Address comments

hegelai · Sep 12, 2023 · 9497015 · 9497015
1 parent e89845e
commit 9497015
Show file tree

Hide file tree

Showing 2 changed files with 16 additions and 6 deletions.
diff --git a/examples/notebooks/FineTuningExperiment.ipynb b/examples/notebooks/FineTuningExperiment.ipynb
@@ -26,14 +26,22 @@
  "## Installations"
  ]
  },
+ {
+ "cell_type": "markdown",
+ "id": "abbc7d2a",
+ "metadata": {},
+ "source": [
+ "You may need to install `prompttools`, `replicate`, and HuggingFace's `datasets` to complete this example."
+ ]
+ },
  {
  "cell_type": "code",
  "execution_count": null,
  "id": "885dabeb",
  "metadata": {},
  "outputs": [],
  "source": [
- "# !pip install --quiet --force-reinstall prompttools"
+ "# !pip install --quiet --force-reinstall prompttools datasets replicate"
  ]
  },
  {
@@ -177,7 +185,7 @@
  "source": [
  "# Fine tuning GPT-3.5 on OpenAI\n",
  "\n",
- "First, we'll create and upload a file for our fine tuning rows, and fine tune the model."
+ "First, we'll create and upload a file for our fine tuning rows, and fine tune the model. The file processing may take a few minutes, and the fine tuning may take a few hours."
  ]
  },
  {
@@ -238,7 +246,7 @@
  "metadata": {},
  "outputs": [],
  "source": [
- "openai.FineTuningJob.create(training_file=\"file-name\", model=\"gpt-3.5-turbo\")"
+ "openai.FineTuningJob.create(training_file=\"YOUR_FILE_NAME_HERE\", model=\"gpt-3.5-turbo\")"
  ]
  },
  {
@@ -549,7 +557,9 @@
  "source": [
  "## Run fine tuning\n",
  "\n",
- "For replicate, you will need to upload your fine-tuning rows to an accessible URL. For our example, we uploaded the file to an S3 bucket and used a signed URL to provide access."
+ "For replicate, you will need to upload your fine-tuning rows to an accessible URL. For our example, we uploaded the file to an S3 bucket and used a signed URL to provide access.\n",
+ "\n",
+ "The fine tuning may take a few hours."
  ]
  },
  {

diff --git a/prompttools/experiment/experiments/_utils.py b/prompttools/experiment/experiments/_utils.py
@@ -42,8 +42,8 @@ def _get_dynamic_columns(df: pd.DataFrame) -> pd.DataFrame:
  unique_counts = df[hashable_columns].nunique()
  columns_with_multiple_unique_values = unique_counts[unique_counts > 1].index
  dfs_to_concat = [df[columns_with_multiple_unique_values], df[unhashable_columns]]
- if 'prompt' in df:
+ if 'prompt' in df and 'prompt' not in dfs_to_concat:
  dfs_to_concat.append(df['prompt'])
- if 'messages' in df:
+ elif 'messages' in df and 'messages' not in dfs_to_concat:
  dfs_to_concat.append(df['messages'])
  return pd.concat(dfs_to_concat, axis=1)