Tim Dettmers
VERIFIED TECHNICAL DOSSIERSources checked

Tim Dettmers

Staff Research Scientist, Ai2 | Creator of bitsandbytes & QLoRA Quantization

2 Verified ArtifactsSource Checked & Attributed

Verified Proof of Work Artifacts

2 items cataloged

Each artifact below represents an authenticated research publication, production code repository, or technical architectural framework directly authored or co-created by Tim Dettmers. Every entry undergoes editorial source verification.

#1
RESEARCH Checked Sep 20, 2026

8-Bit Optimizers via Block-wise Dynamic Quantization with Low-Precision States

A breakthrough algorithmic paper by Tim Dettmers introducing 8-bit Adam and low-precision optimizer states, reducing memory consumption of deep learning training states by 75% without compromising model convergence.

Model & Execution Context:Benchmarked on language modeling, machine translation, and ImageNet classification with PyTorch.
Scope & Limitations

Custom CUDA block-level quantization; optimizer speedups are maximized on modern Tensor Core architectures.

#2
IMPLEMENTATION Checked Sep 20, 2026

bitsandbytes: 8-Bit and 4-Bit CUDA Quantization Primitives

The open-source library behind 8-bit matrix multiplication and 4-bit NormalFloat quantization, enabling consumer GPUs to run and train 70B+ parameter foundation models with negligible accuracy loss.

Model & Execution Context:CUDA, C++, Python, Hugging Face transformers, PEFT, and PyTorch.
Scope & Limitations

Tightly coupled with NVIDIA CUDA architectures; requires compilation of custom C++/CUDA kernels for optimal memory coalescing.