dadeba/ggml-temp

Tensor library for machine learning

ggml

Tensor library for machine learning

⚠️ The quantization formats Q4 and Q8 have been updated: ggerganov/llama.cpp#1508 - requantize any old models

Note that this project is under development and not ready for production use.
Some of the development is currently happening in the llama.cpp and whisper.cpp repos

Features

• Written in C
• 16-bit float support
• 4-bit integer quantization support
• Automatic differentiation (WIP in progress)
• ADAM and L-BFGS optimizers
• Optimized for Apple silicon via NEON intrinsics and Accelerate framework
• On x86 architectures utilzes AVX intrinsics
• No third-party dependencies
• Zero memory allocations during runtime

Roadmap

• Example of GPT-2 inference examples/gpt-2
• Example of GPT-J inference examples/gpt-j
• Example of Whisper inference examples/whisper
• Support 4-bit integer quantization ggerganov#27
• Example of Cerebras-GPT inference examples/gpt-2
• Example of FLAN-T5 inference ggerganov#12
• Example of LLaMA inference ggerganov/llama.cpp
• Example of LLaMA training ggerganov/llama.cpp/examples/baby-llama
• Example of BLOOM inference NouamaneTazi/bloomz.cpp
• Example of RWKV inference saharNooby/rwkv.cpp
• Example of SAM inference
• Idea for GPU support: ggerganov/llama.cpp#915
• Example of StableLM (GPT-NeoX) inference examples/gpt-neox
• Example of BERT inference skeskinen/bert.cpp
• Example of 💫 StarCoder inference examples/starcoder
• Example of MPT inference examples/mpt
• Example of Replit inference examples/replit

Whisper inference (example)

With ggml you can efficiently run Whisper inference on the CPU.

Memory requirements:

Model	Disk	Mem
tiny	75 MB	~280 MB
base	142 MB	~430 MB
small	466 MB	~1.0 GB
medium	1.5 GB	~2.6 GB
large	2.9 GB	~4.7 GB

GPT inference (example)

With ggml you can efficiently run GPT-2 and GPT-J inference on the CPU.

Here is how to run the example programs:

# Build ggml + examples
git clone https://github.com/ggerganov/ggml
cd ggml
mkdir build && cd build
cmake ..
make -j4 gpt-2 gpt-j

# Run the GPT-2 small 117M model
../examples/gpt-2/download-ggml-model.sh 117M
./bin/gpt-2 -m models/gpt-2-117M/ggml-model.bin -p "This is an example"

# Run the GPT-J 6B model (requires 12GB disk space and 16GB CPU RAM)
../examples/gpt-j/download-ggml-model.sh 6B
./bin/gpt-j -m models/gpt-j-6B/ggml-model.bin -p "This is an example"

# Run the Cerebras-GPT 111M model
# Download from: https://huggingface.co/cerebras
python3 ../examples/gpt-2/convert-cerebras-to-ggml.py /path/to/Cerebras-GPT-111M/
./bin/gpt-2 -m /path/to/Cerebras-GPT-111M/ggml-model-f16.bin -p "This is an example"

The inference speeds that I get for the different models on my 32GB MacBook M1 Pro are as follows:

Model	Size	Time / Token
GPT-2	117M	5 ms
GPT-2	345M	12 ms
GPT-2	774M	23 ms
GPT-2	1558M	42 ms
---	---	---
GPT-J	6B	125 ms

For more information, checkout the corresponding programs in the examples folder.

Using cuBLAS

# fix the path to point to your CUDA compiler
cmake -DGGML_CUBLAS=ON -DCMAKE_CUDA_COMPILER=/usr/local/cuda-12.1/bin/nvcc ..

Resources

• GGML - Large Language Models for Everyone: a description of the GGML format provided by the maintainers of the llm Rust crate, which provides Rust bindings for GGML
• marella/ctransformers: Python bindings for GGML models.