Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
|
3 |
+
|
4 |
+
# BEE-spoke-data/Mixtral-GQA-400m-v2
|
5 |
+
|
6 |
+
|
7 |
+
|
8 |
+
|
9 |
+
## testing code
|
10 |
+
|
11 |
+
```python
|
12 |
+
|
13 |
+
# !pip install -U -q transformers datasets accelerate sentencepiece
|
14 |
+
from huggingface_hub import notebook_login
|
15 |
+
|
16 |
+
notebook_login()
|
17 |
+
|
18 |
+
import warnings
|
19 |
+
|
20 |
+
warnings.filterwarnings("ignore")
|
21 |
+
|
22 |
+
# Use a pipeline as a high-level helper
|
23 |
+
from transformers import pipeline
|
24 |
+
|
25 |
+
pipe = pipeline("text-generation", model="BEE-spoke-data/Mixtral-GQA-400m-v2")
|
26 |
+
pipe.model.config.pad_token_id = pipe.model.config.eos_token_id
|
27 |
+
|
28 |
+
import pprint as pp
|
29 |
+
|
30 |
+
prompt = "My favorite is Tori Black because"
|
31 |
+
|
32 |
+
res = pipe(
|
33 |
+
prompt,
|
34 |
+
max_new_tokens=256,
|
35 |
+
top_k=4,
|
36 |
+
penalty_alpha=0.6,
|
37 |
+
use_cache=True,
|
38 |
+
no_repeat_ngram_size=4,
|
39 |
+
repetition_penalty=1.1,
|
40 |
+
renormalize_logits=True,
|
41 |
+
)
|
42 |
+
pp.pprint(res[0])
|
43 |
+
|
44 |
+
```
|