embedding/check_embedding.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoModelForCausalLM, AutoTokenizer\n",
    "device = \"cuda\" # the device to load the model onto\n",
    "\n",
    "model = AutoModelForCausalLM.from_pretrained(\n",
    "    \"./Qwen2-0.5B-Instruct\",\n",
    "    torch_dtype=\"auto\",\n",
    "    device_map=\"auto\"\n",
    ")\n",
    "tokenizer = AutoTokenizer.from_pretrained(\"./Qwen2-0.5B-Instruct\")\n",
    "\n",
    "# prompt = \"Give me a short introduction to large language model.\"\n",
    "# messages = [\n",
    "#     {\"role\": \"system\", \"content\": \"You are a helpful assistant.\"},\n",
    "#     {\"role\": \"user\", \"content\": prompt}\n",
    "# ]\n",
    "# text = tokenizer.apply_chat_template(\n",
    "#     messages,\n",
    "#     tokenize=False,\n",
    "#     add_generation_prompt=True\n",
    "# )\n",
    "# model_inputs = tokenizer([text], return_tensors=\"pt\").to(device)\n",
    "\n",
    "# generated_ids = model.generate(\n",
    "#     model_inputs.input_ids,\n",
    "#     max_new_tokens=512\n",
    "# )\n",
    "# generated_ids = [\n",
    "#     output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n",
    "# ]\n",
    "\n",
    "# response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 添加新的special token"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[27, 1805, 29]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 测试原始的tokenizer\n",
    "tokenizer.encode(\"<image>\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "添加新tokenizer方法：\n",
    "\n",
    "1. 在`tokenizer_config.json`文件里面的`added_tokens_decoder`里面，加上这个东西：\n",
    "```\n",
    "\"151646\": {\n",
    "      \"content\": \"<image>\",\n",
    "      \"lstrip\": false,\n",
    "      \"normalized\": false,\n",
    "      \"rstrip\": false,\n",
    "      \"single_word\": false,\n",
    "      \"special\": true\n",
    "    }\n",
    "```\n",
    "\n",
    "2. 在`tokenizer_config.json`文件里面的`additional_special_tokens` 里面加上 `<image>`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[151646]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 重启以后重新运行\n",
    "tokenizer.encode(\"<image>\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 为什么不需要重新训练embedding 和 lm_head?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "151647"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 分词器长度\n",
    "len(tokenizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Embedding(151936, 896)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# embedding长度\n",
    "model.model.embed_tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Qwen2ForCausalLM(\n",
       "  (model): Qwen2Model(\n",
       "    (embed_tokens): Embedding(151936, 896)\n",
       "    (layers): ModuleList(\n",
       "      (0-23): 24 x Qwen2DecoderLayer(\n",
       "        (self_attn): Qwen2SdpaAttention(\n",
       "          (q_proj): Linear(in_features=896, out_features=896, bias=True)\n",
       "          (k_proj): Linear(in_features=896, out_features=128, bias=True)\n",
       "          (v_proj): Linear(in_features=896, out_features=128, bias=True)\n",
       "          (o_proj): Linear(in_features=896, out_features=896, bias=False)\n",
       "          (rotary_emb): Qwen2RotaryEmbedding()\n",
       "        )\n",
       "        (mlp): Qwen2MLP(\n",
       "          (gate_proj): Linear(in_features=896, out_features=4864, bias=False)\n",
       "          (up_proj): Linear(in_features=896, out_features=4864, bias=False)\n",
       "          (down_proj): Linear(in_features=4864, out_features=896, bias=False)\n",
       "          (act_fn): SiLU()\n",
       "        )\n",
       "        (input_layernorm): Qwen2RMSNorm((896,), eps=1e-06)\n",
       "        (post_attention_layernorm): Qwen2RMSNorm((896,), eps=1e-06)\n",
       "      )\n",
       "    )\n",
       "    (norm): Qwen2RMSNorm((896,), eps=1e-06)\n",
       "  )\n",
       "  (lm_head): Linear(in_features=896, out_features=151936, bias=False)\n",
       ")"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Linear(in_features=896, out_features=151936, bias=False)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# lm_head长度\n",
    "model.lm_head"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "embedding 和 lm_head 留了足够的空间，所以不需要微调。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}