spuuntries/ilearnmate-icts 🔢📝 → 🖼️

▶️ 16 runs 📅 Jul 2025 ⚙️ Cog 0.16.0

About

Example Output

Output

Performance Metrics

185.81s Prediction Time

273.72s Total Time

All Input Parameters

{
  "num_examples_per_side": 3,
  "attributes_to_generate": "girly,modestly,verbose,happy"
}

Input Parameters

seed Type: integer: Seed for reproducibility of example generation and vector training. Set to 0 for random behavior.
num_examples_per_side Type: integerDefault: 3Range: 1 - 10: Number of descriptive examples to generate for each side of the contrast. More examples might lead to better vectors but will increase generation time.
attributes_to_generate Type: stringDefault: girly,modestly,verbose,happy: Comma-separated list of attributes for which to generate control vectors (e.g., 'girly,modestly,verbose,happy')

Output Schema

Output

Type: array • Items Type: string • Items Format: uri

Example Execution Logs

--- Processing attribute: 'girly' ---
Generating antonym for 'girly'...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Generated contrastive pair: ('girly', 'masculine')
Generating 3 examples for each side...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Positive Examples:
- A graceful dancer with poised movements.
- An elegant dresser with impeccable taste in fashion.
- A warm-hearted supporter of charitable causes.
- a very girly person
Negative Examples:
- A protective ally in times of conflict.
- A confident decision-maker in leadership positions.
- A loyal companion through life's challenges.
- a very masculine person
Created dataset with 20 entries.
Training control vector...
  0%|          | 0/2 [00:00<?, ?it/s]
 50%|█████     | 1/2 [00:23<00:23, 23.42s/it]
100%|██████████| 2/2 [00:30<00:00, 13.84s/it]
100%|██████████| 2/2 [00:30<00:00, 15.27s/it]
  0%|          | 0/31 [00:00<?, ?it/s]
 26%|██▌       | 8/31 [00:00<00:00, 77.28it/s]
 58%|█████▊    | 18/31 [00:00<00:00, 90.32it/s]
 94%|█████████▎| 29/31 [00:00<00:00, 98.37it/s]
100%|██████████| 31/31 [00:00<00:00, 97.83it/s]
Successfully generated vector for 'girly' and saved to /tmp/tmpcm3a7bx1/girly.gguf.
--- Processing attribute: 'modestly' ---
Generating antonym for 'modestly'...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Generated contrastive pair: ('modestly', 'ostentatiously')
Generating 3 examples for each side...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Positive Examples:
- A humble and reserved individual.
- A considerate and thoughtful colleague.
- A quiet and introspective observer.
- a very modestly person
Negative Examples:
- A grandiose networker.
- An attention-seeking philanthropist.
- A flamboyant storyteller.
- a very ostentatiously person
Created dataset with 20 entries.
Training control vector...
  0%|          | 0/2 [00:00<?, ?it/s]
 50%|█████     | 1/2 [00:21<00:21, 21.94s/it]
100%|██████████| 2/2 [00:28<00:00, 12.74s/it]
100%|██████████| 2/2 [00:28<00:00, 14.12s/it]
  0%|          | 0/31 [00:00<?, ?it/s]
 13%|█▎        | 4/31 [00:00<00:01, 24.88it/s]
 84%|████████▍ | 26/31 [00:00<00:00, 114.62it/s]
100%|██████████| 31/31 [00:00<00:00, 107.54it/s]
Successfully generated vector for 'modestly' and saved to /tmp/tmplxlmsv6_/modestly.gguf.
--- Processing attribute: 'verbose' ---
Generating antonym for 'verbose'...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Generated contrastive pair: ('verbose', 'terse')
Generating 3 examples for each side...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Positive Examples:
- An eloquent storyteller.
- A persuasive debater.
- A detailed explaner.
- a very verbose person
Negative Examples:
- A succinct advisor.
- A pragmatic problem-solver.
- A direct communicator.
- a very terse person
Created dataset with 20 entries.
Training control vector...
  0%|          | 0/2 [00:00<?, ?it/s]
 50%|█████     | 1/2 [00:21<00:21, 21.42s/it]
100%|██████████| 2/2 [00:28<00:00, 12.75s/it]
100%|██████████| 2/2 [00:28<00:00, 14.05s/it]
  0%|          | 0/31 [00:00<?, ?it/s]
  3%|▎         | 1/31 [00:00<00:03,  7.88it/s]
  6%|▋         | 2/31 [00:00<00:03,  8.61it/s]
 32%|███▏      | 10/31 [00:00<00:00, 36.22it/s]
 52%|█████▏    | 16/31 [00:00<00:00, 41.81it/s]
 68%|██████▊   | 21/31 [00:00<00:00, 44.44it/s]
 87%|████████▋ | 27/31 [00:00<00:00, 46.48it/s]
100%|██████████| 31/31 [00:00<00:00, 40.22it/s]
Successfully generated vector for 'verbose' and saved to /tmp/tmpixvevdb9/verbose.gguf.
--- Processing attribute: 'happy' ---
Generating antonym for 'happy'...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Generated contrastive pair: ('happy', 'sad')
Generating 3 examples for each side...
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Setting `pad_token_id` to `eos_token_id`:32000 for open-end generation.
Positive Examples:
- A contagious smile spreader.
- An enthusiastic supporter in group activities.
- A grateful and appreciative recipient of kindness.
- a very happy person
Negative Examples:
- Withdrawing from social interactions.
- Displaying a lack of enthusiasm or interest in activities.
- Speaking softly and hesitantly with a somber tone.
- a very sad person
Created dataset with 20 entries.
Training control vector...
  0%|          | 0/2 [00:00<?, ?it/s]
 50%|█████     | 1/2 [00:22<00:22, 22.88s/it]
100%|██████████| 2/2 [00:30<00:00, 13.65s/it]
100%|██████████| 2/2 [00:30<00:00, 15.03s/it]
  0%|          | 0/31 [00:00<?, ?it/s]
 16%|█▌        | 5/31 [00:00<00:00, 49.63it/s]
 35%|███▌      | 11/31 [00:00<00:00, 50.00it/s]
 68%|██████▊   | 21/31 [00:00<00:00, 71.37it/s]
100%|██████████| 31/31 [00:00<00:00, 82.62it/s]
Successfully generated vector for 'happy' and saved to /tmp/tmpo506n3b0/happy.gguf.

Version Details

Version ID: d857689481b0efc7f0e810023aea89190322cccffafd1ee0840649a5a66b48ba
Version Created: August 4, 2025

Run on Replicate →