Update colab notebook (#12247)

16a00e05 · Ravi Theja · GitHub · e825f669 · 16a00e05 · 16a00e05
Unverified Commit 16a00e05 authored 11 months ago by Ravi Theja Committed by GitHub 11 months ago
--- a/llama-index-packs/llama-index-packs-raft-dataset/examples/raft_dataset.ipynb
+++ b/llama-index-packs/llama-index-packs-raft-dataset/examples/raft_dataset.ipynb
@@ -11,7 +11,11 @@
    "\n",
    "This LlamaPack implements RAFT: Adapting Language Model to Domain Specific RAG [paper](https://arxiv.org/abs/2403.10131)\n",
    "\n",
-    "Retrieval Augmented FineTuning (RAFT) is a training recipe introduced in this paper that aims to improve the performance of large language models (LLMs) in open-book, in-domain question-answering tasks. Given a question and a set of retrieved documents, RAFT trains the LLM to identify and cite verbatim the most relevant sequences from the documents that help answer the question, while ignoring irrelevant or distracting information. By explicitly training the model to distinguish between relevant and irrelevant information and to provide evidence from the relevant documents, RAFT encourages the LLM to develop better reasoning and explanation abilities, ultimately improving its ability to answer questions accurately and rationally in scenarios where additional context or knowledge is available."
+    "Retrieval Augmented FineTuning (RAFT) is a training recipe introduced in this paper that aims to improve the performance of large language models (LLMs) in open-book, in-domain question-answering tasks. Given a question and a set of retrieved documents, RAFT trains the LLM to identify and cite verbatim the most relevant sequences from the documents that help answer the question, while ignoring irrelevant or distracting information. By explicitly training the model to distinguish between relevant and irrelevant information and to provide evidence from the relevant documents, RAFT encourages the LLM to develop better reasoning and explanation abilities, ultimately improving its ability to answer questions accurately and rationally in scenarios where additional context or knowledge is available.\n",
+    "\n",
+    "A key component of RAFT is how the dataset is generated for fine-tuning. Each QA pair also includes an \"oracle\" document from which the answer to the question can be deduced as well as \"distractor\" documents which are irrelevant. During training this forces the model to learn which information is relevant/irrelevant and also memorize domain knowledge.\n",
+    "\n",
+    "In this notebook we will create `RAFT Dataset` using `RAFTDatasetPack` LlamaPack."
   ]
  },
  {
@@ -28,7 +32,8 @@
   "metadata": {},
   "outputs": [],
   "source": [
-    "!pip install llama-index"
+    "!pip install llama-index\n",
+    "!pip install llama-index-packs-raft-dataset"
   ]
  },
  {
@@ -39,7 +44,7 @@
   "source": [
    "import os\n",
    "\n",
-    "os.environ[\"OPENAI_API_KEY\"] = \"sk-\""
+    "os.environ[\"OPENAI_API_KEY\"] = \"<YOUR OPENAI API KEY>\""
   ]
  },
  {

 %% Cell type:markdown id: tags:
 # RAFT Dataset LlamaPack
 <a href="https://colab.research.google.com/github/run-llama/llama_index/blob/main/llama-index-packs/llama-index-packs-raft-dataset/examples/raft_dataset.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
 This LlamaPack implements RAFT: Adapting Language Model to Domain Specific RAG [paper](https://arxiv.org/abs/2403.10131)
 Retrieval Augmented FineTuning (RAFT) is a training recipe introduced in this paper that aims to improve the performance of large language models (LLMs) in open-book, in-domain question-answering tasks. Given a question and a set of retrieved documents, RAFT trains the LLM to identify and cite verbatim the most relevant sequences from the documents that help answer the question, while ignoring irrelevant or distracting information. By explicitly training the model to distinguish between relevant and irrelevant information and to provide evidence from the relevant documents, RAFT encourages the LLM to develop better reasoning and explanation abilities, ultimately improving its ability to answer questions accurately and rationally in scenarios where additional context or knowledge is available.
+A key component of RAFT is how the dataset is generated for fine-tuning. Each QA pair also includes an "oracle" document from which the answer to the question can be deduced as well as "distractor" documents which are irrelevant. During training this forces the model to learn which information is relevant/irrelevant and also memorize domain knowledge.
+In this notebook we will create `RAFT Dataset` using `RAFTDatasetPack` LlamaPack.
 %% Cell type:markdown id: tags:
 #### Installation
 %% Cell type:code id: tags:
 ``` python
 !pip install llama-index
+!pip install llama-index-packs-raft-dataset
 ```
 %% Cell type:code id: tags:
 ``` python
 import os
-os.environ["OPENAI_API_KEY"] = "sk-"
+os.environ["OPENAI_API_KEY"] = "<YOUR OPENAI API KEY>"
 ```
 %% Cell type:markdown id: tags:
 #### Download Data
 %% Cell type:code id: tags:
 ``` python
 !wget --user-agent "Mozilla" "https://raw.githubusercontent.com/run-llama/llama_index/main/docs/docs/examples/data/paul_graham/paul_graham_essay.txt" -O './paul_graham_essay.txt'
 ```
 %% Cell type:code id: tags:
 ``` python
 from llama_index.packs.raft_dataset import RAFTDatasetPack
 ```
 %% Cell type:code id: tags:
 ``` python
 raft_dataset = RAFTDatasetPack("./paul_graham_essay.txt")
 ```
 %% Cell type:code id: tags:
 ``` python
 # Beware of the costs invloved. This will use GPT-4 for dataset creation.
 # It will also take long time based on the file size.
 dataset = raft_dataset.run()
 ```
 %% Cell type:markdown id: tags:
 The above dataset is HuggingFace Dataset format. You can then save it into `.arrow` or `.jsonl` format and use it for further finetuning.
 %% Cell type:code id: tags:
 ``` python
 output_path = "<output path>"
 # Save as .arrow format
 dataset.save_to_disk(output_path)
 # Save as .jsonl format
 dataset.to_json(output_path + ".jsonl")
 ```
 %% Cell type:markdown id: tags:
 #### You can refer to the original implementation [here](https://github.com/ShishirPatil/gorilla/tree/main/raft)

--- a/llama-index-packs/llama-index-packs-raft-dataset/pyproject.toml
+++ b/llama-index-packs/llama-index-packs-raft-dataset/pyproject.toml
@@ -29,7 +29,7 @@ license = "MIT"
 maintainers = ["ravi-theja"]
 name = "llama-index-packs-raft-dataset"
 readme = "README.md"
-version = "0.1.1"
+version = "0.1.2"
 [tool.poetry.dependencies]
 python = ">=3.8.1,<4.0"