torchao INT8/FP8 weight-only quantized MiniMax-H3 (video+audio omni) for consumer GPUs. Group-offload recipe; load with diffusers (main) + torchao.
Chohan
abhishekchohan
AI & ML interests
None yet
Recent Activity
updated a model 8 days ago
abhishekchohan/Qwen3.8-27B-GPTQ-INT4 published a model 8 days ago
abhishekchohan/Qwen3.8-27B-GPTQ-INT4 updated a model 11 days ago
abhishekchohan/Muse-Glimmer-30B-GPTQ-INT4Organizations
None yet
torchao INT8/FP8 weight-only quantized MiniMax-H3 (video+audio omni) for consumer GPUs. Group-offload recipe; load with diffusers (main) + torchao.
FP8 (torchao float8wo) quantized FLUX.2 Klein pipelines. ~2x smaller, near-lossless. Load with diffusers + torchao.
AWQ-quantized Qwen3 models: 32B, 14B, 8B, and 4B