T-MAC is a kernel library to directly support mixed-precision matrix multiplication (int1/2/3/4 x int8/fp16/fp32) without the need for dequantization by utilizing lookup tables. T-MAC aims to boost low-bit LLM inference on CPUs. T-MAC already offers support for various low-bit models, including W4A16 from GPTQ/gguf, W2A16 from BitDistiller and W1(.58)A8 from BitNet on OSX/Linux/Windows equipped with ARM/Intel CPUs.
T-MAC achieves a token generation throughput of 22 tokens/sec with a single core and 54 tokens/sec with four cores on M2-Ultra for 3B BitNet, which is a 3x speedup compared to SOTA CPU low-bit framework (llama.cpp). T-MAC can even reach 11 tokens/sec on lower-end devices like Raspberry Pi 5.
We evaluate the token generation performance of different models on four different devices: Apple M2-Ultra, Jetson AGX Orin, Raspberry Pi 5 and Surface Book 3. Check datasheet for more details.
We evaluate BitNet-3B and Llama-2-7B (W2) with T-MAC 2-bit and llama.cpp Q2_K, and evaluate Llama-2-7B (W4) with T-MAC 4-bit and llama.cpp Q4_0.
In addition to providing a significant speedup, T-MAC can also match the same performance using fewer CPU cores. For instance, to reach 40 tokens/sec, a throughput that greatly surpasses human reading speed, T-MAC only requires 2 cores, while llama.cpp requires 8 cores. On Jetson AGX Orin, to achieve 10 tokens/sec, a throughput that already meets human reading speed, T-MAC only requires 2 cores, while llama.cpp uses all 12 cores. T-MAC can meet real-time requirements on less powerful devices equipped with fewer CPU cores like Raspberry Pi 5. By using fewer cores, T-MAC can reserve computational resources for other applications and significantly reduce power and energy consumption, both of which are crucial for edge devices.
T-MAC achieves significant speedup at single-threads and consumes much less CPU cores to reach the same throughput
The throughputs of T-MAC are obtained without fast-aggregation. Users can toggle on fast-aggregation through
-fa
to achieve an additional speedup of 10%~20%.
Our GEMM kernels demonstrate superior performance over SOTA low-bit GEMM on CPU. The following figure shows the speedup compared to llama.cpp for llama-7b kernels during token generation (NUM_THREADS=1):
llama.cpp doesn't provide 1-bit kernel implementation, but we can deduce it from the 2-bit, as it won't bring additional speedup according to the 2/3/4-bit results.
Although we haven't integrated multi-batch (N>1) GEMM into llama.cpp, T-MAC can achieve significant speedup due to reduced computaional cost, which ensures superior performance on prompt evaluation and multi-batch token generation. The following figures shows the speedup compared to llama.cpp using OpenBLAS backend (NUM_THREADS=1):
M2-Ultra is an exception as it is equipped with a specially designed AMX coprocessor to accelerate multi-batch GEMM. However, T-MAC can still achieve comparable performance at 2-bit.
By replacing heavy fused-multiply-add instructions with table lookup instructions, T-MAC significantly reduces power consumption. Combined with the speedup, T-MAC ultimately results in a substantial decrease in total energy consumption.
Multi-threading power/energy consumption on M2-Ultra for three models, M1: Llama-2-7B (W4), M2: Llama-2-7B (W2) and M3: BitNet-3B
Data sampled with powermetrics.
T-MAC achieves comparable 2-bit mpGEMM performance compared to CUDA GPU on Jetson AGX Orin. While the CUDA GPU outperforms the CPU in executing kernels other than mpGEMM, making the end-to-end performance of T-MAC (CPU) slightly slower, T-MAC can deliver considerable savings in power and energy consumption.
Framework | Throughput (tokens/sec) | Power (W) | Energy (J/token) |
---|---|---|---|
llama.cpp (CPU) | 7.08 | 15.0 | 2.12 |
llama.cpp (GPU) | 20.03 | 30.8 | 1.54 |
T-MAC (CPU) | 15.62 | 10.4 | 0.66 |
Throughput/power/energy comparison for Llama-2-7B (W2) on NVIDIA Jetson AGX Orin (NUM_THREADS=12 for CPU)
Data sampled with jetson-stats under power mode MAXN.
- Python (3.8 recommended)
- virtualenv
- cmake>=3.22
First, install cmake
, zstd
(dependency of llvm) and libomp
(dependency of tvm). Homebrew is recommended:
brew install cmake zlib libomp
If
zstd
is installed through homebrew, thancmake
should also be installed through homebrew to ensure thatzstd
can be found bycmake
.
Install t_mac
from the source (please run in a virtualenv
):
git clone --recursive https://github.com/microsoft/T-MAC.git
# in virtualenv
pip install . -v # or pip install -e . -v
source build/t-mac-envs.sh
The command will download clang+llvm and build tvm from source. So it might take a bit of time.
Install cmake>=3.22 from Official Page.
Then install TVM build dependencies:
sudo apt install build-essential libtinfo-dev zlib1g-dev libzstd-dev libxml2-dev
Install t_mac
from the source (please run in a virtualenv
):
git clone --recursive https://github.com/microsoft/T-MAC.git
# in virtualenv
pip install . -v # or pip install -e . -v
source build/t-mac-envs.sh
The command will download clang+llvm and build tvm from source. So it might take a bit of time.
Due to lack of stable clang+llvm prebuilt on Windows, Conda + Visual Studio is recommended to install dependencies.
First, install Visual Studio 2019 and toggle on Desk development with C++
and C++ Clang tools for Windows
. Then, create conda environment within Developer PowerShell for VS 2019
:
git clone --recursive https://github.com/microsoft/T-MAC.git
cd T-MAC
conda env create --file conda\tvm-build-environment.yaml
conda activate tvm-build
If you are using Visual Studio 2022, replace
llvmdev =14.0.6
withllvmdev =17.0.6
in the yaml file.
After that, build TVM with:
cd 3rdparty\tvm
mkdir build
cp cmake\config.cmake build
Append set(USE_LLVM llvm-config)
to build\config.cmake
.
cd build
cmake .. -A x64
cmake --build . --config Release -- /m
Install t_mac
from the source:
cd ..\..\..\ # back to project root directory
$env:MANUAL_BUILD = "1"
$env:PYTHONPATH = "$pwd\3rdparty\tvm\python"
pip install . -v # or pip install -e . -v
The following process could be more complicated. However, if your deployment scenerio doesn't require a native build, you can use WSL/docker and follow the Ubuntu guide.
First, install Visual Studio 2022(/2019) and toggle on Desk development with C++
. Then, create conda environment within Developer PowerShell for VS 20XX
.
git clone --recursive https://github.com/microsoft/T-MAC.git
cd T-MAC
conda env create --file conda\tvm-build-environment.yaml
conda activate tvm-build
Remember to replace llvmdev =14.0.6
with llvmdev =17.0.6
in the yaml file if you are using Visual Studio 2022 (which is recommended on ARM64 for better performance).
After that, build TVM with:
cd 3rdparty\tvm
mkdir build
cp cmake\config.cmake build
Append set(USE_LLVM llvm-config)
to build\config.cmake
.
cd build
cmake .. -A x64 # Build TVM in x64, as Python and dependencies are x64
cmake --build . --config Release -- /m
If you encounter errors like
string sub-command regex, mode replace: regex "$" matched an empty string.
during runningcmake .. -A x64
while building TVM, don't worry, and just runcmake .. -A x64
again. Check this issue of LLVM for more details.
As clang tools in Visual Studio are in fact emulated x64 tools, we recommend to install the native arm64 tools manually.
- Install CMake from Offical Windows ARM installer.
- Download Ninja from Release Page and add to Path.
- Install Clang from Release Page.
Please note:
- The conda environment should only be used while building TVM.
conda deactivate tvm-build
after building TVM.
Starting the following commands in virtualenv and outside of Developer Command Prompt/Powershell for VS to ensure our native clang tools are used.
Install t_mac
from the source:
cd ..\..\..\ # back to project root directory
$env:MANUAL_BUILD = "1"
$env:PYTHONPATH = "$pwd\3rdparty\tvm\python"
# In virtualenv
pip install wmi # To detect the native ARM64 CPU within x86_64 python
pip install . -v # or pip install -e . -v
After that, you can verify the installation through: python -c "import t_mac; print(t_mac.__version__); from tvm.contrib.clang import find_clang; print(find_clang())"
.
Currently, we supports end-to-end inference through llama.cpp integration.
We have provided an all-in-one script. Invoke it with:
pip install 3rdparty/llama.cpp/gguf-py
huggingface-cli download 1bitLLM/bitnet_b1_58-3B --local-dir ${model_dir}
python tools/run_pipeline.py -o ${model_dir}
An example output:
Running STEP.0: Compile kernels
Running command in /Users/user/jianyu/T-MAC/deploy:
python compile.py -o tuned -da -nt 4 -tb -gc -gs 128 -ags 64 -t -m hf-bitnet-3b -r
Running STEP.1: Build T-MAC C++ CMakeFiles
Running command in /Users/user/jianyu/T-MAC/build:
cmake -DCMAKE_INSTALL_PREFIX=/Users/user/jianyu/T-MAC/install ..
Running STEP.2: Install T-MAC C++
Running command in /Users/user/jianyu/T-MAC/build:
cmake --build . --target install --config Release
Running STEP.3: Convert HF to GGUF
Running command in /Users/user/jianyu/T-MAC/3rdparty/llama.cpp:
python convert-hf-to-gguf-t-mac.py /Users/user/Downloads/test_models/hf-bitnet-3B --outtype i2 --outfile /Users/user/Downloads/test_models/hf-bitnet-3B/ggml-model.i2.gguf --kcfg /Users/user/jianyu/T-MAC/install/lib/kcfg.ini
Running STEP.4: Build llama.cpp CMakeFiles
Running command in /Users/user/jianyu/T-MAC/3rdparty/llama.cpp/build:
cmake .. -DLLAMA_TMAC=ON -DCMAKE_PREFIX_PATH=/Users/user/jianyu/T-MAC/install/lib/cmake/t-mac -DCMAKE_BUILD_TYPE=Release -DLLAMA_LLAMAFILE_DEFAULT=OFF -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++
Running STEP.5: Build llama.cpp
Running command in /Users/user/jianyu/T-MAC/3rdparty/llama.cpp/build:
cmake --build . --target main --config Release
Running STEP.6: Run inference
Running command in /Users/user/jianyu/T-MAC/3rdparty/llama.cpp/build:
/Users/user/jianyu/T-MAC/3rdparty/llama.cpp/build/bin/main -m /Users/user/Downloads/test_models/hf-bitnet-3B/ggml-model.i2.gguf -n 128 -t 4 -p Microsoft Corporation is an American multinational corporation and technology company headquartered in Redmond, Washington. -b 1 -ngl 0 -c 2048
Check logs/2024-07-15-17-10-11.log for inference output
Check e2e.md for detailed and advanced usage.
LLM inference incurs significant computational cost. Low-bit quantization, a widely adopted technique, introduces the challenge of mixed-precision GEMM (mpGEMM), which is not directly supported by hardware and requires convert/dequant operations.
We propose the use of a lookup table (LUT) to support mpGEMM. Our method involves the following key technniques:
- Given the low precision of weights, we group one-bit weights (e.g., into groups of 4), precompute all possible partial sums, and then use a LUT to store them.
- We employ shift and accumulate operations to support scalable bits from 1 to 4.
- On a CPU, we utilize tbl/pshuf instructions for fast table lookup.
- We reduce the table size from
$2^n$ to$2^{n-1}$ , incorporating a sign bit to accelerate LUT precomputation.
Our method exhibits several notable characteristics:
- T-MAC shows a linear scaling ratio of FLOPs and inference latency relative to the number of bits. This contrasts with traditional convert-based methods, which fail to achieve additional speedup when reducing from 4 bits to lower bits.
- T-MAC inherently supports bit-wise computation for int1/2/3/4, eliminating the need for dequantization. Furthermore, it accommodates all types of activations (e.g., fp8, fp16, int8) using fast table lookup and add instructions, bypassing the need for poorly supported fused-multiply-add instructions.
- T-MAC holds the potential to realize performance gains across all processing units (PUs).
If you find this repository useful, please use the following BibTeX entry for citation.
@misc{wei2024tmaccpurenaissancetable,
title={T-MAC: CPU Renaissance via Table Lookup for Low-Bit LLM Deployment on Edge},
author={Jianyu Wei and Shijie Cao and Ting Cao and Lingxiao Ma and Lei Wang and Yanyong Zhang and Mao Yang},
year={2024},
eprint={2407.00088},
archivePrefix={arXiv},
primaryClass={cs.DC},
url={https://arxiv.org/abs/2407.00088},
}