Update README.md

README.md

@@ -3,7 +3,7 @@ size_categories:
 - 1M<n<10M
 viewer: false
 ---
-# Imagenet.int8: Entire Imagenet dataset in 5GB.
+# Imagenet.int8: Entire Imagenet dataset in 5GB
 
 <p align="center">
   <img src="contents/vae.png" alt="small" width="800">
@@ -36,31 +36,36 @@ So clearly, it doesn't make sense to download entire Imagenet and process with V
 (BTW If you think you'll need higher precision, you can always further fine-tune your model on higher precision. But I doubt that.)
 
 
-# How to use?
+# How do I use this?
 
-First download this.
+First download this. You can use `huggingface-cli` for that. 
 
 ```bash
-do huggingface-cli download --repo-type dataset cloneofsimo/imagenet.int8 --local-dir ./vae_mds
+# Pro tip : use `hf_transfer` to get faster download speed.
+pip install hf_transfer
+export HF_HUB_ENABLE_HF_TRANSFER=True
+# actual download script. 
+huggingface-cli download --repo-type dataset cloneofsimo/imagenet.int8 --local-dir ./vae_mds
 ```
 
+
+
 Then, you need to install [streaming dataset](https://github.com/mosaicml/streaming) to use this. The dataset is MDS format.
 
 ```bash
 pip install mosaicml-streaming
 ```
 
-Then, you can use this dataset like this:
+Then, you can very simply use the dataset like this:
 
-```python
+(for more info on using Mosaic's StreamingDataset and MDS format, [reference here](https://docs.mosaicml.com/projects/streaming/en/stable/index.html))
 
+```python
 from streaming.base.format.mds.encodings import Encoding, _encodings
 import numpy as np
 from typing import Any
 import torch
 from streaming import StreamingDataset
-from diffusers.models import AutoencoderKL
-from diffusers.image_processor import VaeImageProcessor
 
 class uint8(Encoding):
     def encode(self, obj: Any) -> bytes:
@@ -72,7 +77,6 @@ class uint8(Encoding):
 
 _encodings["uint8"] = uint8
 
-
 remote_train_dir = "./vae_mds" # this is the path you installed this dataset.
 local_train_dir = "./local_train_dir"
 
@@ -93,15 +97,18 @@ train_dataloader = torch.utils.data.DataLoader(
 )
 ```
 
-Thats the dataloader! Now, below is the example usage. Notice how you have to reshape as data is flattened.
+By default, batch will have three attributes: `vae_output`, `label`, `label_as_text`.
 
-```
-###### Example Usage. Let's see if we can get the 5th image. BTW shuffle plz
+Thats the dataloader! Now, below is the example usage. Notice how you have to reshape the data back to `(B, 4, 32, 32)` as they are decoded flattened.
+
+```python
+###### Example Usage. Decode back the 5th image. BTW shuffle plz
+from diffusers.models import AutoencoderKL
+from diffusers.image_processor import VaeImageProcessor
 
 model = "stabilityai/your-stable-diffusion-model"
 vae = AutoencoderKL.from_pretrained("stabilityai/sdxl-vae").to("cuda:0")
 
-
 batch = next(iter(train_dataloader))
 
 i = 5
@@ -120,7 +127,9 @@ img.save("5th_image.png")
 
 Enjoy!
 
-# Citation
+# Citations
+
+If you find this material helpful, consider citation!
 
 ```bibtex
 @misc{imagenet_int8,