refactor

{src/cua_lite → cua_lite}/data/preproc/opencua/README.md

@@ -23,7 +23,7 @@ Standardize AgentNet preprocessing using the Qwen3-VL action spaces. By saving i
 > TODO (zwc): Please scale it up and push the complete dataset to huggingface. The script is running slowly and needs to be sped up.
 ```shell
 # process 1/256 of the dataset
-python src/cua_lite/data/preproc/opencua/opencua.py \
+python cua_lite/data/preproc/opencua/opencua.py \
     --jsonl_path ".data/huggingface/xlangai/AgentNet/agentnet_ubuntu_5k.jsonl" \
     --extracted_images_dir ".data/huggingface/xlangai/AgentNet/extracted_ubuntu_images" \
     --output_path ".data/tmp/scalecua/ubuntu" \
@@ -36,7 +36,7 @@ python src/cua_lite/data/preproc/opencua/opencua.py \
 Unzip the file to separate images and text. This prevents redundant image saving during the text preprocessing later on.
 > TODO (zwc): The script is running slowly and needs to be sped up.
 ```shell
-python src/cua_lite/data/unzip.py \
+python cua_lite/data/unzip.py \
     --input_path ".data/tmp/scalecua/ubuntu/shard-00000-of-00256" \
     --output_path ".data/unzipped/scalecua/ubuntu/shard-00000-of-00256" \
     --overwrite