docs: revert changes to lora example

Author: MStokluska

examples/fine-tuning/lora/lora_sft-distributed.ipynb

@@ -2,6 +2,7 @@
   "cells": [
     {
       "cell_type": "markdown",
+      "id": "bb4d8595-c34f-4ab0-a0d7-f41d6aec95cd",
       "metadata": {},
       "source": [
         "## LoRA/QLoRA Fine-Tuning with Kubeflow Trainer and Training Hub on OpenShift AI\n",
@@ -32,22 +33,24 @@
         "```sql\n",
         "SELECT AVG(salary) FROM employees WHERE department = 'engineering'\n",
         "```"
-      ],
-      "id": "bb4d8595-c34f-4ab0-a0d7-f41d6aec95cd"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "ea83ac4b-06f1-4c5d-b9fa-372cd3dd5ad2",
       "metadata": {},
       "source": [
         "## Setup\n",
         "\n",
         "First, import the required dependencies."
-      ],
-      "id": "ea83ac4b-06f1-4c5d-b9fa-372cd3dd5ad2"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "bccf4f5f-244b-4283-a9a1-765f1ff5a89c",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Standard library imports\n",
         "import json\n",
@@ -57,14 +60,14 @@
         "\n",
         "from datasets import load_dataset\n",
         "from kubernetes import client as k8s"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "bccf4f5f-244b-4283-a9a1-765f1ff5a89c"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "79ee4db1-b144-4c93-92f9-37a195033649",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Configure logging to show only essential information\n",
         "logging.basicConfig(\n",
@@ -79,22 +82,22 @@
         "logging.getLogger(\"torch\").setLevel(logging.WARNING)\n",
         "\n",
         "print(\"✅ Logging configured for notebook environment\")"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "79ee4db1-b144-4c93-92f9-37a195033649"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "fdef7bfb-c5b8-49bc-ae03-b1d19bdd6541",
       "metadata": {},
       "source": [
         "## Authenticate to your OpenShift Cluster"
-      ],
-      "id": "fdef7bfb-c5b8-49bc-ae03-b1d19bdd6541"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "88ce03ef-189e-409d-a8e0-8dd486de83e9",
       "metadata": {},
+      "outputs": [],
       "source": [
         "api_server = \"<REPLACE WITH OPENSHIFT SERVER>\"\n",
         "token = \"<REPLACE WITH OPENSHIFT TOKEN>\"\n",
@@ -106,13 +109,11 @@
         "# configuration.verify_ssl = False\n",
         "configuration.api_key = {\"authorization\": f\"Bearer {token}\"}\n",
         "api_client = k8s.ApiClient(configuration)"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "88ce03ef-189e-409d-a8e0-8dd486de83e9"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "5219c28e-5d66-4fdb-b746-a7616336a50e",
       "metadata": {},
       "source": [
         "## 1. Load and Explore the Dataset\n",
@@ -122,23 +123,25 @@
         "    Natural language questions\n",
         "    Database schema context (CREATE TABLE statements)\n",
         "    Corresponding SQL queries"
-      ],
-      "id": "5219c28e-5d66-4fdb-b746-a7616336a50e"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "1d953abf-5777-4a2a-a6e2-541a2a405202",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Load the dataset\n",
         "dataset = load_dataset(\"b-mc2/sql-create-context\", split=\"train\")"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "1d953abf-5777-4a2a-a6e2-541a2a405202"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "1ad43a62-ae35-4f07-8380-ec7998d8b377",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Converting the format of the intial messages.\n",
         "def convert_to_messages(example):\n",
@@ -168,24 +171,24 @@
         "sample_converted = convert_to_messages(dataset[0])\n",
         "print(\"Converted format:\")\n",
         "print(json.dumps(sample_converted, indent=2))"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "1ad43a62-ae35-4f07-8380-ec7998d8b377"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "5940a0b6-4ab7-413e-96da-4c5d48acad53",
       "metadata": {},
       "source": [
         "## 2. Prepare Training Data\n",
         "\n",
         "Training Hub expects data in the chat template format with a messages field containing the conversation. We'll convert each example into a user message (question + context) and an assistant message (SQL query)."
-      ],
-      "id": "5940a0b6-4ab7-413e-96da-4c5d48acad53"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "a2ac0f3e-d897-43ba-9f3e-14369ba1bef7",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Training Dataset Preparation.\n",
         "print(f\"Dataset size: {len(dataset)} examples\")\n",
@@ -227,15 +230,13 @@
         "print(f\"Training data saved to: {training_file}\")\n",
         "print(f\"File size: {training_file.stat().st_size / 1024:.1f} KB\")\n",
         "\n",
-        "data_path = f\"/opt/app-root/src/{PVC_PATH}/lora_text_sql_output/train_data.jsonl\"\n",
+        "data_path = f\"{PVC_PATH}/lora_text_sql_output/train_data.jsonl\"\n",
         "print(data_path)"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "a2ac0f3e-d897-43ba-9f3e-14369ba1bef7"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "9de91052-984d-4b2b-b60e-8c6d3086b94c",
       "metadata": {},
       "source": [
         "## 3. Configure and Run LoRA Training\n",
@@ -251,12 +252,14 @@
         "\n",
         "    load_in_4bit: Enable 4-bit quantization to reduce memory\n",
         "    bnb_4bit_quant_type: Quantization type ('nf4' recommended)\n"
-      ],
-      "id": "9de91052-984d-4b2b-b60e-8c6d3086b94c"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "8016ae6f-a01a-4714-a72b-9080acad89c4",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Training configuration\n",
         "MODEL_NAME = \"Qwen/Qwen2.5-1.5B-Instruct\"\n",
@@ -332,53 +335,53 @@
         "    \"checkpoint_at_epoch\": 2,\n",
         "}\n",
         "params"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "8016ae6f-a01a-4714-a72b-9080acad89c4"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "2302cfd9-42f2-4b94-b7c5-dbb4e4ac559b",
       "metadata": {},
       "source": [
         "## Training with LORA SFT and Kubeflow Trainer\n",
         "Launch a training job via Kubeflow Trainer with configured hyperparameters."
-      ],
-      "id": "2302cfd9-42f2-4b94-b7c5-dbb4e4ac559b"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "3eace042-f662-4b36-9fb0-c6f543c4efb3",
       "metadata": {},
+      "outputs": [],
       "source": [
         "from kubeflow.common.types import KubernetesBackendConfig\n",
         "from kubeflow.trainer import TrainerClient\n",
         "from kubeflow.trainer.rhai import TrainingHubAlgorithms, TrainingHubTrainer\n",
         "\n",
         "backend_cfg = KubernetesBackendConfig(client_configuration=api_client.configuration)\n",
         "client = TrainerClient(backend_cfg)"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "3eace042-f662-4b36-9fb0-c6f543c4efb3"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "bb066eae-39ce-4c20-91e0-86aa7afde30a",
       "metadata": {},
+      "outputs": [],
       "source": [
         "for runtime in client.list_runtimes():\n",
         "    if runtime.name == \"training-hub\":\n",
         "        th_runtime = runtime\n",
         "        print(\"Found runtime: \" + str(th_runtime))"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "bb066eae-39ce-4c20-91e0-86aa7afde30a"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "76ec647f-0c8b-48bb-9acc-69bda8315c1e",
       "metadata": {
         "scrolled": true
       },
+      "outputs": [],
       "source": [
         "from kubeflow.trainer.options.kubernetes import (\n",
         "    ContainerOverride,\n",
@@ -438,35 +441,35 @@
         ")\n",
         "\n",
         "print(job_name)"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "76ec647f-0c8b-48bb-9acc-69bda8315c1e"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "f5dbda75-8ed3-4872-b4c3-e7d495e7d20b",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Follow job logs\n",
         "logs = client.get_job_logs(job_name, follow=True)\n",
         "for line in logs:\n",
         "    print(line)"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "f5dbda75-8ed3-4872-b4c3-e7d495e7d20b"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "3c663d1d-d87c-43d0-95e9-530307252fab",
       "metadata": {},
       "source": [
         "## Loading the Model from the Desired Checkpoints."
-      ],
-      "id": "3c663d1d-d87c-43d0-95e9-530307252fab"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "7e73e162-3046-45a3-86d7-c7464c07d6ed",
       "metadata": {},
+      "outputs": [],
       "source": [
         "import glob\n",
         "import os\n",
@@ -509,14 +512,14 @@
         "    model.eval()\n",
         "    tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)\n",
         "    print(\"Loaded model with HuggingFace/PEFT (CPU compatible)\")"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "7e73e162-3046-45a3-86d7-c7464c07d6ed"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "aa4c95b7-75e1-444c-8245-58ef4f043507",
       "metadata": {},
+      "outputs": [],
       "source": [
         "def generate_sql(question: str, schema: str, max_tokens: int = 256) -> str:\n",
         "    \"\"\"\n",
@@ -564,24 +567,24 @@
         "    )\n",
         "\n",
         "    return response.strip()"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "aa4c95b7-75e1-444c-8245-58ef4f043507"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "3a44ae5a-521b-4335-bc29-e1cb915ee8fa",
       "metadata": {},
       "source": [
         "## Test the Trained Model\n",
         "\n",
         "Let's load the trained model and test it on some SQL generation examples."
-      ],
-      "id": "3a44ae5a-521b-4335-bc29-e1cb915ee8fa"
+      ]
     },
     {
       "cell_type": "code",
+      "execution_count": null,
+      "id": "4f60325a-4216-4bb6-a095-17923831b28c",
       "metadata": {},
+      "outputs": [],
       "source": [
         "# Test with examples from the dataset\n",
         "test_examples = [\n",
@@ -610,20 +613,17 @@
         "    sql = generate_sql(example[\"question\"], example[\"schema\"])\n",
         "    print(f\"Generated SQL: {sql}\")\n",
         "    print(\"-\" * 60)"
-      ],
-      "execution_count": null,
-      "outputs": [],
-      "id": "4f60325a-4216-4bb6-a095-17923831b28c"
+      ]
     },
     {
       "cell_type": "markdown",
+      "id": "6bd68750-7e65-4a7b-ad11-80658ed68ba7",
       "metadata": {},
       "source": [
         "## Final Analysis and Summary\n",
         "In this notebook, we demonstrated how LORA/QLORA can be used fine tuning Qwen 2.5 1.5B Instruct model, \n",
         "we were able to fine tune the model to understand natural languages to sql queries generation."
-      ],
-      "id": "6bd68750-7e65-4a7b-ad11-80658ed68ba7"
+      ]
     }
   ],
   "metadata": {
@@ -647,4 +647,4 @@
   },
   "nbformat": 4,
   "nbformat_minor": 5
-}
+}