Add relative path functionality for colab
Browse files- test_pretrained.ipynb +10 -7
test_pretrained.ipynb
CHANGED
@@ -35,17 +35,20 @@
|
|
35 |
},
|
36 |
{
|
37 |
"cell_type": "code",
|
38 |
-
"execution_count":
|
39 |
"metadata": {},
|
40 |
"outputs": [],
|
41 |
"source": [
|
42 |
"current_path = \"./\"\n",
|
43 |
"\n",
|
|
|
|
|
|
|
44 |
"if is_google_colab:\n",
|
45 |
" hugging_face_path = snapshot_download(\n",
|
46 |
" repo_id=\"USC-Applied-NLP-Group/SQL-Generation\",\n",
|
47 |
" repo_type=\"model\", \n",
|
48 |
-
" allow_patterns=[\"src/*\"],
|
49 |
" )\n",
|
50 |
" sys.path.append(hugging_face_path)\n",
|
51 |
" current_path = hugging_face_path"
|
@@ -70,7 +73,7 @@
|
|
70 |
},
|
71 |
{
|
72 |
"cell_type": "code",
|
73 |
-
"execution_count":
|
74 |
"metadata": {},
|
75 |
"outputs": [
|
76 |
{
|
@@ -88,7 +91,7 @@
|
|
88 |
],
|
89 |
"source": [
|
90 |
"# Load dataset and check length\n",
|
91 |
-
"df = pd.read_csv(
|
92 |
"print(\"Total dataset examples: \" + str(len(df)))\n",
|
93 |
"print(\"\\n\")\n",
|
94 |
"\n",
|
@@ -108,7 +111,7 @@
|
|
108 |
},
|
109 |
{
|
110 |
"cell_type": "code",
|
111 |
-
"execution_count":
|
112 |
"metadata": {},
|
113 |
"outputs": [],
|
114 |
"source": [
|
@@ -116,8 +119,8 @@
|
|
116 |
"device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
|
117 |
"\n",
|
118 |
"# Load model and tokenizer\n",
|
119 |
-
"tokenizer = AutoTokenizer.from_pretrained(\"
|
120 |
-
"model = AutoModelForCausalLM.from_pretrained(\"
|
121 |
"model.generation_config.pad_token_id = tokenizer.pad_token_id"
|
122 |
]
|
123 |
},
|
|
|
35 |
},
|
36 |
{
|
37 |
"cell_type": "code",
|
38 |
+
"execution_count": null,
|
39 |
"metadata": {},
|
40 |
"outputs": [],
|
41 |
"source": [
|
42 |
"current_path = \"./\"\n",
|
43 |
"\n",
|
44 |
+
"def get_path(rel_path):\n",
|
45 |
+
" return os.path.join(current_path, rel_path)\n",
|
46 |
+
"\n",
|
47 |
"if is_google_colab:\n",
|
48 |
" hugging_face_path = snapshot_download(\n",
|
49 |
" repo_id=\"USC-Applied-NLP-Group/SQL-Generation\",\n",
|
50 |
" repo_type=\"model\", \n",
|
51 |
+
" allow_patterns=[\"src/*\", \"train-data/*\", \"deepseek-coder-1.3b-instruct/*\"], \n",
|
52 |
" )\n",
|
53 |
" sys.path.append(hugging_face_path)\n",
|
54 |
" current_path = hugging_face_path"
|
|
|
73 |
},
|
74 |
{
|
75 |
"cell_type": "code",
|
76 |
+
"execution_count": null,
|
77 |
"metadata": {},
|
78 |
"outputs": [
|
79 |
{
|
|
|
91 |
],
|
92 |
"source": [
|
93 |
"# Load dataset and check length\n",
|
94 |
+
"df = pd.read_csv(get_path(\"train-data/sql_train.tsv\"), sep=\"\\t\")\n",
|
95 |
"print(\"Total dataset examples: \" + str(len(df)))\n",
|
96 |
"print(\"\\n\")\n",
|
97 |
"\n",
|
|
|
111 |
},
|
112 |
{
|
113 |
"cell_type": "code",
|
114 |
+
"execution_count": null,
|
115 |
"metadata": {},
|
116 |
"outputs": [],
|
117 |
"source": [
|
|
|
119 |
"device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
|
120 |
"\n",
|
121 |
"# Load model and tokenizer\n",
|
122 |
+
"tokenizer = AutoTokenizer.from_pretrained(get_path(\"deepseek-coder-1.3b-instruct\"))\n",
|
123 |
+
"model = AutoModelForCausalLM.from_pretrained(get_path(\"deepseek-coder-1.3b-instruct\"), torch_dtype=torch.bfloat16, device_map=device) \n",
|
124 |
"model.generation_config.pad_token_id = tokenizer.pad_token_id"
|
125 |
]
|
126 |
},
|