Spaces:
Paused
Paused
Commit
•
65fe553
1
Parent(s):
61d0f76
bug fix
Browse files
app.py
CHANGED
@@ -1,24 +1,13 @@
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
4 |
-
|
5 |
-
#model = AutoModelForCausalLM.from_pretrained("checkpoint_500",trust_remote_code=True)
|
6 |
-
|
7 |
-
model_name = "microsoft/phi-2"
|
8 |
-
|
9 |
import os
|
10 |
token = os.environ.get("HUGGING_FACE_TOKEN")
|
11 |
|
12 |
-
|
13 |
-
#bnb_config = BitsAndBytesConfig(
|
14 |
-
# load_in_4bit=True,
|
15 |
-
# bnb_4bit_quant_type="nf4",
|
16 |
-
# bnb_4bit_compute_dtype=torch.float16,
|
17 |
-
#)
|
18 |
|
19 |
model = AutoModelForCausalLM.from_pretrained(
|
20 |
model_name,
|
21 |
-
#quantization_config=bnb_config,
|
22 |
use_auth_token=token,
|
23 |
trust_remote_code=True
|
24 |
)
|
@@ -28,25 +17,28 @@ model.load_adapter("checkpoint_500")
|
|
28 |
tokenizer = AutoTokenizer.from_pretrained("checkpoint_500", trust_remote_code=True)
|
29 |
tokenizer.pad_token = tokenizer.eos_token
|
30 |
|
|
|
31 |
def inference(prompt, count):
|
32 |
count = int(count)
|
33 |
pipe = pipeline(task="text-generation", model=model, tokenizer=tokenizer)
|
34 |
result = pipe(f"{prompt}",max_new_tokens=count)
|
35 |
-
|
36 |
-
return
|
37 |
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
]
|
42 |
-
|
43 |
|
44 |
demo = gr.Interface(
|
45 |
inference,
|
46 |
-
inputs = [
|
47 |
-
|
48 |
-
|
49 |
-
|
|
|
|
|
|
|
50 |
examples = examples
|
51 |
)
|
|
|
52 |
demo.launch()
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
|
|
|
|
|
|
|
|
|
|
4 |
import os
|
5 |
token = os.environ.get("HUGGING_FACE_TOKEN")
|
6 |
|
7 |
+
model_name = "microsoft/phi-2"
|
|
|
|
|
|
|
|
|
|
|
8 |
|
9 |
model = AutoModelForCausalLM.from_pretrained(
|
10 |
model_name,
|
|
|
11 |
use_auth_token=token,
|
12 |
trust_remote_code=True
|
13 |
)
|
|
|
17 |
tokenizer = AutoTokenizer.from_pretrained("checkpoint_500", trust_remote_code=True)
|
18 |
tokenizer.pad_token = tokenizer.eos_token
|
19 |
|
20 |
+
|
21 |
def inference(prompt, count):
|
22 |
count = int(count)
|
23 |
pipe = pipeline(task="text-generation", model=model, tokenizer=tokenizer)
|
24 |
result = pipe(f"{prompt}",max_new_tokens=count)
|
25 |
+
output = result[0]['generated_text']
|
26 |
+
return output
|
27 |
|
28 |
+
examples = [
|
29 |
+
["What is LLM?","50"]
|
30 |
+
]
|
|
|
|
|
31 |
|
32 |
demo = gr.Interface(
|
33 |
inference,
|
34 |
+
inputs = [
|
35 |
+
gr.Textbox(placeholder="Enter a prompt"),
|
36 |
+
gr.Textbox(placeholder="Enter number of characters you want to generate")
|
37 |
+
],
|
38 |
+
outputs = [
|
39 |
+
gr.Textbox(label="Generated text")
|
40 |
+
],
|
41 |
examples = examples
|
42 |
)
|
43 |
+
|
44 |
demo.launch()
|