Prepare fine-tune for colab
Browse files- finetune_model.ipynb +86 -219
finetune_model.ipynb
CHANGED
@@ -7,6 +7,73 @@
|
|
7 |
"# Finetune DeepSeek Coder 1.3B for NBA Kaggle Database SQLite Generation"
|
8 |
]
|
9 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
{
|
11 |
"cell_type": "markdown",
|
12 |
"metadata": {},
|
@@ -28,185 +95,7 @@
|
|
28 |
}
|
29 |
],
|
30 |
"source": [
|
31 |
-
"
|
32 |
-
"Database Schema and Explanations\n",
|
33 |
-
"\n",
|
34 |
-
"team Table\n",
|
35 |
-
"Stores information about NBA teams.\n",
|
36 |
-
"CREATE TABLE IF NOT EXISTS \"team\" (\n",
|
37 |
-
" \"id\" TEXT PRIMARY KEY, -- Unique identifier for the team\n",
|
38 |
-
" \"full_name\" TEXT, -- Full official name of the team (e.g., \"Los Angeles Lakers\")\n",
|
39 |
-
" \"abbreviation\" TEXT, -- Shortened team name (e.g., \"LAL\")\n",
|
40 |
-
" \"nickname\" TEXT, -- Commonly used nickname for the team (e.g., \"Lakers\")\n",
|
41 |
-
" \"city\" TEXT, -- City where the team is based\n",
|
42 |
-
" \"state\" TEXT, -- State where the team is located\n",
|
43 |
-
" \"year_founded\" REAL -- Year the team was established\n",
|
44 |
-
");\n",
|
45 |
-
"\n",
|
46 |
-
"game Table\n",
|
47 |
-
"Contains detailed statistics for each NBA game, including home and away team performance.\n",
|
48 |
-
"CREATE TABLE IF NOT EXISTS \"game\" (\n",
|
49 |
-
" \"season_id\" TEXT, -- Season identifier, formatted as \"2YYYY\" (e.g., \"21970\" for the 1970 season)\n",
|
50 |
-
" \"team_id_home\" TEXT, -- ID of the home team (matches \"id\" in team table)\n",
|
51 |
-
" \"team_abbreviation_home\" TEXT, -- Abbreviation of the home team\n",
|
52 |
-
" \"team_name_home\" TEXT, -- Full name of the home team\n",
|
53 |
-
" \"game_id\" TEXT PRIMARY KEY, -- Unique identifier for the game\n",
|
54 |
-
" \"game_date\" TIMESTAMP, -- Date the game was played (YYYY-MM-DD format)\n",
|
55 |
-
" \"matchup_home\" TEXT, -- Matchup details including opponent (e.g., \"LAL vs. BOS\")\n",
|
56 |
-
" \"wl_home\" TEXT, -- \"W\" if the home team won, \"L\" if they lost\n",
|
57 |
-
" \"min\" INTEGER, -- Total minutes played in the game\n",
|
58 |
-
" \"fgm_home\" REAL, -- Field goals made by the home team\n",
|
59 |
-
" \"fga_home\" REAL, -- Field goals attempted by the home team\n",
|
60 |
-
" \"fg_pct_home\" REAL, -- Field goal percentage of the home team\n",
|
61 |
-
" \"fg3m_home\" REAL, -- Three-point field goals made by the home team\n",
|
62 |
-
" \"fg3a_home\" REAL, -- Three-point attempts by the home team\n",
|
63 |
-
" \"fg3_pct_home\" REAL, -- Three-point field goal percentage of the home team\n",
|
64 |
-
" \"ftm_home\" REAL, -- Free throws made by the home team\n",
|
65 |
-
" \"fta_home\" REAL, -- Free throws attempted by the home team\n",
|
66 |
-
" \"ft_pct_home\" REAL, -- Free throw percentage of the home team\n",
|
67 |
-
" \"oreb_home\" REAL, -- Offensive rebounds by the home team\n",
|
68 |
-
" \"dreb_home\" REAL, -- Defensive rebounds by the home team\n",
|
69 |
-
" \"reb_home\" REAL, -- Total rebounds by the home team\n",
|
70 |
-
" \"ast_home\" REAL, -- Assists by the home team\n",
|
71 |
-
" \"stl_home\" REAL, -- Steals by the home team\n",
|
72 |
-
" \"blk_home\" REAL, -- Blocks by the home team\n",
|
73 |
-
" \"tov_home\" REAL, -- Turnovers by the home team\n",
|
74 |
-
" \"pf_home\" REAL, -- Personal fouls by the home team\n",
|
75 |
-
" \"pts_home\" REAL, -- Total points scored by the home team\n",
|
76 |
-
" \"plus_minus_home\" INTEGER, -- Plus/minus rating for the home team\n",
|
77 |
-
" \"video_available_home\" INTEGER, -- Indicates whether video is available (1 = Yes, 0 = No)\n",
|
78 |
-
" \"team_id_away\" TEXT, -- ID of the away team\n",
|
79 |
-
" \"team_abbreviation_away\" TEXT, -- Abbreviation of the away team\n",
|
80 |
-
" \"team_name_away\" TEXT, -- Full name of the away team\n",
|
81 |
-
" \"matchup_away\" TEXT, -- Matchup details from the away teamβs perspective\n",
|
82 |
-
" \"wl_away\" TEXT, -- \"W\" if the away team won, \"L\" if they lost\n",
|
83 |
-
" \"fgm_away\" REAL, -- Field goals made by the away team\n",
|
84 |
-
" \"fga_away\" REAL, -- Field goals attempted by the away team\n",
|
85 |
-
" \"fg_pct_away\" REAL, -- Field goal percentage of the away team\n",
|
86 |
-
" \"fg3m_away\" REAL, -- Three-point field goals made by the away team\n",
|
87 |
-
" \"fg3a_away\" REAL, -- Three-point attempts by the away team\n",
|
88 |
-
" \"fg3_pct_away\" REAL, -- Three-point field goal percentage of the away team\n",
|
89 |
-
" \"ftm_away\" REAL, -- Free throws made by the away team\n",
|
90 |
-
" \"fta_away\" REAL, -- Free throws attempted by the away team\n",
|
91 |
-
" \"ft_pct_away\" REAL, -- Free throw percentage of the away team\n",
|
92 |
-
" \"oreb_away\" REAL, -- Offensive rebounds by the away team\n",
|
93 |
-
" \"dreb_away\" REAL, -- Defensive rebounds by the away team\n",
|
94 |
-
" \"reb_away\" REAL, -- Total rebounds by the away team\n",
|
95 |
-
" \"ast_away\" REAL, -- Assists by the away team\n",
|
96 |
-
" \"stl_away\" REAL, -- Steals by the away team\n",
|
97 |
-
" \"blk_away\" REAL, -- Blocks by the away team\n",
|
98 |
-
" \"tov_away\" REAL, -- Turnovers by the away team\n",
|
99 |
-
" \"pf_away\" REAL, -- Personal fouls by the away team\n",
|
100 |
-
" \"pts_away\" REAL, -- Total points scored by the away team\n",
|
101 |
-
" \"plus_minus_away\" INTEGER, -- Plus/minus rating for the away team\n",
|
102 |
-
" \"video_available_away\" INTEGER, -- Indicates whether video is available (1 = Yes, 0 = No)\n",
|
103 |
-
" \"season_type\" TEXT -- Regular season or playoffs\n",
|
104 |
-
");\n",
|
105 |
-
"\n",
|
106 |
-
"other_stats Table\n",
|
107 |
-
"Stores additional statistics, linked to the game table via game_id.\n",
|
108 |
-
"CREATE TABLE IF NOT EXISTS \"other_stats\" (\n",
|
109 |
-
" \"game_id\" TEXT, -- Unique game identifier, matches id column from game table\n",
|
110 |
-
" \"league_id\" TEXT, -- League identifier\n",
|
111 |
-
" \"team_id_home\" TEXT, -- Home team identifier\n",
|
112 |
-
" \"team_abbreviation_home\" TEXT, -- Home team abbreviation\n",
|
113 |
-
" \"team_city_home\" TEXT, -- Home team city\n",
|
114 |
-
" \"pts_paint_home\" INTEGER, -- Points in the paint by the home team\n",
|
115 |
-
" \"pts_2nd_chance_home\" INTEGER, -- Second chance points by the home team\n",
|
116 |
-
" \"pts_fb_home\" INTEGER, -- Fast break points by the home team\n",
|
117 |
-
" \"largest_lead_home\" INTEGER,-- Largest lead by the home team\n",
|
118 |
-
" \"lead_changes\" INTEGER, -- Number of lead changes \n",
|
119 |
-
" \"times_tied\" INTEGER, -- Number of times the score was tied\n",
|
120 |
-
" \"team_turnovers_home\" INTEGER, -- Home team turnovers\n",
|
121 |
-
" \"total_turnovers_home\" INTEGER, -- Total turnovers by the home team\n",
|
122 |
-
" \"team_rebounds_home\" INTEGER, -- Home team rebounds\n",
|
123 |
-
" \"pts_off_to_home\" INTEGER, -- Points off turnovers by the home team\n",
|
124 |
-
" \"team_id_away\" TEXT, -- Away team identifier\n",
|
125 |
-
" \"team_abbreviation_away\" TEXT, -- Away team abbreviation\n",
|
126 |
-
" \"pts_paint_away\" INTEGER, -- Points in the paint by the away team\n",
|
127 |
-
" \"pts_2nd_chance_away\" INTEGER, -- Second chance points by the away team\n",
|
128 |
-
" \"pts_fb_away\" INTEGER, -- Fast break points by the away team\n",
|
129 |
-
" \"largest_lead_away\" INTEGER,-- Largest lead by the away team\n",
|
130 |
-
" \"team_turnovers_away\" INTEGER, -- Away team turnovers\n",
|
131 |
-
" \"total_turnovers_away\" INTEGER, -- Total turnovers by the away team\n",
|
132 |
-
" \"team_rebounds_away\" INTEGER, -- Away team rebounds\n",
|
133 |
-
" \"pts_off_to_away\" INTEGER -- Points off turnovers by the away team\n",
|
134 |
-
");\n",
|
135 |
-
"\n",
|
136 |
-
"\n",
|
137 |
-
"Team Name Information\n",
|
138 |
-
"In the plaintext user questions, only the full team names will be used, but in the queries you may use the full team names or the abbreviations. \n",
|
139 |
-
"The full team names can be used with the game table, while the abbreviations should be used with the other_stats table.\n",
|
140 |
-
"Notice they are separated by the | character in the following list:\n",
|
141 |
-
"\n",
|
142 |
-
"Atlanta Hawks|ATL\n",
|
143 |
-
"Boston Celtics|BOS\n",
|
144 |
-
"Cleveland Cavaliers|CLE\n",
|
145 |
-
"New Orleans Pelicans|NOP\n",
|
146 |
-
"Chicago Bulls|CHI\n",
|
147 |
-
"Dallas Mavericks|DAL\n",
|
148 |
-
"Denver Nuggets|DEN\n",
|
149 |
-
"Golden State Warriors|GSW\n",
|
150 |
-
"Houston Rockets|HOU\n",
|
151 |
-
"Los Angeles Clippers|LAC\n",
|
152 |
-
"Los Angeles Lakers|LAL\n",
|
153 |
-
"Miami Heat|MIA\n",
|
154 |
-
"Milwaukee Bucks|MIL\n",
|
155 |
-
"Minnesota Timberwolves|MIN\n",
|
156 |
-
"Brooklyn Nets|BKN\n",
|
157 |
-
"New York Knicks|NYK\n",
|
158 |
-
"Orlando Magic|ORL\n",
|
159 |
-
"Indiana Pacers|IND\n",
|
160 |
-
"Philadelphia 76ers|PHI\n",
|
161 |
-
"Phoenix Suns|PHX\n",
|
162 |
-
"Portland Trail Blazers|POR\n",
|
163 |
-
"Sacramento Kings|SAC\n",
|
164 |
-
"San Antonio Spurs|SAS\n",
|
165 |
-
"Oklahoma City Thunder|OKC\n",
|
166 |
-
"Toronto Raptors|TOR\n",
|
167 |
-
"Utah Jazz|UTA\n",
|
168 |
-
"Memphis Grizzlies|MEM\n",
|
169 |
-
"Washington Wizards|WAS\n",
|
170 |
-
"Detroit Pistons|DET\n",
|
171 |
-
"Charlotte Hornets|CHA\n",
|
172 |
-
"\n",
|
173 |
-
"Query Guidelines\n",
|
174 |
-
"Use team_name_home and team_name_away to match teams to the game table. Use team_abbreviation_home and team_abbreviation away to match teams to the other_stats table.\n",
|
175 |
-
"\n",
|
176 |
-
"To filter by season, use season_id = '2YYYY'.\n",
|
177 |
-
"\n",
|
178 |
-
"Example: To get statistics from 2005, use a statement like: season_id = '22005'. To get statistics from 1972, use a statement like: season_id = \"21972\". To get statistics from 2015, use a statement like: season_id = \"22015\".\n",
|
179 |
-
"\n",
|
180 |
-
"Ensure queries return relevant columns and avoid unnecessary joins.\n",
|
181 |
-
"\n",
|
182 |
-
"Example User Requests and SQLite Queries\n",
|
183 |
-
"Request:\n",
|
184 |
-
"\"What is the most points the Los Angeles Lakers have ever scored at home?\"\n",
|
185 |
-
"SQLite:\n",
|
186 |
-
"SELECT MAX(pts_home) FROM game WHERE team_name_home = 'Los Angeles Lakers';\n",
|
187 |
-
"\n",
|
188 |
-
"Request:\n",
|
189 |
-
"\"Which teams are located in the state of California?\"\n",
|
190 |
-
"SQLite:\n",
|
191 |
-
"SELECT full_name FROM team WHERE state = 'California';\n",
|
192 |
-
"\n",
|
193 |
-
"Request:\n",
|
194 |
-
"\"Which team had the highest number of team turnovers in an away game?\"\n",
|
195 |
-
"SQLite:\n",
|
196 |
-
"SELECT team_abbreviation_away FROM other_stats ORDER BY team_turnovers_away DESC LIMIT 1;\n",
|
197 |
-
"\n",
|
198 |
-
"Request:\n",
|
199 |
-
"\"Which teams were founded before 1979?\"\n",
|
200 |
-
"SQLite:\n",
|
201 |
-
"SELECT full_name FROM team WHERE year_founded < 1979;\n",
|
202 |
-
"\n",
|
203 |
-
"Request:\n",
|
204 |
-
"\"Find the Boston Celtics largest home victory margin in the 2008 season.\"\n",
|
205 |
-
"SQLite:\n",
|
206 |
-
"SELECT MAX(pts_home - pts_away) AS biggest_win FROM game WHERE team_name_home = 'Boston Celtics' AND season_id = '22008';\n",
|
207 |
-
"\n",
|
208 |
-
"Generate only the SQLite query prefaced by SQLite: and no other text, do not output an explanation of the query. Now generate an SQLite query for the following user request. Request:\n",
|
209 |
-
"\"\"\"\n",
|
210 |
"\n",
|
211 |
"print(len(input_prompt))"
|
212 |
]
|
@@ -220,30 +109,14 @@
|
|
220 |
},
|
221 |
{
|
222 |
"cell_type": "code",
|
223 |
-
"execution_count":
|
224 |
"metadata": {},
|
225 |
"outputs": [
|
226 |
{
|
227 |
"name": "stderr",
|
228 |
"output_type": "stream",
|
229 |
"text": [
|
230 |
-
"
|
231 |
-
" from .autonotebook import tqdm as notebook_tqdm\n"
|
232 |
-
]
|
233 |
-
},
|
234 |
-
{
|
235 |
-
"name": "stdout",
|
236 |
-
"output_type": "stream",
|
237 |
-
"text": [
|
238 |
-
"WARNING:tensorflow:From c:\\Users\\Dean\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\site-packages\\tf_keras\\src\\losses.py:2976: The name tf.losses.sparse_softmax_cross_entropy is deprecated. Please use tf.compat.v1.losses.sparse_softmax_cross_entropy instead.\n",
|
239 |
-
"\n"
|
240 |
-
]
|
241 |
-
},
|
242 |
-
{
|
243 |
-
"name": "stderr",
|
244 |
-
"output_type": "stream",
|
245 |
-
"text": [
|
246 |
-
"C:\\Users\\Dean\\AppData\\Local\\Temp\\ipykernel_6496\\2921743792.py:18: FutureWarning: DataFrame.applymap has been deprecated. Use DataFrame.map instead.\n",
|
247 |
" df = df.applymap(lambda x: re.sub(r'\\s+', ' ', x) if isinstance(x, str) else x)\n"
|
248 |
]
|
249 |
},
|
@@ -274,7 +147,7 @@
|
|
274 |
"name": "stderr",
|
275 |
"output_type": "stream",
|
276 |
"text": [
|
277 |
-
"Map: 100%|ββββββββββ| 1044/1044 [00:
|
278 |
]
|
279 |
},
|
280 |
{
|
@@ -295,22 +168,14 @@
|
|
295 |
}
|
296 |
],
|
297 |
"source": [
|
298 |
-
"
|
299 |
-
"import torch\n",
|
300 |
-
"from datasets import Dataset\n",
|
301 |
-
"from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, Trainer, BitsAndBytesConfig, EarlyStoppingCallback, PreTrainedTokenizer\n",
|
302 |
-
"from torch.utils.data import DataLoader\n",
|
303 |
-
"from peft import LoraConfig, get_peft_model, TaskType\n",
|
304 |
-
"import os\n",
|
305 |
-
"import re\n",
|
306 |
-
"import numpy as np\n",
|
307 |
"\n",
|
308 |
"# Model output directories\n",
|
309 |
-
"MODEL_DIR = \"
|
310 |
-
"VAL_OUTPUT = \"val-16.hf\"\n",
|
311 |
"\n",
|
312 |
"# Load dataset\n",
|
313 |
-
"df = pd.read_csv(\"
|
314 |
"\n",
|
315 |
"df = df.applymap(lambda x: re.sub(r'\\s+', ' ', x) if isinstance(x, str) else x)\n",
|
316 |
"\n",
|
@@ -319,14 +184,16 @@
|
|
319 |
"print(df.head())\n",
|
320 |
"\n",
|
321 |
"# Load tokenizer\n",
|
322 |
-
"model_name = \"
|
323 |
"tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
|
324 |
"\n",
|
325 |
"# Enable 8-bit quantization for lower memory usage\n",
|
326 |
-
"bnb_config =
|
327 |
-
"
|
328 |
-
"
|
329 |
-
"
|
|
|
|
|
330 |
"\n",
|
331 |
"# Load model with quantization\n",
|
332 |
"#device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
|
@@ -870,7 +737,7 @@
|
|
870 |
},
|
871 |
{
|
872 |
"cell_type": "code",
|
873 |
-
"execution_count":
|
874 |
"metadata": {},
|
875 |
"outputs": [
|
876 |
{
|
@@ -893,7 +760,7 @@
|
|
893 |
"print(prompt_length)\n",
|
894 |
"\n",
|
895 |
"# Create connection to sqlite3 database\n",
|
896 |
-
"connection = sql.connect('
|
897 |
"cursor = connection.cursor()\n",
|
898 |
"\n",
|
899 |
"for v in val_dataset:\n",
|
@@ -4248,7 +4115,7 @@
|
|
4248 |
],
|
4249 |
"metadata": {
|
4250 |
"kernelspec": {
|
4251 |
-
"display_name": "
|
4252 |
"language": "python",
|
4253 |
"name": "python3"
|
4254 |
},
|
@@ -4262,7 +4129,7 @@
|
|
4262 |
"name": "python",
|
4263 |
"nbconvert_exporter": "python",
|
4264 |
"pygments_lexer": "ipython3",
|
4265 |
-
"version": "3.
|
4266 |
}
|
4267 |
},
|
4268 |
"nbformat": 4,
|
|
|
7 |
"# Finetune DeepSeek Coder 1.3B for NBA Kaggle Database SQLite Generation"
|
8 |
]
|
9 |
},
|
10 |
+
{
|
11 |
+
"cell_type": "code",
|
12 |
+
"execution_count": null,
|
13 |
+
"metadata": {},
|
14 |
+
"outputs": [
|
15 |
+
{
|
16 |
+
"name": "stderr",
|
17 |
+
"output_type": "stream",
|
18 |
+
"text": [
|
19 |
+
"/opt/anaconda3/envs/CSCI544/lib/python3.11/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
|
20 |
+
" from .autonotebook import tqdm as notebook_tqdm\n"
|
21 |
+
]
|
22 |
+
}
|
23 |
+
],
|
24 |
+
"source": [
|
25 |
+
"import pandas as pd\n",
|
26 |
+
"import torch\n",
|
27 |
+
"from datasets import Dataset\n",
|
28 |
+
"from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, Trainer, BitsAndBytesConfig, EarlyStoppingCallback, PreTrainedTokenizer\n",
|
29 |
+
"from torch.utils.data import DataLoader\n",
|
30 |
+
"import sys\n",
|
31 |
+
"from peft import LoraConfig, get_peft_model, TaskType\n",
|
32 |
+
"from huggingface_hub import snapshot_download\n",
|
33 |
+
"import os\n",
|
34 |
+
"import re\n",
|
35 |
+
"import numpy as np"
|
36 |
+
]
|
37 |
+
},
|
38 |
+
{
|
39 |
+
"cell_type": "code",
|
40 |
+
"execution_count": null,
|
41 |
+
"metadata": {},
|
42 |
+
"outputs": [],
|
43 |
+
"source": [
|
44 |
+
"is_google_colab = False\n",
|
45 |
+
"use_bnb = True"
|
46 |
+
]
|
47 |
+
},
|
48 |
+
{
|
49 |
+
"cell_type": "code",
|
50 |
+
"execution_count": null,
|
51 |
+
"metadata": {},
|
52 |
+
"outputs": [],
|
53 |
+
"source": [
|
54 |
+
"current_read_path = \"./\"\n",
|
55 |
+
"current_write_path = \"./\"\n",
|
56 |
+
"\n",
|
57 |
+
"def read_path(rel_path):\n",
|
58 |
+
" return os.path.join(current_read_path, rel_path)\n",
|
59 |
+
"\n",
|
60 |
+
"def write_path(rel_path):\n",
|
61 |
+
" return os.path.join(current_write_path, rel_path)\n",
|
62 |
+
"\n",
|
63 |
+
"if is_google_colab:\n",
|
64 |
+
" from google.colab import drive\n",
|
65 |
+
" drive.mount('/content/drive')\n",
|
66 |
+
" current_write_path = \"/content/drive/MyDrive/sql_gen\"\n",
|
67 |
+
"\n",
|
68 |
+
" hugging_face_path = snapshot_download(\n",
|
69 |
+
" repo_id=\"USC-Applied-NLP-Group/SQL-Generation\",\n",
|
70 |
+
" repo_type=\"model\", \n",
|
71 |
+
" allow_patterns=[\"src/*\", \"train-data/*\", \"deepseek-coder-1.3b-instruct/*\", \"nba-data/*\"], \n",
|
72 |
+
" )\n",
|
73 |
+
" sys.path.append(hugging_face_path)\n",
|
74 |
+
" current_path = hugging_face_path"
|
75 |
+
]
|
76 |
+
},
|
77 |
{
|
78 |
"cell_type": "markdown",
|
79 |
"metadata": {},
|
|
|
95 |
}
|
96 |
],
|
97 |
"source": [
|
98 |
+
"from src.prompts.prompt import input_text as input_prompt\n",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
"\n",
|
100 |
"print(len(input_prompt))"
|
101 |
]
|
|
|
109 |
},
|
110 |
{
|
111 |
"cell_type": "code",
|
112 |
+
"execution_count": null,
|
113 |
"metadata": {},
|
114 |
"outputs": [
|
115 |
{
|
116 |
"name": "stderr",
|
117 |
"output_type": "stream",
|
118 |
"text": [
|
119 |
+
"/var/folders/g0/47tr69v179dg7w6zyphp9b280000gn/T/ipykernel_35112/48906000.py:8: FutureWarning: DataFrame.applymap has been deprecated. Use DataFrame.map instead.\n",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
120 |
" df = df.applymap(lambda x: re.sub(r'\\s+', ' ', x) if isinstance(x, str) else x)\n"
|
121 |
]
|
122 |
},
|
|
|
147 |
"name": "stderr",
|
148 |
"output_type": "stream",
|
149 |
"text": [
|
150 |
+
"Map: 100%|ββββββββββ| 1044/1044 [00:17<00:00, 59.19 examples/s]"
|
151 |
]
|
152 |
},
|
153 |
{
|
|
|
168 |
}
|
169 |
],
|
170 |
"source": [
|
171 |
+
"\n",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
"\n",
|
173 |
"# Model output directories\n",
|
174 |
+
"MODEL_DIR = write_path(\"fine-tuned-model-16-test\")\n",
|
175 |
+
"VAL_OUTPUT = write_path(\"val-16.hf\")\n",
|
176 |
"\n",
|
177 |
"# Load dataset\n",
|
178 |
+
"df = pd.read_csv(read_path(\"train-data/sql_train.tsv\"), sep='\\t')\n",
|
179 |
"\n",
|
180 |
"df = df.applymap(lambda x: re.sub(r'\\s+', ' ', x) if isinstance(x, str) else x)\n",
|
181 |
"\n",
|
|
|
184 |
"print(df.head())\n",
|
185 |
"\n",
|
186 |
"# Load tokenizer\n",
|
187 |
+
"model_name = read_path(\"deepseek-coder-1.3b-instruct\")\n",
|
188 |
"tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
|
189 |
"\n",
|
190 |
"# Enable 8-bit quantization for lower memory usage\n",
|
191 |
+
"bnb_config = None\n",
|
192 |
+
"if use_bnb:\n",
|
193 |
+
" bnb_config = BitsAndBytesConfig(\n",
|
194 |
+
" load_in_8bit=True, \n",
|
195 |
+
" bnb_8bit_compute_dtype=torch.float16\n",
|
196 |
+
" )\n",
|
197 |
"\n",
|
198 |
"# Load model with quantization\n",
|
199 |
"#device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
|
|
|
737 |
},
|
738 |
{
|
739 |
"cell_type": "code",
|
740 |
+
"execution_count": null,
|
741 |
"metadata": {},
|
742 |
"outputs": [
|
743 |
{
|
|
|
760 |
"print(prompt_length)\n",
|
761 |
"\n",
|
762 |
"# Create connection to sqlite3 database\n",
|
763 |
+
"connection = sql.connect(read_path('nba-data/nba.sqlite'))\n",
|
764 |
"cursor = connection.cursor()\n",
|
765 |
"\n",
|
766 |
"for v in val_dataset:\n",
|
|
|
4115 |
],
|
4116 |
"metadata": {
|
4117 |
"kernelspec": {
|
4118 |
+
"display_name": "CSCI544",
|
4119 |
"language": "python",
|
4120 |
"name": "python3"
|
4121 |
},
|
|
|
4129 |
"name": "python",
|
4130 |
"nbconvert_exporter": "python",
|
4131 |
"pygments_lexer": "ipython3",
|
4132 |
+
"version": "3.11.11"
|
4133 |
}
|
4134 |
},
|
4135 |
"nbformat": 4,
|