From 3e86c6a1a88f12859e35fa7e7f07cf4efb5a5335 Mon Sep 17 00:00:00 2001
From: thiswillbeyourgithub
 <26625900+thiswillbeyourgithub@users.noreply.github.com>
Date: Thu, 11 Sep 2025 08:33:30 +0200
Subject: [PATCH 1/4] fix: hilarious typo

Signed-off-by: thiswillbeyourgithub <26625900+thiswillbeyourgithub@users.noreply.github.com>
---
 README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 32c90b2..bb61a35 100644
--- a/README.md
+++ b/README.md
@@ -33,7 +33,7 @@ f
 #    num_rows: 20
 # })
 
-# it has this sgaoe
+# it has this shape
 ds_a = Dataset.from_parquet(str(f)).with_format("torch")
 ds_a[0:2]['hidden_states'].shape # [batch, layers, tokens, hidden_states]
 # torch.Size([2, 25, 1, 896])

From e7aa9e00b7078c54c9e515feb42ecf702a2f6db0 Mon Sep 17 00:00:00 2001
From: thiswillbeyourgithub
 <26625900+thiswillbeyourgithub@users.noreply.github.com>
Date: Thu, 11 Sep 2025 08:33:45 +0200
Subject: [PATCH 2/4] doc: mention the nbs better

Signed-off-by: thiswillbeyourgithub <26625900+thiswillbeyourgithub@users.noreply.github.com>
---
 README.md | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/README.md b/README.md
index bb61a35..2dbd06f 100644
--- a/README.md
+++ b/README.md
@@ -9,7 +9,12 @@ Install using
 pip install git+https://github.com/wassname/activation_store.git
 ```
 
-Example
+
+## Examples
+
+Full examples can be found in the [nbs folder](./nbs).
+
+
 ```py
 layer_groups = {'mlp.down_proj': [
   'model.layers.21.mlp.down_proj',
@@ -47,9 +52,6 @@ git clone https//github.com/wassname/activation_store.git
 uv sync
 ```
 
-see examples in `nbs` folder.
-
-
 ## TODO:
 
 - [x] test compression: it's not worth the [complexity](https://github.com/EleutherAI/elk/blob/84e99a36a5050881d85f1510a2486ce46ac1f942/elk/extraction/extraction.py#L382)

From e44e4fbd1abb5dd199b3c1d660e063d0669d558a Mon Sep 17 00:00:00 2001
From: thiswillbeyourgithub
 <26625900+thiswillbeyourgithub@users.noreply.github.com>
Date: Thu, 11 Sep 2025 08:42:54 +0200
Subject: [PATCH 3/4] doc: use keyword arguments in the example

Signed-off-by: thiswillbeyourgithub <26625900+thiswillbeyourgithub@users.noreply.github.com>
---
 README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 2dbd06f..c0e10b0 100644
--- a/README.md
+++ b/README.md
@@ -30,7 +30,7 @@ layer_groups = {'mlp.down_proj': [
   'model.layers.23.mlp.up_proj']}
 
 # collect activations into a huggingface dataset
-f = activation_store(ds, model, layers=layer_groups)
+f = activation_store(loader=ds, model=model, layers=layer_groups)
 f
 # > Generating train split: 0 examples [00:00, ? examples/s]
 # Dataset({

From a1dd6d3d8527817916a6e181292dfa0d5cb99f6e Mon Sep 17 00:00:00 2001
From: thiswillbeyourgithub
 <26625900+thiswillbeyourgithub@users.noreply.github.com>
Date: Thu, 11 Sep 2025 08:43:42 +0200
Subject: [PATCH 4/4] wording

Signed-off-by: thiswillbeyourgithub <26625900+thiswillbeyourgithub@users.noreply.github.com>
---
 README.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index c0e10b0..19dac9c 100644
--- a/README.md
+++ b/README.md
@@ -1,10 +1,10 @@
 # activation_store
 
-Utility library to persistently store transformer activations on disk.
+Utility library to persistently store transformer activations on disk as huggingface dataset.
 
-These activations can be quite large (layers x batch x sequence x hidden_size), so generating them to disk helps avoid out of memory errors.
+As these activations can be quite large (layers x batch x sequence x hidden_size), generating them to disk helps avoid out of memory errors.
 
-Install using 
+Install using
 ```
 pip install git+https://github.com/wassname/activation_store.git
 ```