diff --git a/.gitignore b/.gitignore
index 2561d7e..37b8676 100644
--- a/.gitignore
+++ b/.gitignore
@@ -11,4 +11,5 @@ __pycache__/
hw2/code/checkpoints/
hw2/code/visualized/
hw3/code/data/
-hw3/code/checkpoints/
\ No newline at end of file
+hw3/code/checkpoints/
+hw4/code/workdirs/
\ No newline at end of file
diff --git a/hw4/code/attnvis.ipynb b/hw4/code/attnvis.ipynb
new file mode 100644
index 0000000..b910419
--- /dev/null
+++ b/hw4/code/attnvis.ipynb
@@ -0,0 +1,163 @@
+{
+ "cells": [
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Input texts: +++如梦令\n",
+ "昨夜雨疏风骤。浓睡不消残酒。试问卷帘人，却道海棠依旧。知否。知否。应是绿肥红瘦。\n"
+ ]
+ }
+ ],
+ "source": [
+ "import os\n",
+ "from contextlib import nullcontext\n",
+ "import torch\n",
+ "from model import GPTConfig, GPT\n",
+ "from bertviz import head_view\n",
+ "from dataset import Converter, LMDataset\n",
+ "\n",
+ "# set random seed for reproducibility\n",
+ "seed = 2024\n",
+ "torch.manual_seed(seed)\n",
+ "torch.cuda.manual_seed(seed)\n",
+ "torch.cuda.manual_seed_all(seed)\n",
+ "torch.backends.cudnn.deterministic = True\n",
+ "torch.backends.cuda.matmul.allow_tf32 = True # allow tf32 on matmul\n",
+ "torch.backends.cudnn.allow_tf32 = True # allow tf32 on cudnn\n",
+ "\n",
+ "#################################################\n",
+ "# \n",
+ "model_name = 'mygpt'\n",
+ "ckpt_path = 'workdirs/quansongci'\n",
+ "data_root = 'data/quansongci'\n",
+ "vis_text_path = 'data/vis/vis_1.txt'\n",
+ "#################################################\n",
+ "\n",
+ "device = 'cpu'\n",
+ "\n",
+ "dataset = LMDataset(data_root, 'train')\n",
+ "converter = Converter(dataset.stoi, dataset.itos)\n",
+ "\n",
+ "\n",
+ "with open(vis_text_path, 'r', encoding='utf-8') as f:\n",
+ " start = f.read()\n",
+ "start_ids = converter.single_encode(start)\n",
+ "start_texts = [c for c in start]\n",
+ "x = (torch.tensor(start_ids, dtype=torch.long, device=device)[None, ...])\n",
+ "print(f\"Input texts: {start}\")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "id": "c0792738",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "loading model params from workdirs/quansongci/best.pth\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ ""
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ " \n",
+ " \n",
+ "
\n",
+ " Layer: \n",
+ " \n",
+ " \n",
+ "
\n",
+ "