Shorten dataset download instructions

README.md

@@ -40,31 +40,14 @@ Each version's README describes its repairs, exclusions and any sampling. `tasks
 
 ## Download retained tasks
 
-Download one version without downloading its images:
-
 ```python
 from huggingface_hub import hf_hub_download
 
-source = "camel-ai__SETA-Env_filtered_stages_3_4_5"
 parquet = hf_hub_download(
-    repo_id="FWeindel/validated-tasks",
+    "FWeindel/validated-tasks",
+    "camel-ai__SETA-Env_filtered_stages_3_4_5/tasks.parquet",
     repo_type="dataset",
-    filename=f"{source}/tasks.parquet",
-    local_dir="data",
 )
 ```
 
-When you pass this original Parquet to the [validation runner](https://github.com/MLI-lab/OT-next-data/blob/main/validation/README.md), it automatically downloads and caches the published Apptainer images needed by the selected tasks. Environments without a published image are built normally.
-
-To download the retained tasks and all available cached images together:
-
-```python
-from huggingface_hub import snapshot_download
-
-snapshot_download(
-    repo_id="FWeindel/validated-tasks",
-    repo_type="dataset",
-    allow_patterns=[f"{source}/tasks.parquet", f"{source}/cached_images/**"],
-    local_dir="data",
-)
-```
+The [validation runner](https://github.com/MLI-lab/OT-next-data/blob/main/validation/README.md) automatically downloads the required cached Apptainer images when you run these tasks. Missing images are built normally.