mlconvexai
commited on
Commit
•
7efd050
1
Parent(s):
0530606
Upload Poro_34B_GPTQ_quantization.ipynb
Browse files- Poro_34B_GPTQ_quantization.ipynb +736 -0
Poro_34B_GPTQ_quantization.ipynb
ADDED
@@ -0,0 +1,736 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "markdown",
|
5 |
+
"id": "c9399417-92ea-4474-a6cb-ce1ecf14f8ea",
|
6 |
+
"metadata": {},
|
7 |
+
"source": [
|
8 |
+
"# Poro 34B GPTQ quantization"
|
9 |
+
]
|
10 |
+
},
|
11 |
+
{
|
12 |
+
"cell_type": "markdown",
|
13 |
+
"id": "8bea76a0-0cce-461e-b167-2f1b6207395e",
|
14 |
+
"metadata": {},
|
15 |
+
"source": [
|
16 |
+
"## Step 1: Import transformers libraries and check the CUDA availability"
|
17 |
+
]
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"cell_type": "code",
|
21 |
+
"execution_count": 1,
|
22 |
+
"id": "1ca2fc08-52ed-4ca3-b849-fbcc72df11f6",
|
23 |
+
"metadata": {},
|
24 |
+
"outputs": [],
|
25 |
+
"source": [
|
26 |
+
"from transformers import AutoModelForCausalLM, AutoTokenizer, GPTQConfig"
|
27 |
+
]
|
28 |
+
},
|
29 |
+
{
|
30 |
+
"cell_type": "code",
|
31 |
+
"execution_count": 2,
|
32 |
+
"id": "97e1ee06-325a-4ca5-8426-39ee43fd02f1",
|
33 |
+
"metadata": {},
|
34 |
+
"outputs": [],
|
35 |
+
"source": [
|
36 |
+
"import torch"
|
37 |
+
]
|
38 |
+
},
|
39 |
+
{
|
40 |
+
"cell_type": "code",
|
41 |
+
"execution_count": 4,
|
42 |
+
"id": "17be0537-e39a-4ad7-b29b-6f4f7d72ead7",
|
43 |
+
"metadata": {},
|
44 |
+
"outputs": [
|
45 |
+
{
|
46 |
+
"data": {
|
47 |
+
"text/plain": [
|
48 |
+
"'2.2.1+cu121'"
|
49 |
+
]
|
50 |
+
},
|
51 |
+
"execution_count": 4,
|
52 |
+
"metadata": {},
|
53 |
+
"output_type": "execute_result"
|
54 |
+
}
|
55 |
+
],
|
56 |
+
"source": [
|
57 |
+
"torch.__version__"
|
58 |
+
]
|
59 |
+
},
|
60 |
+
{
|
61 |
+
"cell_type": "code",
|
62 |
+
"execution_count": 5,
|
63 |
+
"id": "e05ee325-ce5d-49b6-985e-c66ff88ee3e5",
|
64 |
+
"metadata": {},
|
65 |
+
"outputs": [
|
66 |
+
{
|
67 |
+
"data": {
|
68 |
+
"text/plain": [
|
69 |
+
"True"
|
70 |
+
]
|
71 |
+
},
|
72 |
+
"execution_count": 5,
|
73 |
+
"metadata": {},
|
74 |
+
"output_type": "execute_result"
|
75 |
+
}
|
76 |
+
],
|
77 |
+
"source": [
|
78 |
+
"torch.cuda.is_available()"
|
79 |
+
]
|
80 |
+
},
|
81 |
+
{
|
82 |
+
"cell_type": "code",
|
83 |
+
"execution_count": null,
|
84 |
+
"id": "c8114af7-2cdb-425f-ab8a-2d35462c2977",
|
85 |
+
"metadata": {},
|
86 |
+
"outputs": [],
|
87 |
+
"source": []
|
88 |
+
},
|
89 |
+
{
|
90 |
+
"cell_type": "markdown",
|
91 |
+
"id": "495fc0f8-ecc9-4c76-8251-2829246ee68a",
|
92 |
+
"metadata": {},
|
93 |
+
"source": [
|
94 |
+
"## Step 2: Load the original Poro 34B model from Huggingface and save it locally"
|
95 |
+
]
|
96 |
+
},
|
97 |
+
{
|
98 |
+
"cell_type": "code",
|
99 |
+
"execution_count": 3,
|
100 |
+
"id": "a5a24fba-71e7-4192-aafc-f95648b261d4",
|
101 |
+
"metadata": {},
|
102 |
+
"outputs": [],
|
103 |
+
"source": [
|
104 |
+
"model_name='LumiOpen/Poro-34B'"
|
105 |
+
]
|
106 |
+
},
|
107 |
+
{
|
108 |
+
"cell_type": "code",
|
109 |
+
"execution_count": 4,
|
110 |
+
"id": "148eeafd-6aae-440e-b30d-5ebdd1a8a4a5",
|
111 |
+
"metadata": {},
|
112 |
+
"outputs": [
|
113 |
+
{
|
114 |
+
"data": {
|
115 |
+
"application/vnd.jupyter.widget-view+json": {
|
116 |
+
"model_id": "c6a1ba90df9147489c1c4af10080d933",
|
117 |
+
"version_major": 2,
|
118 |
+
"version_minor": 0
|
119 |
+
},
|
120 |
+
"text/plain": [
|
121 |
+
"tokenizer_config.json: 0%| | 0.00/286 [00:00<?, ?B/s]"
|
122 |
+
]
|
123 |
+
},
|
124 |
+
"metadata": {},
|
125 |
+
"output_type": "display_data"
|
126 |
+
},
|
127 |
+
{
|
128 |
+
"data": {
|
129 |
+
"application/vnd.jupyter.widget-view+json": {
|
130 |
+
"model_id": "4c407d194d1742b091947f92ad455236",
|
131 |
+
"version_major": 2,
|
132 |
+
"version_minor": 0
|
133 |
+
},
|
134 |
+
"text/plain": [
|
135 |
+
"tokenizer.json: 0%| | 0.00/5.64M [00:00<?, ?B/s]"
|
136 |
+
]
|
137 |
+
},
|
138 |
+
"metadata": {},
|
139 |
+
"output_type": "display_data"
|
140 |
+
},
|
141 |
+
{
|
142 |
+
"data": {
|
143 |
+
"application/vnd.jupyter.widget-view+json": {
|
144 |
+
"model_id": "0392890e4392402086239952185801b1",
|
145 |
+
"version_major": 2,
|
146 |
+
"version_minor": 0
|
147 |
+
},
|
148 |
+
"text/plain": [
|
149 |
+
"special_tokens_map.json: 0%| | 0.00/545 [00:00<?, ?B/s]"
|
150 |
+
]
|
151 |
+
},
|
152 |
+
"metadata": {},
|
153 |
+
"output_type": "display_data"
|
154 |
+
}
|
155 |
+
],
|
156 |
+
"source": [
|
157 |
+
"org_tokenizer = AutoTokenizer.from_pretrained(model_name)"
|
158 |
+
]
|
159 |
+
},
|
160 |
+
{
|
161 |
+
"cell_type": "code",
|
162 |
+
"execution_count": null,
|
163 |
+
"id": "3d3ac738-3bc6-4c4f-bfc2-81692a80a662",
|
164 |
+
"metadata": {},
|
165 |
+
"outputs": [
|
166 |
+
{
|
167 |
+
"data": {
|
168 |
+
"application/vnd.jupyter.widget-view+json": {
|
169 |
+
"model_id": "14cac38e76464e748d261f4398b62085",
|
170 |
+
"version_major": 2,
|
171 |
+
"version_minor": 0
|
172 |
+
},
|
173 |
+
"text/plain": [
|
174 |
+
"config.json: 0%| | 0.00/697 [00:00<?, ?B/s]"
|
175 |
+
]
|
176 |
+
},
|
177 |
+
"metadata": {},
|
178 |
+
"output_type": "display_data"
|
179 |
+
},
|
180 |
+
{
|
181 |
+
"data": {
|
182 |
+
"application/vnd.jupyter.widget-view+json": {
|
183 |
+
"model_id": "383ffbe929054df2beb163e89af423eb",
|
184 |
+
"version_major": 2,
|
185 |
+
"version_minor": 0
|
186 |
+
},
|
187 |
+
"text/plain": [
|
188 |
+
"model.safetensors.index.json: 0%| | 0.00/57.0k [00:00<?, ?B/s]"
|
189 |
+
]
|
190 |
+
},
|
191 |
+
"metadata": {},
|
192 |
+
"output_type": "display_data"
|
193 |
+
},
|
194 |
+
{
|
195 |
+
"data": {
|
196 |
+
"application/vnd.jupyter.widget-view+json": {
|
197 |
+
"model_id": "22f290c533504c78891a85764d4a4dee",
|
198 |
+
"version_major": 2,
|
199 |
+
"version_minor": 0
|
200 |
+
},
|
201 |
+
"text/plain": [
|
202 |
+
"Downloading shards: 0%| | 0/14 [00:00<?, ?it/s]"
|
203 |
+
]
|
204 |
+
},
|
205 |
+
"metadata": {},
|
206 |
+
"output_type": "display_data"
|
207 |
+
},
|
208 |
+
{
|
209 |
+
"data": {
|
210 |
+
"application/vnd.jupyter.widget-view+json": {
|
211 |
+
"model_id": "a52ddab3f4c04bcd968d481e53d4ce83",
|
212 |
+
"version_major": 2,
|
213 |
+
"version_minor": 0
|
214 |
+
},
|
215 |
+
"text/plain": [
|
216 |
+
"model-00001-of-00014.safetensors: 0%| | 0.00/4.71G [00:00<?, ?B/s]"
|
217 |
+
]
|
218 |
+
},
|
219 |
+
"metadata": {},
|
220 |
+
"output_type": "display_data"
|
221 |
+
},
|
222 |
+
{
|
223 |
+
"data": {
|
224 |
+
"application/vnd.jupyter.widget-view+json": {
|
225 |
+
"model_id": "20afc515d7e64b0193d936d8fdb17a1e",
|
226 |
+
"version_major": 2,
|
227 |
+
"version_minor": 0
|
228 |
+
},
|
229 |
+
"text/plain": [
|
230 |
+
"model-00002-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
231 |
+
]
|
232 |
+
},
|
233 |
+
"metadata": {},
|
234 |
+
"output_type": "display_data"
|
235 |
+
},
|
236 |
+
{
|
237 |
+
"data": {
|
238 |
+
"application/vnd.jupyter.widget-view+json": {
|
239 |
+
"model_id": "431fbc52ce0a49219cb73664bfd1f9a6",
|
240 |
+
"version_major": 2,
|
241 |
+
"version_minor": 0
|
242 |
+
},
|
243 |
+
"text/plain": [
|
244 |
+
"model-00003-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
245 |
+
]
|
246 |
+
},
|
247 |
+
"metadata": {},
|
248 |
+
"output_type": "display_data"
|
249 |
+
},
|
250 |
+
{
|
251 |
+
"data": {
|
252 |
+
"application/vnd.jupyter.widget-view+json": {
|
253 |
+
"model_id": "fd7453a0ddd64b3299173e7d5586dab1",
|
254 |
+
"version_major": 2,
|
255 |
+
"version_minor": 0
|
256 |
+
},
|
257 |
+
"text/plain": [
|
258 |
+
"model-00004-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
259 |
+
]
|
260 |
+
},
|
261 |
+
"metadata": {},
|
262 |
+
"output_type": "display_data"
|
263 |
+
},
|
264 |
+
{
|
265 |
+
"data": {
|
266 |
+
"application/vnd.jupyter.widget-view+json": {
|
267 |
+
"model_id": "44114f0a82144b3ebdc5e1ae083bfbf7",
|
268 |
+
"version_major": 2,
|
269 |
+
"version_minor": 0
|
270 |
+
},
|
271 |
+
"text/plain": [
|
272 |
+
"model-00005-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
273 |
+
]
|
274 |
+
},
|
275 |
+
"metadata": {},
|
276 |
+
"output_type": "display_data"
|
277 |
+
},
|
278 |
+
{
|
279 |
+
"data": {
|
280 |
+
"application/vnd.jupyter.widget-view+json": {
|
281 |
+
"model_id": "f8052d0df811492eb6af45897fe568d2",
|
282 |
+
"version_major": 2,
|
283 |
+
"version_minor": 0
|
284 |
+
},
|
285 |
+
"text/plain": [
|
286 |
+
"model-00006-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
287 |
+
]
|
288 |
+
},
|
289 |
+
"metadata": {},
|
290 |
+
"output_type": "display_data"
|
291 |
+
},
|
292 |
+
{
|
293 |
+
"data": {
|
294 |
+
"application/vnd.jupyter.widget-view+json": {
|
295 |
+
"model_id": "787b3d7a208348f3ad0141b6c839faf1",
|
296 |
+
"version_major": 2,
|
297 |
+
"version_minor": 0
|
298 |
+
},
|
299 |
+
"text/plain": [
|
300 |
+
"model-00007-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
301 |
+
]
|
302 |
+
},
|
303 |
+
"metadata": {},
|
304 |
+
"output_type": "display_data"
|
305 |
+
},
|
306 |
+
{
|
307 |
+
"data": {
|
308 |
+
"application/vnd.jupyter.widget-view+json": {
|
309 |
+
"model_id": "acc82eeb37af48088a00b9e4537d4de6",
|
310 |
+
"version_major": 2,
|
311 |
+
"version_minor": 0
|
312 |
+
},
|
313 |
+
"text/plain": [
|
314 |
+
"model-00008-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
315 |
+
]
|
316 |
+
},
|
317 |
+
"metadata": {},
|
318 |
+
"output_type": "display_data"
|
319 |
+
},
|
320 |
+
{
|
321 |
+
"data": {
|
322 |
+
"application/vnd.jupyter.widget-view+json": {
|
323 |
+
"model_id": "588c9bec1e15411e929538a83a820356",
|
324 |
+
"version_major": 2,
|
325 |
+
"version_minor": 0
|
326 |
+
},
|
327 |
+
"text/plain": [
|
328 |
+
"model-00009-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
329 |
+
]
|
330 |
+
},
|
331 |
+
"metadata": {},
|
332 |
+
"output_type": "display_data"
|
333 |
+
},
|
334 |
+
{
|
335 |
+
"data": {
|
336 |
+
"application/vnd.jupyter.widget-view+json": {
|
337 |
+
"model_id": "70d44bda93724d62a069dc9c7d3abb02",
|
338 |
+
"version_major": 2,
|
339 |
+
"version_minor": 0
|
340 |
+
},
|
341 |
+
"text/plain": [
|
342 |
+
"model-00010-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
343 |
+
]
|
344 |
+
},
|
345 |
+
"metadata": {},
|
346 |
+
"output_type": "display_data"
|
347 |
+
},
|
348 |
+
{
|
349 |
+
"data": {
|
350 |
+
"application/vnd.jupyter.widget-view+json": {
|
351 |
+
"model_id": "4ef2e73b8204473b868043ce8d547148",
|
352 |
+
"version_major": 2,
|
353 |
+
"version_minor": 0
|
354 |
+
},
|
355 |
+
"text/plain": [
|
356 |
+
"model-00011-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
357 |
+
]
|
358 |
+
},
|
359 |
+
"metadata": {},
|
360 |
+
"output_type": "display_data"
|
361 |
+
},
|
362 |
+
{
|
363 |
+
"data": {
|
364 |
+
"application/vnd.jupyter.widget-view+json": {
|
365 |
+
"model_id": "3b463343388d42079f6f4452bb9931ce",
|
366 |
+
"version_major": 2,
|
367 |
+
"version_minor": 0
|
368 |
+
},
|
369 |
+
"text/plain": [
|
370 |
+
"model-00012-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
371 |
+
]
|
372 |
+
},
|
373 |
+
"metadata": {},
|
374 |
+
"output_type": "display_data"
|
375 |
+
},
|
376 |
+
{
|
377 |
+
"data": {
|
378 |
+
"application/vnd.jupyter.widget-view+json": {
|
379 |
+
"model_id": "1eb6c9827bdc481784d3069d04f7c318",
|
380 |
+
"version_major": 2,
|
381 |
+
"version_minor": 0
|
382 |
+
},
|
383 |
+
"text/plain": [
|
384 |
+
"model-00013-of-00014.safetensors: 0%| | 0.00/4.93G [00:00<?, ?B/s]"
|
385 |
+
]
|
386 |
+
},
|
387 |
+
"metadata": {},
|
388 |
+
"output_type": "display_data"
|
389 |
+
},
|
390 |
+
{
|
391 |
+
"data": {
|
392 |
+
"application/vnd.jupyter.widget-view+json": {
|
393 |
+
"model_id": "5662ef8d064b4fa9b70c9b1d1329fdd9",
|
394 |
+
"version_major": 2,
|
395 |
+
"version_minor": 0
|
396 |
+
},
|
397 |
+
"text/plain": [
|
398 |
+
"model-00014-of-00014.safetensors: 0%| | 0.00/4.52G [00:00<?, ?B/s]"
|
399 |
+
]
|
400 |
+
},
|
401 |
+
"metadata": {},
|
402 |
+
"output_type": "display_data"
|
403 |
+
},
|
404 |
+
{
|
405 |
+
"data": {
|
406 |
+
"application/vnd.jupyter.widget-view+json": {
|
407 |
+
"model_id": "7b7bd1bd16fb46068153e7b6b2f90b29",
|
408 |
+
"version_major": 2,
|
409 |
+
"version_minor": 0
|
410 |
+
},
|
411 |
+
"text/plain": [
|
412 |
+
"Loading checkpoint shards: 0%| | 0/14 [00:00<?, ?it/s]"
|
413 |
+
]
|
414 |
+
},
|
415 |
+
"metadata": {},
|
416 |
+
"output_type": "display_data"
|
417 |
+
}
|
418 |
+
],
|
419 |
+
"source": [
|
420 |
+
"branch = \"1000B\"\n",
|
421 |
+
"org_model = AutoModelForCausalLM.from_pretrained(model_name,\n",
|
422 |
+
" torch_dtype=torch.bfloat16,\n",
|
423 |
+
" revision=branch,\n",
|
424 |
+
")"
|
425 |
+
]
|
426 |
+
},
|
427 |
+
{
|
428 |
+
"cell_type": "code",
|
429 |
+
"execution_count": 6,
|
430 |
+
"id": "4edb55ba-908f-4070-8af3-92c7cf33f8d0",
|
431 |
+
"metadata": {},
|
432 |
+
"outputs": [],
|
433 |
+
"source": [
|
434 |
+
"model_configuration = org_model.config"
|
435 |
+
]
|
436 |
+
},
|
437 |
+
{
|
438 |
+
"cell_type": "code",
|
439 |
+
"execution_count": 7,
|
440 |
+
"id": "69107145-6808-4add-83fe-c3577893d724",
|
441 |
+
"metadata": {},
|
442 |
+
"outputs": [],
|
443 |
+
"source": [
|
444 |
+
"# original model configuration is missing the sequence length parameter\n",
|
445 |
+
"model_configuration.sequence_length = 2048"
|
446 |
+
]
|
447 |
+
},
|
448 |
+
{
|
449 |
+
"cell_type": "code",
|
450 |
+
"execution_count": 8,
|
451 |
+
"id": "33225c1e-6205-4ee0-95e9-2b15a2bf9a68",
|
452 |
+
"metadata": {},
|
453 |
+
"outputs": [
|
454 |
+
{
|
455 |
+
"data": {
|
456 |
+
"text/plain": [
|
457 |
+
"('Poro-34B/tokenizer_config.json',\n",
|
458 |
+
" 'Poro-34B/special_tokens_map.json',\n",
|
459 |
+
" 'Poro-34B/tokenizer.json')"
|
460 |
+
]
|
461 |
+
},
|
462 |
+
"execution_count": 8,
|
463 |
+
"metadata": {},
|
464 |
+
"output_type": "execute_result"
|
465 |
+
}
|
466 |
+
],
|
467 |
+
"source": [
|
468 |
+
"# Poro 34B is saved locally (this is not required but provides faster processing if there is a need for multiple runs)\n",
|
469 |
+
"org_model.save_pretrained(\"Poro-34B\", max_shard_size=\"5GB\",safe_serialization=True)\n",
|
470 |
+
"org_tokenizer.save_pretrained(\"Poro-34B\")"
|
471 |
+
]
|
472 |
+
},
|
473 |
+
{
|
474 |
+
"cell_type": "code",
|
475 |
+
"execution_count": null,
|
476 |
+
"id": "8d28eb15-26aa-4aed-be2b-2e67dd243e92",
|
477 |
+
"metadata": {},
|
478 |
+
"outputs": [],
|
479 |
+
"source": []
|
480 |
+
},
|
481 |
+
{
|
482 |
+
"cell_type": "markdown",
|
483 |
+
"id": "1e3ea881-658f-41fc-b09c-df711219653d",
|
484 |
+
"metadata": {},
|
485 |
+
"source": [
|
486 |
+
"## Step 3: Fine-tuned parameters are loaded from local Poro-34B-Lora-185 directory and merged"
|
487 |
+
]
|
488 |
+
},
|
489 |
+
{
|
490 |
+
"cell_type": "code",
|
491 |
+
"execution_count": 9,
|
492 |
+
"id": "c9cd3923-76ed-42d0-a882-5959cf0abf18",
|
493 |
+
"metadata": {},
|
494 |
+
"outputs": [],
|
495 |
+
"source": [
|
496 |
+
"from peft import PeftModel"
|
497 |
+
]
|
498 |
+
},
|
499 |
+
{
|
500 |
+
"cell_type": "code",
|
501 |
+
"execution_count": 10,
|
502 |
+
"id": "118992d6-336e-4eb2-9392-9ca77081ece7",
|
503 |
+
"metadata": {},
|
504 |
+
"outputs": [],
|
505 |
+
"source": [
|
506 |
+
"model_id2 = \"Poro-34B-Lora-185\""
|
507 |
+
]
|
508 |
+
},
|
509 |
+
{
|
510 |
+
"cell_type": "code",
|
511 |
+
"execution_count": 12,
|
512 |
+
"id": "10281b50-740f-4e2f-b2cc-f2d24a7e2f77",
|
513 |
+
"metadata": {},
|
514 |
+
"outputs": [],
|
515 |
+
"source": [
|
516 |
+
"loaded_model = PeftModel.from_pretrained(org_model,model_id2,is_trainable=True)"
|
517 |
+
]
|
518 |
+
},
|
519 |
+
{
|
520 |
+
"cell_type": "code",
|
521 |
+
"execution_count": 13,
|
522 |
+
"id": "c4ba0fc3-1cea-4e15-b302-d6353d1e970e",
|
523 |
+
"metadata": {},
|
524 |
+
"outputs": [],
|
525 |
+
"source": [
|
526 |
+
"# Fine-tuned weights are merged to original Poro 34B model\n",
|
527 |
+
"merged_model = loaded_model.merge_and_unload()"
|
528 |
+
]
|
529 |
+
},
|
530 |
+
{
|
531 |
+
"cell_type": "code",
|
532 |
+
"execution_count": 14,
|
533 |
+
"id": "07818cdc-d227-4abf-9437-8be3081aeb11",
|
534 |
+
"metadata": {},
|
535 |
+
"outputs": [
|
536 |
+
{
|
537 |
+
"data": {
|
538 |
+
"text/plain": [
|
539 |
+
"('Poro-34B-185c/tokenizer_config.json',\n",
|
540 |
+
" 'Poro-34B-185c/special_tokens_map.json',\n",
|
541 |
+
" 'Poro-34B-185c/tokenizer.json')"
|
542 |
+
]
|
543 |
+
},
|
544 |
+
"execution_count": 14,
|
545 |
+
"metadata": {},
|
546 |
+
"output_type": "execute_result"
|
547 |
+
}
|
548 |
+
],
|
549 |
+
"source": [
|
550 |
+
"# Merged model is saved locally\n",
|
551 |
+
"merged_model.save_pretrained(\"Poro-34B-185c\", max_shard_size=\"5GB\",safe_serialization=True)\n",
|
552 |
+
"org_tokenizer.save_pretrained(\"Poro-34B-185c\")"
|
553 |
+
]
|
554 |
+
},
|
555 |
+
{
|
556 |
+
"cell_type": "code",
|
557 |
+
"execution_count": null,
|
558 |
+
"id": "f18349b3-2fd8-4927-9ba4-a442d6217e1b",
|
559 |
+
"metadata": {},
|
560 |
+
"outputs": [],
|
561 |
+
"source": []
|
562 |
+
},
|
563 |
+
{
|
564 |
+
"cell_type": "markdown",
|
565 |
+
"id": "93d3bbdc-ceaf-4b19-b5e6-05f4d11cf275",
|
566 |
+
"metadata": {},
|
567 |
+
"source": [
|
568 |
+
"## Step 4: GPTQ quantization is applied to merged fine-tuned model"
|
569 |
+
]
|
570 |
+
},
|
571 |
+
{
|
572 |
+
"cell_type": "code",
|
573 |
+
"execution_count": 15,
|
574 |
+
"id": "e80c7627-da1f-4cb0-a079-6748806c0a0e",
|
575 |
+
"metadata": {},
|
576 |
+
"outputs": [],
|
577 |
+
"source": [
|
578 |
+
"model_id = \"Poro-34B-185c\""
|
579 |
+
]
|
580 |
+
},
|
581 |
+
{
|
582 |
+
"cell_type": "code",
|
583 |
+
"execution_count": 16,
|
584 |
+
"id": "2af3a06a-0d71-4c0e-aa3b-6f352c278bf2",
|
585 |
+
"metadata": {},
|
586 |
+
"outputs": [],
|
587 |
+
"source": [
|
588 |
+
"tokenizer = AutoTokenizer.from_pretrained(model_id)"
|
589 |
+
]
|
590 |
+
},
|
591 |
+
{
|
592 |
+
"cell_type": "code",
|
593 |
+
"execution_count": 17,
|
594 |
+
"id": "a302205a-0206-43b1-9bef-ed67547520f6",
|
595 |
+
"metadata": {},
|
596 |
+
"outputs": [],
|
597 |
+
"source": [
|
598 |
+
"# Dataset is a list of strings, we have here only one string to show the process\n",
|
599 |
+
"dataset = [\"Peruuta ensin vanhaan osoitteeseen tilattu uutiskirje kirjeen alareunan “Peruuta tilaus” -linkistä.\\nTilaa uutiskirje uudelleen oikeaan osoitteeseen.\"]"
|
600 |
+
]
|
601 |
+
},
|
602 |
+
{
|
603 |
+
"cell_type": "code",
|
604 |
+
"execution_count": 18,
|
605 |
+
"id": "2fabbc2e-436d-4ce1-ad89-4f30bdd977fa",
|
606 |
+
"metadata": {},
|
607 |
+
"outputs": [],
|
608 |
+
"source": [
|
609 |
+
"gptq_config = GPTQConfig(bits=4, dataset = dataset, tokenizer=tokenizer)"
|
610 |
+
]
|
611 |
+
},
|
612 |
+
{
|
613 |
+
"cell_type": "code",
|
614 |
+
"execution_count": null,
|
615 |
+
"id": "b6148609-3111-40a8-b748-fc876b9869f9",
|
616 |
+
"metadata": {},
|
617 |
+
"outputs": [],
|
618 |
+
"source": [
|
619 |
+
"model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=gptq_config,low_cpu_mem_usage=True)"
|
620 |
+
]
|
621 |
+
},
|
622 |
+
{
|
623 |
+
"cell_type": "code",
|
624 |
+
"execution_count": null,
|
625 |
+
"id": "59a2a544-9ee1-4f4e-b26a-d91de5e8f321",
|
626 |
+
"metadata": {},
|
627 |
+
"outputs": [],
|
628 |
+
"source": [
|
629 |
+
"# Quantized model and tokenizer are saved locally\n",
|
630 |
+
"model.save_pretrained(\"Poro-34B-GPTQ-SGroup\", use_safetensors=True)\n",
|
631 |
+
"tokenizer.save_pretrained(\"Poro-34B-GPTQ-SGroup\")"
|
632 |
+
]
|
633 |
+
},
|
634 |
+
{
|
635 |
+
"cell_type": "code",
|
636 |
+
"execution_count": null,
|
637 |
+
"id": "5e012ca3-966a-4480-aae3-6c2b67e6dde6",
|
638 |
+
"metadata": {},
|
639 |
+
"outputs": [],
|
640 |
+
"source": [
|
641 |
+
"# Login to Huggingface\n",
|
642 |
+
"from huggingface_hub import notebook_login\n",
|
643 |
+
"notebook_login()"
|
644 |
+
]
|
645 |
+
},
|
646 |
+
{
|
647 |
+
"cell_type": "code",
|
648 |
+
"execution_count": null,
|
649 |
+
"id": "587afbc0-8b81-4807-bed8-6af145845b95",
|
650 |
+
"metadata": {},
|
651 |
+
"outputs": [],
|
652 |
+
"source": [
|
653 |
+
"# Quantized model and tokenizer are saved to Huggingface\n",
|
654 |
+
"model.push_to_hub(\"Poro-34B-GPTQ-SGroup\", use_safetensors=True)\n",
|
655 |
+
"tokenizer.push_to_hub(\"Poro-34B-GPTQ-SGroup\")"
|
656 |
+
]
|
657 |
+
},
|
658 |
+
{
|
659 |
+
"cell_type": "code",
|
660 |
+
"execution_count": null,
|
661 |
+
"id": "f02d72bb-e75b-415f-b791-254246c5f971",
|
662 |
+
"metadata": {},
|
663 |
+
"outputs": [],
|
664 |
+
"source": []
|
665 |
+
},
|
666 |
+
{
|
667 |
+
"cell_type": "code",
|
668 |
+
"execution_count": null,
|
669 |
+
"id": "df85b2dc-22e0-40da-b1c1-fee4095c31be",
|
670 |
+
"metadata": {},
|
671 |
+
"outputs": [],
|
672 |
+
"source": []
|
673 |
+
},
|
674 |
+
{
|
675 |
+
"cell_type": "code",
|
676 |
+
"execution_count": null,
|
677 |
+
"id": "802bf734-e951-4aa0-9512-99ff7bf952f9",
|
678 |
+
"metadata": {},
|
679 |
+
"outputs": [],
|
680 |
+
"source": []
|
681 |
+
},
|
682 |
+
{
|
683 |
+
"cell_type": "code",
|
684 |
+
"execution_count": null,
|
685 |
+
"id": "f287018e-07b4-4080-b286-e905059f2f90",
|
686 |
+
"metadata": {},
|
687 |
+
"outputs": [],
|
688 |
+
"source": []
|
689 |
+
},
|
690 |
+
{
|
691 |
+
"cell_type": "code",
|
692 |
+
"execution_count": null,
|
693 |
+
"id": "6baf3c59-611b-47e0-9737-d25952c98c70",
|
694 |
+
"metadata": {},
|
695 |
+
"outputs": [],
|
696 |
+
"source": []
|
697 |
+
},
|
698 |
+
{
|
699 |
+
"cell_type": "code",
|
700 |
+
"execution_count": null,
|
701 |
+
"id": "84f6aef7-ff91-4bac-add3-3e3c6b4667ca",
|
702 |
+
"metadata": {},
|
703 |
+
"outputs": [],
|
704 |
+
"source": []
|
705 |
+
},
|
706 |
+
{
|
707 |
+
"cell_type": "code",
|
708 |
+
"execution_count": null,
|
709 |
+
"id": "8feaaca1-b05e-4e4e-97ce-18931d908eb7",
|
710 |
+
"metadata": {},
|
711 |
+
"outputs": [],
|
712 |
+
"source": []
|
713 |
+
}
|
714 |
+
],
|
715 |
+
"metadata": {
|
716 |
+
"kernelspec": {
|
717 |
+
"display_name": "Python 3",
|
718 |
+
"language": "python",
|
719 |
+
"name": "python3"
|
720 |
+
},
|
721 |
+
"language_info": {
|
722 |
+
"codemirror_mode": {
|
723 |
+
"name": "ipython",
|
724 |
+
"version": 3
|
725 |
+
},
|
726 |
+
"file_extension": ".py",
|
727 |
+
"mimetype": "text/x-python",
|
728 |
+
"name": "python",
|
729 |
+
"nbconvert_exporter": "python",
|
730 |
+
"pygments_lexer": "ipython3",
|
731 |
+
"version": "3.8.8"
|
732 |
+
}
|
733 |
+
},
|
734 |
+
"nbformat": 4,
|
735 |
+
"nbformat_minor": 5
|
736 |
+
}
|