Zihao Ye
Zihao Ye
Hi all, I just switched from pyls backend to mspyls backend, but I found the completion list returned by mspyls is not as convenient as pyls, for example (I would...
## 🐛 Bug The Half Precision SpMM is buggy. ## To Reproduce I tried to apply SpMM on both fp16 feature and fp32 feature: ```python import dgl import torch as...
I moved the discussion from tvm forum to here because relax has not been upstreamed yet. In [relax wiki](https://github.com/octoml/relax/wiki/Relax-Architecture-Overview), ```python value = match_shape(lhs, pattern) ``` did a pattern matching that...
This PR supports compilation and deployment of MOSS model, especially moss-moon-003-sft.
# The issue Currently, our tokenizer.cpp port only supports load from [a single json file](https://github.com/mlc-ai/mlc-llm/blob/5bdcc86a632c7105ac2b874d7d255685839dd204/3rdparty/tokenizers-cpp/tokenizers.h#L81-L84), which is the [legacy format](https://huggingface.co/docs/transformers/v4.28.1/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.save_pretrained) of hugging face tokenizer that is only applicable to fast...
Implement #31 .
by @junrushao
LShift, RShift and Pow
# Milestone - [ ] Blitz course to SparseTIR (due Nov 17) - [ ] System Overview (for developers) (due Nov 20) - [ ] Tutorial: SpMM operator and optimization...