Revert notebook

Author: abdulfatir

notebooks/chronos-2-quickstart.ipynb

@@ -37,7 +37,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%pip install 'chronos-forecasting>=2.1[extras]' 'matplotlib'"
+    "%pip install 'chronos-forecasting>=2.1' 'pandas[pyarrow]' 'matplotlib'"
    ]
   },
   {
@@ -1403,9 +1403,7 @@
    "source": [
     "## Fine-Tuning\n",
     "\n",
-    "Chronos-2 supports fine-tuning on your own data. You may either fine-tune all weights of the model (_full fine-tuning_) or a [low rank adapter (LoRA)](https://huggingface.co/docs/peft/en/package_reference/lora), which significantly reduces the number of trainable parameters.\n",
-    "\n",
-    "**Note:** Fine-tuning functionality is intended for advanced users. The default fine-tuning hyperparameters may not always improve accuracy for your specific use case. We recommend experimenting with different hyperparameters. "
+    "Chronos-2 supports fine-tuning on your own data."
    ]
   },
   {
@@ -1417,17 +1415,13 @@
     "\n",
     "The `fit` method accepts:\n",
     "- `inputs`: Time series for fine-tuning (same format as predict_quantiles)\n",
-    "- `finetune_mode`: `\"full\"` or `\"lora\"`\n",
-    "- `lora_config`: The [`LoraConfig`](https://huggingface.co/docs/peft/en/package_reference/lora#peft.LoraConfig), in case `finetune_mode=\"lora\"`\n",
     "- `prediction_length`: Forecast horizon for fine-tuning\n",
     "- `validation_inputs`: Optional validation data (same format as inputs)\n",
-    "- `learning_rate`: Optimizer learning rate (default: 1e-6, we recommend a higher learning rate such as 1e-5 for LoRA)\n",
+    "- `learning_rate`: Optimizer learning rate (default: 1e-5)\n",
     "- `num_steps`: Number of training steps (default: 1000)\n",
     "- `batch_size`: Batch size for training (default: 256)\n",
     "\n",
-    "Returns a new pipeline with the fine-tuned model.\n",
-    "\n",
-    "Please read the docstring for details about specific arguments."
+    "Returns a new pipeline with the fine-tuned model."
    ]
   },
   {
@@ -1514,7 +1508,7 @@
     }
    ],
    "source": [
-    "# Fine-tune the model by default full fine-tuning will be performed\n",
+    "# Fine-tune the model\n",
     "finetuned_pipeline = pipeline.fit(\n",
     "    inputs=train_inputs,\n",
     "    prediction_length=13,\n",
@@ -1565,104 +1559,17 @@
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": 21,
-   "id": "7944046c",
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Could not estimate the number of tokens of the input, floating-point operations will not be computed\n"
-     ]
-    },
-    {
-     "data": {
-      "text/html": [
-       "\n",
-       "    <div>\n",
-       "      \n",
-       "      <progress value='50' max='50' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
-       "      [50/50 00:05, Epoch 1/9223372036854775807]\n",
-       "    </div>\n",
-       "    <table border=\"1\" class=\"dataframe\">\n",
-       "  <thead>\n",
-       " <tr style=\"text-align: left;\">\n",
-       "      <th>Step</th>\n",
-       "      <th>Training Loss</th>\n",
-       "    </tr>\n",
-       "  </thead>\n",
-       "  <tbody>\n",
-       "    <tr>\n",
-       "      <td>10</td>\n",
-       "      <td>0.778100</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <td>20</td>\n",
-       "      <td>0.852700</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <td>30</td>\n",
-       "      <td>0.981700</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <td>40</td>\n",
-       "      <td>0.830200</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <td>50</td>\n",
-       "      <td>0.859900</td>\n",
-       "    </tr>\n",
-       "  </tbody>\n",
-       "</table><p>"
-      ],
-      "text/plain": [
-       "<IPython.core.display.HTML object>"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    }
-   ],
-   "source": [
-    "# Fine-tune the model with LoRA\n",
-    "lora_finetuned_pipeline = pipeline.fit(\n",
-    "    inputs=train_inputs,\n",
-    "    prediction_length=13,\n",
-    "    num_steps=50,  # few fine-tuning steps for a quick demo\n",
-    "    learning_rate=1e-5,\n",
-    "    batch_size=32,\n",
-    "    logging_steps=10,\n",
-    "    finetune_mode=\"lora\",\n",
-    ")"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 22,
-   "id": "44e5d367",
+   "cell_type": "markdown",
+   "id": "91083481",
    "metadata": {},
-   "outputs": [],
    "source": [
-    "# Use the LoRA fine-tuned model for predictions\n",
-    "lora_finetuned_pred_df = lora_finetuned_pipeline.predict_df(\n",
-    "    sales_context_df,\n",
-    "    future_df=sales_future_df,\n",
-    "    prediction_length=13,\n",
-    "    quantile_levels=[0.1, 0.5, 0.9],\n",
-    "    id_column=\"id\",\n",
-    "    timestamp_column=\"timestamp\",\n",
-    "    target=\"Sales\",\n",
-    ")"
+    "**Note:** Fine-tuning functionality is intended for advanced users. The default fine-tuning hyperparameters may not always improve accuracy for your specific use case. We recommend experimenting with different hyperparameters. "
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "7c899976",
+   "cell_type": "markdown",
+   "id": "771d7f6a",
    "metadata": {},
-   "outputs": [],
    "source": []
   }
  ],