Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

FP8 byte dequant: pass enable_fp_fusion at the launch (#14416)

Daniel Hand568d16d3ddf
diff --git a/unsloth/kernels/fp8.py b/unsloth/kernels/fp8.py
index fda34d8d430..4e5564ad3dc 100644
--- a/unsloth/kernels/fp8.py
+++ b/unsloth/kernels/fp8.py
@@ -734,6 +734,7 @@ def _e4m3fn_bytes_dequant(
BLOCK_R = BLOCK_R,
BLOCK_C = BLOCK_C,
LONG_INDEXING = long_indexing(x, weight_scale, out),
+ enable_fp_fusion = True,
num_warps = num_warps,
)
return out