cicdatopea
commited on
update ipex result
Browse files
README.md
CHANGED
@@ -18,14 +18,15 @@ Please follow the license of the original model.
|
|
18 |
|
19 |
## How To Use
|
20 |
|
21 |
-
### INT4 Inference on CPU with Qbits
|
22 |
|
23 |
|
24 |
**pip3 install auto-round** (it will install intel-extension-for-pytorch and intel-extension-for-transformers both). For intel cpu, it will prioritize using intel-extension-for-pytorch , for other cpus, it will prioritize using intel-extension-for-transformers.
|
25 |
|
26 |
-
**To make sure to use qbits with intel-extension-for-transformers, please uninstall intel-extension-for-pytorch
|
27 |
-
|
28 |
|
|
|
|
|
29 |
|
30 |
~~~python
|
31 |
from auto_round import AutoRoundConfig ##must import for autoround format
|
@@ -161,7 +162,91 @@ prompt = "Please give a brief introduction of DeepSeek company."
|
|
161 |
"""DeepSeek Artificial Intelligence Co., Ltd. (referred to as "DeepSeek" or "深度求索") , founded in 2023, is a Chinese company dedicated to making AGI a reality"""
|
162 |
~~~
|
163 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
164 |
|
|
|
|
|
|
|
165 |
|
166 |
### INT4 Inference on CUDA(have not tested, maybe need 8X80G GPU)
|
167 |
|
|
|
18 |
|
19 |
## How To Use
|
20 |
|
21 |
+
### INT4 Inference on CPU with Qbits(Recommend)
|
22 |
|
23 |
|
24 |
**pip3 install auto-round** (it will install intel-extension-for-pytorch and intel-extension-for-transformers both). For intel cpu, it will prioritize using intel-extension-for-pytorch , for other cpus, it will prioritize using intel-extension-for-transformers.
|
25 |
|
26 |
+
**To make sure to use qbits with intel-extension-for-transformers, please uninstall intel-extension-for-pytorch**
|
|
|
27 |
|
28 |
+
intel-extension-for-transformers faster repacking, slower inference,higher accuracy
|
29 |
+
intel-extension-for-pytorch much slower repacking, faster inferecne, lower accuracy
|
30 |
|
31 |
~~~python
|
32 |
from auto_round import AutoRoundConfig ##must import for autoround format
|
|
|
162 |
"""DeepSeek Artificial Intelligence Co., Ltd. (referred to as "DeepSeek" or "深度求索") , founded in 2023, is a Chinese company dedicated to making AGI a reality"""
|
163 |
~~~
|
164 |
|
165 |
+
### INT4 Inference on CPU with IPEX
|
166 |
+
**pip3 install auto-round** (it will install intel-extension-for-pytorch and intel-extension-for-transformers both). For intel cpu, it will prioritize using intel-extension-for-pytorch , for other cpus, it will prioritize using intel-extension-for-transformers.
|
167 |
+
|
168 |
+
**To make sure to use intel-extension-for-pytorch, please uninstall intel-extension-for-transformers**
|
169 |
+
|
170 |
+
use the same code above
|
171 |
+
|
172 |
+
|
173 |
+
```python
|
174 |
+
Prompt: 9.11和9.8哪个数字大
|
175 |
+
Generated: 要比较 **9.11** 和 **9.8** 的大小,可以按照以下步骤进行:
|
176 |
+
|
177 |
+
1. **比较整数部分**:
|
178 |
+
- 两个数的整数部分都是 **9**,所以整数部分相同。
|
179 |
+
|
180 |
+
2. **比较小数部分**:
|
181 |
+
- **9.11** 的小数部分是 **0.11**
|
182 |
+
- **9.8** 的小数部分是 **0.8**
|
183 |
+
|
184 |
+
3. **统一小数位数**:
|
185 |
+
- 将 **0.8** 转换为 **0.80**,以便于比较。
|
186 |
+
|
187 |
+
4. **直接比较小数部分**:
|
188 |
+
- **0.80** > **0.11**
|
189 |
+
|
190 |
+
因此,**9.8** 大于 **9.11**。
|
191 |
+
|
192 |
+
最终答案:\boxed{9.8}
|
193 |
+
|
194 |
+
```
|
195 |
+
-------------------------------------
|
196 |
+
Prompt: strawberry中有几个r?
|
197 |
+
Generated: ### 第一步:理解问题
|
198 |
+
|
199 |
+
首先,我需要明确问题的含义。问题是:“strawberry中有几个r?”。这里的“strawberry”是一个英文单词,意思是“草莓”。问题问的是这个单
|
200 |
+
词中有多少个字母“r”。
|
201 |
+
|
202 |
+
### 第二步:分解单词
|
203 |
+
|
204 |
+
为了找出“strawberry”中有多少个“r”,我需要将这个单词分解成单个字母。让我们逐个字母来看:
|
205 |
+
|
206 |
+
- s
|
207 |
+
- t
|
208 |
+
- r
|
209 |
+
- a
|
210 |
+
- w
|
211 |
+
- b
|
212 |
+
- e
|
213 |
+
- r
|
214 |
+
- r
|
215 |
+
- y
|
216 |
+
|
217 |
+
### 第三步:识别字母“r”
|
218 |
+
|
219 |
+
现在,我需要找出这些字母中哪些是“r”。让我们逐一检查:
|
220 |
+
|
221 |
+
1. s - 不是r
|
222 |
+
2. t - 不是r
|
223 |
+
3. r - 是r
|
224 |
+
4. a - 不是r
|
225 |
+
5. w - 不是r
|
226 |
+
6. b
|
227 |
+
|
228 |
+
-------------------------------------
|
229 |
+
Prompt: How many r in strawberry.
|
230 |
+
Generated: The word "strawberry" contains **3 "r"s.
|
231 |
+
|
232 |
+
-------------------------------------
|
233 |
+
Prompt: There is a girl who likes adventure,
|
234 |
+
Generated: That sounds like the start of an exciting story! A girl who loves adventure could be the protagonist of countless thrilling tales. Here are a few ideas to spark your imagination:
|
235 |
+
|
236 |
+
1. **The Explorer of Lost Lands**: She discovers a hidden map leading to a forgotten civilization deep in the jungle. Along the way, she faces wild animals, solves ancient puzzles, and uncovers secrets about her own family.
|
237 |
+
|
238 |
+
2. **The Skybound Adventurer**: She builds or finds a mysterious airship and sets off to explore floating islands, sky cities, and uncharted clouds. Along the way, she encounters sky pirates, befriends mythical creatures, and learns to navigate the winds.
|
239 |
+
|
240 |
+
3. **The Time Traveler**: She stumbles upon a device that allows her to travel through time. She visits ancient civilizations, future worlds, and pivotal moments in history, all while trying to fix a timeline that’
|
241 |
+
|
242 |
+
|
243 |
+
-----------------------------------------
|
244 |
+
Prompt: Please give a brief introduction of DeepSeek company.
|
245 |
+
Generated: DeepSeek Artificial Intelligence Co., Ltd. (referred to as "DeepSeek" or "深度求索") , founded in 2023, is a Chinese company dedicated to making AGI a reality.
|
246 |
|
247 |
+
-----------------------------------------
|
248 |
+
Prompt: hello
|
249 |
+
Generated: Hello! How can I assist you today? 😊
|
250 |
|
251 |
### INT4 Inference on CUDA(have not tested, maybe need 8X80G GPU)
|
252 |
|