Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,48 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
---
|
4 |
+
|
5 |
+
This is a new kind of model optimization.
|
6 |
+
|
7 |
+
A paper is currently being written on the technique.
|
8 |
+
|
9 |
+
## Quickstart
|
10 |
+
|
11 |
+
This code snippets show how to get quickly started with running the model on a GPU:
|
12 |
+
|
13 |
+
```python
|
14 |
+
import torch
|
15 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
16 |
+
|
17 |
+
torch.random.manual_seed(0)
|
18 |
+
model_id = "microsoft/Phi-3-medium-4k-instruct"
|
19 |
+
model = AutoModelForCausalLM.from_pretrained(
|
20 |
+
model_id,
|
21 |
+
device_map="cuda",
|
22 |
+
torch_dtype="auto",
|
23 |
+
trust_remote_code=True,
|
24 |
+
)
|
25 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
26 |
+
|
27 |
+
messages = [
|
28 |
+
{"role": "user", "content": "Can you provide ways to eat combinations of bananas and dragonfruits?"},
|
29 |
+
{"role": "assistant", "content": "Sure! Here are some ways to eat bananas and dragonfruits together: 1. Banana and dragonfruit smoothie: Blend bananas and dragonfruits together with some milk and honey. 2. Banana and dragonfruit salad: Mix sliced bananas and dragonfruits together with some lemon juice and honey."},
|
30 |
+
{"role": "user", "content": "What about solving an 2x + 3 = 7 equation?"},
|
31 |
+
]
|
32 |
+
|
33 |
+
pipe = pipeline(
|
34 |
+
"text-generation",
|
35 |
+
model=model,
|
36 |
+
tokenizer=tokenizer,
|
37 |
+
)
|
38 |
+
|
39 |
+
generation_args = {
|
40 |
+
"max_new_tokens": 500,
|
41 |
+
"return_full_text": False,
|
42 |
+
"temperature": 0.0,
|
43 |
+
"do_sample": False,
|
44 |
+
}
|
45 |
+
|
46 |
+
output = pipe(messages, **generation_args)
|
47 |
+
print(output[0]['generated_text'])
|
48 |
+
```
|