bf16

Here are 12 public repositories matching this topic...

angelolamonaca / PyTorch-Precision-Converter

A flexible utility for converting tensor precision in PyTorch models and safetensors files, enabling efficient deployment across various platforms.

machine-learning deep-learning utilities deployment toolkit optimization pytorch fp16 model-compression model-conversion model-checkpoint bf16 tensor-precision efficient-deployment

Updated Aug 24, 2023
Python

jcartu / qwen36-27b-blackwell-inference-study

Star

Systematic 24-hour benchmark study of Qwen3.6-27B inference on dual NVIDIA RTX PRO 6000 Blackwell SM120 (TP=2). 8 experiments comparing repne/vllm fork vs upstream vLLM across FP8/BF16/NVFP4/Q8_0 quants and MTP/DFlash speculative decoding. Peak: 2,083 tok/s at c=32. Quality: KLD vs BF16 = 0.0018 (noise floor).

benchmark inference blackwell bf16 fp8 vllm qwen speculative-decoding qwen3 nvfp4 rtx-pro-6000

Updated Jun 3, 2026
Python

arcxteam / gguf-convert-model

Star

Auto GGUF Converter for HuggingFace Hub Models with Multiple Quantizations (GGUF Format)

machine-learning cmake ai tensorflow transformers huggingface ai-models huggingface-models bf16 safetensors llama-cpp gguf gguf-models gguf-quantization gguf-editor convert-gguf

Updated Feb 11, 2026
Python

PRITHIVSAKTHIUR / DeepSeek-OCR-experimental

Star

DeepSeek-OCR-experimental is an advanced, multi-purpose visual document intelligence and object localization sandbox. Powered by the unredacted prithivMLmods/DeepSeek-OCR-Latest-BF16.I64-v2.0 architecture, this suite is designed to deliver highly accurate, structure-aware image text extractions.

torch pytorch accelerate matplotlib addict gradio opencv-python torchvision i64 huggingface-transformers timm bf16 huggingface-spaces flash-attention-2 deepseek-ocr easydict

Updated Jun 27, 2026
Python

Merserk / ComfyUI-Precision-Detector

Star

Three ComfyUI nodes that show the real runtime precision of MODEL, CLIP, and VAE objects.

ai pytorch quantization precision dtype fp16 int8 bf16 int4 fp8 comfyui fp4 comfyui-custom-node nvfp4 mxfp8

Updated Jul 14, 2026
Python

wpferrell / Bigsmall

Star

Lossless AI model compression - ~34% smaller with bit-identical weights; the autopilot profiles your machine, picks the highest fidelity that runs, and streams models bigger than your RAM.

pytorch autopilot entropy-coding model-compression huggingface lossless-compression kv-cache bf16 llm fp8 safetensors streaming-inference

Updated Jun 13, 2026
Python

CarrotSwordsman / H20-LLM-Cookbook

Star

Reproducible benchmark suite and tuned Triton fused-MoE configs for NVIDIA H20 LLM inference. 24 configs, 36 perf data points, geomean 1.09× / peak 1.74× speedup.

benchmark triton moe hopper h20 kernel-tuning mixture-of-experts bf16 llm fp8 vllm llm-inference qwen mixtral deepseek sglang triton-kernels nvidia-h20 fused-moe

Updated Jun 2, 2026
Python

arithy / softfloatpy

Sponsor

Star

A Python binding of Berkeley SoftFloat

floating-point ieee-754 ieee754 fp16 binary16 quadruple-precision softfloat bfloat16 bfloat bf16 binary128 fp128

Updated Jan 20, 2026
Python

Atri2-code / Distributed-ML-Efficient-Inference

Star

Distributed GPT-2 fine-tuning with PyTorch FSDP and BF16 mixed precision, INT8 post-training quantisation, a custom Triton quantisation kernel achieving 1.4x throughput over unfused PyTorch, and a full FP32 vs BF16 vs INT8 benchmark suite. 13 tests passing.

machine-learning deep-learning pytorch triton int8 distributed-training efficient-inference mixed-precision mlops gpt2 quantisation bf16 llm fsdp

Updated May 23, 2026
Python

nbbllxx0 / Fused-Gather-GEMM-Scatter-Kernels

Star

Matrix-free 3D SIMP topology optimization with fused gather-GEMM-scatter CUDA kernels on NVIDIA RTX 4090. Companion code for arXiv:2604.18020.

cuda scientific-computing gpu-computing topology-optimization cupy finite-element-methods matrix-free tensorcore bf16 rtx4090

Updated Apr 21, 2026
Python

Banaxi-Tech / banana-fp12

Star

FP12 AI Models.

cuda transformers pytorch quantization fp16 model-compression bf16 llm fp8

Updated May 23, 2026
Python

Orbital-ML-Lab / qwen-fp8-mtp-sidecar-for-mtplx

Star

Recover the native MTP predictor missing from the 8-bit MLX Qwen3.8-27B-Uncensored package, build a BF16 sidecar, and reproduce a 15.59 → 48.75 tok/s controlled M4 Max result with MTPLX.

mtp mlx model-conversion apple-silicon bf16 fp8 safetensors local-llm qwen speculative-decoding qwen3 multi-token-prediction m4-max mtplx qwen3-8-27b

Updated Aug 19, 2026
Python

Improve this page

Add a description, image, and links to the bf16 topic page so that developers can more easily learn about it.

Curate this topic

Add this topic to your repo

To associate your repository with the bf16 topic, visit your repo's landing page and select "manage topics."

Learn more