{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1a848dfb-4083-4d7c-af83-82e663d1f964",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer\n",
    "\n",
    "MODEL_ID = \"/workspace/mixtral-reasoning-output/checkpoint-275/\"\n",
    "\n",
    "model = AutoModelForCausalLM.from_pretrained(\n",
    "        MODEL_ID,\n",
    "        device_map=\"auto\",\n",
    "        torch_dtype=torch.bfloat16,\n",
    "        attn_implementation=\"flash_attention_2\",\n",
    "        trust_remote_code=True\n",
    "        )\n",
    "        \n",
    "tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "419c3212-c843-4102-850c-ec2e83e5401a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model.eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "002d4810-e64d-4553-9256-d5a95bad07da",
   "metadata": {},
   "outputs": [],
   "source": [
    "system_prompt = \"detailed thinking on\"\n",
    "user_prompt = \"\"\"Triangle $ABC$ has a right angle at $B$. Points $D$ and $E$ are chosen on $\\overline{AC}$ and $\\overline{BC}$, respectively, such that $AB = BE = ED = DC = 2$. Find the area of $\\triangle CDE$.\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c6ba5056-67ea-40d7-a6bb-fdb2d750cfc7",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fix the pad token issue\n",
    "if tokenizer.pad_token is None or tokenizer.pad_token_id == tokenizer.eos_token_id:\n",
    "    tokenizer.pad_token = tokenizer.unk_token\n",
    "    tokenizer.pad_token_id = tokenizer.unk_token_id\n",
    "\n",
    "# Verify the fix\n",
    "print(f\"EOS token ID: {tokenizer.eos_token_id}\")\n",
    "print(f\"PAD token ID: {tokenizer.pad_token_id}\")\n",
    "print(f\"UNK token ID: {tokenizer.unk_token_id}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3af2af88-aa27-4670-a78b-bea00bc07414",
   "metadata": {},
   "outputs": [],
   "source": [
    "messages = [\n",
    "    {\"role\": \"system\", \"content\": system_prompt},\n",
    "    {\"role\": \"user\", \"content\": user_prompt}\n",
    "]\n",
    "\n",
    "# Tokenize input\n",
    "input_ids = tokenizer.apply_chat_template(\n",
    "    messages,\n",
    "    tokenize=True,\n",
    "    add_generation_prompt=True,\n",
    "    return_tensors=\"pt\"\n",
    ").to(\"cuda\")\n",
    "\n",
    "# Create streamer - TextStreamer automatically prints to stdout\n",
    "streamer = TextStreamer(\n",
    "    tokenizer, \n",
    "    skip_special_tokens=False,\n",
    "    skip_prompt=False,\n",
    ")\n",
    "\n",
    "# Generate with streamer - no threading needed with TextStreamer\n",
    "model.generate(\n",
    "    input_ids=input_ids,\n",
    "    pad_token_id=tokenizer.eos_token_id\n",
    "    streamer=streamer,\n",
    "    max_new_tokens=16383,\n",
    "    temperature=0.5,\n",
    "    top_p=0.95,\n",
    "    top_k=40,\n",
    "    repetition_penalty=1.2,\n",
    "    do_sample=True,\n",
    "    #use_cache=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2f4daeb6-02c6-4376-acc8-2b34fbb9fbd7",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}