Change the Download Command to account for duplicate filenames

Downloading directly via wget produces issues because of the common crawl datasets have duplicate filenames across the various years.
This creates issues pausing and resuming the download process and also verifying checksums.

This PR changes the download command to download using the subdirectory structure present in the url.

e.g.
`https://data.together.xyz/redpajama-data-1T/v1.0.0/arxiv/arxiv_023827cd-7ee8-42e6-aa7b-661731f4c70f.jsonl`

downloads to

`arxiv/arxiv_023827cd-7ee8-42e6-aa7b-661731f4c70f.jsonl`

README.md

@@ -15,10 +15,17 @@ ds = load_dataset("togethercomputer/RedPajama-Data-1T")
 ```
 
 Or you can directly download the files using the following command:
+
 ```
-wget -i https://data.together.xyz/redpajama-data-1T/v1.0.0/urls.txt
+wget 'https://data.together.xyz/redpajama-data-1T/v1.0.0/urls.txt'
+while read line; do
+    dload_loc=${line#https://data.together.xyz/redpajama-data-1T/v1.0.0/}
+    mkdir -p $(dirname $dload_loc)
+    wget "$line" -O "$dload_loc"
+done < urls.txt
 ```
 
+
 A smaller 1B-token sample of the dataset can be found [here](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T-Sample).
 
 A full set of scripts to recreate the dataset from scratch can be found [here](https://github.com/togethercomputer/RedPajama-Data).