abidlabs's picture
|
download
raw
1.57 kB

Gradio process executor: demo and benchmarks

Supporting artifacts for https://github.com/gradio-app/gradio/pull/14003, moved out of the core library PR. The demo, benchmark runners, historical results, raw samples, and plots are preserved unchanged under demo/process_executor/.

See demo/process_executor/RESULTS.md for measurements and demo/process_executor/README.md for execution commands and limitations. The validation notes in RESULTS.md describe the original benchmark run; subsequent CI fixes and validation are recorded in the PR.

Restore and reproduce

These artifacts preserve the earlier API snapshot. Use commit 659eda098039fc1c60de7a68dda133e8e2aa891c of the process-executor PR with its matching Python client and dependencies installed; the current PR API has since changed. From the checkout root, download these files, preserving their repository-relative paths:

from pathlib import Path
from huggingface_hub import HfApi
api = HfApi()
bucket = "abidlabs/gradio-process-executor-pr14003"
files = [item for item in api.list_bucket_tree(bucket) if item.type == "file" and item.path.startswith("demo/")]
api.download_bucket_files(bucket, [(item.path, Path(item.path)) for item in files])

Then follow the commands in demo/process_executor/README.md. The adapters use the existing sibling Gradio demos, so run them inside the checkout rather than as a standalone package. Restoring adds untracked demo files; they need not be committed.

manifest.json records the source commit and SHA-256 checksums of all original files.

Xet Storage Details

Size:
1.57 kB
·
Xet hash:
c218d1e0e7524266a4232c136ca480bcfa52ed060eabd5ede0db4299999c6c53

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.