eval
- [Testing] Return BenchmarkResult in local_run and rpc_run
- [TIR] Blockize keeping T.init in inner block when outer block does no reduction
- [TIR][Schedule] Derive Nonnegative Bounds from Shape Var
- [TIR] Support cross-threaad reduction lowering with thread-broadcasting rewrite
- [Python] Enhance Wheel Packaging
- [CMake] Support LLVM-16 static linking
mlc-ai/relax
2 pull requests
mlc-ai/package
1 pull request
2
contributions
in private repositories
Jul 14 – Jul 15