Spaces:
Running
on
T4
Running
on
T4
ffreemt
commited on
Commit
·
187485d
1
Parent(s):
ad51515
Update
Browse files
app.py
CHANGED
@@ -297,8 +297,11 @@ with gr.Blocks(
|
|
297 |
f"""
|
298 |
## {model_name.lower()}
|
299 |
|
300 |
-
* temperature range: .51 and up; higher temperature implies more randomness. Suggested temperature for chatting and creative writing is around 1.1 while it should be set to 0.51-1.0 for summarizing
|
301 |
-
|
|
|
|
|
|
|
302 |
```python
|
303 |
from gradio_client import Client
|
304 |
|
|
|
297 |
f"""
|
298 |
## {model_name.lower()}
|
299 |
|
300 |
+
* temperature range: .51 and up; higher temperature implies more randomness. Suggested temperature for chatting and creative writing is around 1.1 while it should be set to 0.51-1.0 for summarizing and translation.
|
301 |
+
* Set `repetition_penalty` to 2.1 or higher for a chatty conversation (more unpredictable and undesirable output). Lower it to 1.1 or smaller if more focused anwsers are desired (for example for translations or fact-oriented queries).
|
302 |
+
* Smaller `top_k` probably will result in smoothier sentences.
|
303 |
+
(`top_k=0` is equivalent to `top_k` equal to very very big though.) Consult `transformers` documentation for more details.
|
304 |
+
* An API is available at https://mikeee-qwen-7b-chat.hf.space/ that can be queried, e.g., in python
|
305 |
```python
|
306 |
from gradio_client import Client
|
307 |
|