|
@@ -181,29 +181,6 @@
|
|
|
"tokenizer = tiktoken.get_encoding(\"gpt2\")"
|
|
"tokenizer = tiktoken.get_encoding(\"gpt2\")"
|
|
|
]
|
|
]
|
|
|
},
|
|
},
|
|
|
- {
|
|
|
|
|
- "cell_type": "markdown",
|
|
|
|
|
- "id": "b319df48-8a14-4beb-966c-6b49e255c6ba",
|
|
|
|
|
- "metadata": {},
|
|
|
|
|
- "source": [
|
|
|
|
|
- "In addition, we have to make sure that we use the `format_input_phi` function: "
|
|
|
|
|
- ]
|
|
|
|
|
- },
|
|
|
|
|
- {
|
|
|
|
|
- "cell_type": "markdown",
|
|
|
|
|
- "id": "ed41ec0e-6a5a-44f1-bb3c-b7d0724cb4e1",
|
|
|
|
|
- "metadata": {},
|
|
|
|
|
- "source": [
|
|
|
|
|
- "```python\n",
|
|
|
|
|
- " train_losses, val_losses, tokens_seen = train_model_simple(\n",
|
|
|
|
|
- " model, train_loader, val_loader, optimizer, device,\n",
|
|
|
|
|
- " num_epochs=num_epochs, eval_freq=5, eval_iter=5,\n",
|
|
|
|
|
- " start_context=format_input_phi(val_data[0]), # New: `Use format_input_phi` function\n",
|
|
|
|
|
- " tokenizer=tokenizer\n",
|
|
|
|
|
- " )\n",
|
|
|
|
|
- "```"
|
|
|
|
|
- ]
|
|
|
|
|
- },
|
|
|
|
|
{
|
|
{
|
|
|
"cell_type": "markdown",
|
|
"cell_type": "markdown",
|
|
|
"id": "e0650926-c39f-4442-8116-cb7494416f28",
|
|
"id": "e0650926-c39f-4442-8116-cb7494416f28",
|