Repository navigation

[GGUF] Reduce peak RAM usage by casting dequantized tensors early during load - #45386

Merged

SunMarc merged 5 commits into

huggingface:mainfrom

UsamaKenway:gguf-early-dtype-casting

Apr 20, 2026

Merged

[GGUF] Reduce peak RAM usage by casting dequantized tensors early during load#45386
SunMarc merged 5 commits into
huggingface:mainfrom
UsamaKenway:gguf-early-dtype-casting

Commits on Apr 12, 2026
- [GGUF] cast dequantized tensors to target dtype during load
  UsamaKenway
  committed
  f51fd9d
- Merge branch 'huggingface:main' into gguf-early-dtype-casting
  UsamaKenway
  authored
  b45024e
Commits on Apr 13, 2026
- [GGUF] refac dtype, quantization casting
  UsamaKenway
  committed
  1dc1aab
Commits on Apr 14, 2026
- [GGUF] refac dtype
  UsamaKenway
  committed
  04c27ac
Commits on Apr 15, 2026
- Merge branch 'main' into gguf-early-dtype-casting
  SunMarc
  authored
  7bee92f