error when runing 'Snippet to concepts generation' script

[gnohgnailoug]

Issue: KeyError 'seed' When Running Snippet to Concepts Generation Script

Description

When attempting to run the 'Snippet to concepts generation' script, I encounter a KeyError related to a missing 'seed' key in the input data.

Error Output

  File "/share/home/starcoder2-self-align/src/star_align/self_ossinstruct.py", line 514, in <module>
    asyncio.run(main())
  File "/share/home/anaconda3/envs/starcoder-generate/lib/python3.10/asyncio/runners.py", line 44, in run
    return loop.run_until_complete(main)
  File "/share/home/anaconda3/envs/starcoder-generate/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete
    return future.result()
  File "/share/home/starcoder2-self-align/src/star_align/self_ossinstruct.py", line 411, in main
    kwargs = build_kwargs(args.instruct_mode, example)
  File "/share/home/starcoder2-self-align/src/star_align/self_ossinstruct.py", line 307, in build_kwargs
    kwargs["snippet"] = example["seed"]
KeyError: 'seed'

Command Used

python src/star_align/self_ossinstruct.py \
    --instruct_mode "S->C" \
    --seed_data_files seed.jsonl \
    --max_new_data 50000 \
    --tag concept_gen \
    --temperature 0.7 \
    --seed_code_start_index 0 \
    --model bigcode/starcoder2-15b \
    --num_fewshots 8 \
    --num_batched_requests 32 \
    --num_sample_per_request 1

Source of Seed Data

The seed data file used is sourced from the following URL:
bigcode/python-stack-v1-functions-filtered-sc2

Steps to Reproduce

1. Setup the environment using the provided command.
2. Run the script as shown in the command section.

Actual Behavior

The script fails with a KeyError indicating that the 'seed' key is missing from the input data examples.

Additional Information

• Environment: Conda environment with Python 3.10

[gnohgnailoug]

I found that the dataset bigcode/python-stack-v1-functions-filtered-sc2 did not have a 'seed' key. Adding the following line resolved the issue:

if args.instruct_mode == "S->C":
    raw_dataset = raw_dataset.map(lambda example: {'seed': example['content']} if 'content' in example else example)