demo_hf.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Copyright 2024 Tsinghua University and ByteDance.\n",
    "#\n",
    "# Licensed under the MIT License (the \"License\");\n",
    "# you may not use this file except in compliance with the License.\n",
    "# You may obtain a copy of the License at\n",
    "#\n",
    "#     https://opensource.org/license/mit\n",
    "#\n",
    "# Unless required by applicable law or agreed to in writing, software\n",
    "# distributed under the License is distributed on an \"AS IS\" BASIS,\n",
    "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
    "# See the License for the specific language governing permissions and\n",
    "# limitations under the License."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Config\n",
    "# Use q8 quantization with BitsAndBytes\n",
    "USE_Q8_QUANTIZATION = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "    A very simple demo to load ChatTS model and use it.\n",
    "\"\"\"\n",
    "from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer, AutoProcessor\n",
    "import torch\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import os\n",
    "\n",
    "if USE_Q8_QUANTIZATION:\n",
    "    from transformers import BitsAndBytesConfig \n",
    "\n",
    "from chatts.encoding_utils import eval_prompt_to_encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a1871e49e92d4a17873ef4a545be8eac",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading checkpoint shards:   0%|          | 0/6 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Set Environment and Load model\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\n",
    "if USE_Q8_QUANTIZATION:\n",
    "    quantization_config = BitsAndBytesConfig(load_in_8bit=True)\n",
    "    model = AutoModelForCausalLM.from_pretrained(\"./ckpt\", trust_remote_code=True, device_map='cuda:0', quantization_config=quantization_config, torch_dtype=\"auto\")\n",
    "else:\n",
    "    model = AutoModelForCausalLM.from_pretrained(\"./ckpt\", trust_remote_code=True, device_map='cuda:0', torch_dtype='float16')\n",
    "tokenizer = AutoTokenizer.from_pretrained(\"./ckpt\", trust_remote_code=True)\n",
    "processor = AutoProcessor.from_pretrained(\"./ckpt\", trust_remote_code=True, tokenizer=tokenizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1200x300 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Generate a time series example\n",
    "SEQ_LEN = 256\n",
    "x = np.arange(SEQ_LEN)\n",
    "# TS1: A simple sin signal with a sudden decrease\n",
    "ts1 = np.sin(x / 10) * 5.0\n",
    "ts1[100:] -= 10.0\n",
    "# TS2: A increasing trend with a upward spike\n",
    "ts2 = x * 0.05\n",
    "ts2[103] += 10.0\n",
    "\n",
    "# Plot TS1 and TS2\n",
    "plt.figure(figsize=(12, 3))\n",
    "plt.subplot(211)\n",
    "plt.plot(ts1)\n",
    "plt.title('TS1')\n",
    "plt.subplot(212)\n",
    "plt.plot(ts2)\n",
    "plt.title('TS2')\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generating...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--------> Generated Answer\n",
      "Yes, there is a correlated local fluctuations between them. An upward spike with an amplitude of 9.88 occurred between point 104 and point 107 in TS2, while a sudden decrease with an amplitude of 12.35 occurred between point 98 and point 103 in TS1, indicating a possible correlation in terms of local fluctuations.\n"
     ]
    }
   ],
   "source": [
    "# Convert time series to encoding\n",
    "prompt = f\"I have 2 time series. TS1 is of length {SEQ_LEN}: <ts><ts/>; TS2 if of length {SEQ_LEN}: <ts><ts/>. Please analyze the local changes in these time series first and then conclude if these time series showing local changes near the same time?\"\n",
    "\n",
    "# Apply Chat Template\n",
    "prompt = f\"<|im_start|>system\\nYou are a helpful assistant.<|im_end|><|im_start|>user\\n{prompt}<|im_end|><|im_start|>assistant\\n\"\n",
    "\n",
    "# Convert to tensor\n",
    "inputs = processor(text=[prompt], timeseries=[ts1, ts2], padding=True, return_tensors=\"pt\")\n",
    "streamer = TextStreamer(tokenizer)\n",
    "\n",
    "# Input into model\n",
    "print('Generating...')\n",
    "outputs = model.generate(\n",
    "                **inputs,\n",
    "                max_new_tokens=300,\n",
    "                streamer=streamer\n",
    "            )\n",
    "\n",
    "# Show output\n",
    "input_len = inputs['attention_mask'][0].sum().item()\n",
    "output = outputs[0][input_len:]\n",
    "text_out = tokenizer.decode(output, skip_special_tokens=True)\n",
    "print('--------> Generated Answer')\n",
    "print(text_out)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.11.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}