Skip to content

Commit

Permalink
📝 Small edits to docs
Browse files Browse the repository at this point in the history
  • Loading branch information
i-be-snek committed Oct 1, 2024
1 parent c2825bd commit 0eb1056
Showing 1 changed file with 3 additions and 4 deletions.
7 changes: 3 additions & 4 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,7 @@ The process below describes crucial parts of our pipeline.

#### (Step 0) Run your experiments!

##### Run prompt experiments on OpenAI models
If you use OpenAI models, there is a way to save your cost with running experiments in batch.
We developed a series of prompts for our database as follows
- V_0 is a list of prompts used in the NLP2024 paper (Please use this in https://github.com/VUB-HYDR/Wikimpacts/releases/tag/v0.1)
Expand All @@ -45,22 +46,20 @@ We developed a series of prompts for our database as follows
- V_3_3 is a version based on V_3_2, but we put the infobox and the whole article in the end of each prompt
- V_4 is the one with two prompts for each impact category, one prompt for L1/2 and one for L3
- V_5 is the one with three prompts for each impact category

Before you run our pipeline, please choose a version of prompts to proceed, which can be revised in the beginning of **run_prompts.py**

```shell
from Database.Prompts.prompts import V_3 as target_prompts
```

#### (Step 1) Raw input
#### (Step 1) Raw input
Choose the raw file contains the text you need to process, please use the clear raw file name, for example "wiki_dev_whole_infobox_20240729_70single_events" which indicates the article resource (wiki), the experiment set (dev), the article structure (whole_infobox), the generation date (20240729) and the number of events (70single_events); the api env you want to use, specially for OpenAI models, it's mandatory; the decription of the experiment such as "all_categories_V3", the prompt category such as "all", and the batch file location you want to store the batch file (this is not mandatory, but it's good to check if you create correct batch file)

#### (Step 2) GPT models
Choose the model you want to apply. The default model is `gpt-4o-2024-05-13`

Below is a command example you can refer to run the script:

```shell
```python
poetry run python3 Database/Prompts/run_prompts.py
--filename wiki_dev_whole_infobox_20240729_70single_events.json
--raw_dir Database/Wiki_dev_test_articles
Expand Down

0 comments on commit 0eb1056

Please sign in to comment.