README: plain phrasing

README.md

@@ -18,18 +18,16 @@ Precomputed [UMT5-XXL](https://huggingface.co/google/umt5-xxl) text embeddings f
 ([`flex-pi/robotwin_3d`](https://huggingface.co/datasets/flex-pi/robotwin_3d)), as consumed by
 Wan2.2-TI2V-5B / FastWAM. Precomputing these costs substantial GPU time; this cache skips it.
 
-## ⚠️ The cache key is the *templated* prompt, not the raw task string
+## Cache key
 
-Task strings from `meta/tasks.jsonl` are wrapped in a fixed template before encoding. Hashing the
-raw task string will **not** find anything.
+Task strings from `meta/tasks.jsonl` are wrapped in a fixed template before encoding, and the cache
+key is the sha256 of that templated prompt:
 
 ```python
 DEFAULT_PROMPT = "A video recorded from a robot's point of view executing the following instruction: {task}"
 key = hashlib.sha256(DEFAULT_PROMPT.format(task=task).encode("utf-8")).hexdigest()
 ```
 
-Worked example:
-
 | | |
 |---|---|
 | task | `Lift the medium-sized green bottle ensuring it remains upright.` |
@@ -51,14 +49,14 @@ keys in row order):
 | `contexts` | `[N, 128, 4096]` | `bfloat16` |
 | `masks` | `[N, 128]` | `bool` |
 
-`masks` marks valid tokens (mean 40.9 of 128). Values past the mask are **not** zero — they are raw
-T5 outputs. FastWAM zeroes them on load (`context[~mask] = 0`); do the same if your model does not
-apply the mask.
+`masks` marks valid tokens (mean 40.9 of 128). Values past the mask are raw T5 outputs rather than
+zeros. FastWAM zeroes them on load (`context[~mask] = 0`); do the same if your model does not apply
+the mask.
 
 ## Converting back to the per-file `.pt` layout
 
 FastWAM reads a flat directory of `{key}.t5_len128.wan22ti2v5b.pt` files, each a `{"context", "mask"}`
-dict — **not** shards. This script reconstructs exactly that layout:
+dict, rather than shards. This script reconstructs that layout:
 
 ```python
 # unpack_to_pt.py -- rebuild the flat .pt cache from the sharded release.