Upload README.md with huggingface_hub

README.md

@@ -23,10 +23,11 @@ for local Apple Silicon inference with
 [`mlx-gen`](https://github.com/lpalbou/mlx-gen).
 
 The source repository ships FP32 weights (~27 GiB) that inference runtimes cast to BF16 at load
-time. This repack stores the runtime dtypes directly, halving the download to ~13.7 GiB with no
+time. This repack stores the runtime dtypes directly, reducing the download to ~15.6 GiB (42% smaller) with no
 change in output:
 
-- UMT5 text encoder: BF16.
+- UMT5 text encoder: BF16, with the `wo` feed-forward projections stored in FP32 exactly as
+  the runtime keeps them (`_keep_in_fp32_modules`).
 - Renderer transformer: BF16, with the runtime FP32 keep-set (norms, scale-shift tables, and
   embedding layers) stored in FP32 exactly as the loader produces them.
 - VAE: FP32, unchanged.
@@ -63,5 +64,5 @@ for workflows, guidance defaults, and task-specific recipes.
 
 Tensors are bit-exact casts of the pinned source revision
 (`ff4c5d4d2d31365c2ffeb30e9753065ee18f58ce`): every tensor equals the value the mlx-gen loader
-produces from the FP32 original, so generation output is identical to running from the source
-repository at the same settings and seed.
+produces from the FP32 original, and generation output was verified bit-identical to the source
+repository on image and video use cases at the same settings and seed (max pixel diff 0).