UnstableLlama/love-aligned-31b

Link all Love-Aligned EXL3 variants

Link all Love-Aligned EXL3 variants

Update README.md

Keep advanced settings collapsed and simplify prompt examples

Repurpose Space for Love-Aligned-31B BF16 chat

torchao int8 weight-only + flash-linear-attention (bnb int8 was ~4 tok/s)

int8 (bitsandbytes) on large slice; scale GPU duration to max_new_tokens

Add requirements.txt

Add README.md

Add app.py

initial commit