First commit and deploy
Browse files- app.py +55 -0
- requirements.txt +5 -0
app.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import transformers
|
2 |
+
import torch
|
3 |
+
import gc
|
4 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
5 |
+
import gradio as gr
|
6 |
+
|
7 |
+
|
8 |
+
# Load the model
|
9 |
+
model = "tiiuae/falcon-7b-instruct"
|
10 |
+
instruction = "Draft an apology email to a customer who experienced a delay in their order and provide reassurance that the issue has been resolved"
|
11 |
+
|
12 |
+
|
13 |
+
tokenizer = AutoTokenizer.from_pretrained(model)
|
14 |
+
pipeline = transformers.pipeline(
|
15 |
+
"text-generation",
|
16 |
+
model=model,
|
17 |
+
tokenizer=tokenizer,
|
18 |
+
torch_dtype=torch.bfloat16,
|
19 |
+
trust_remote_code=True,
|
20 |
+
device_map="auto",
|
21 |
+
)
|
22 |
+
|
23 |
+
def predict(instruction: str):
|
24 |
+
"""
|
25 |
+
The `predict` function takes an instruction as input and uses a pre-trained language model to
|
26 |
+
generate a predicted sequence of text based on the instruction.
|
27 |
+
|
28 |
+
:param instruction: The instruction parameter is a string that represents the input for which you
|
29 |
+
want to generate a prediction. It could be a question, a prompt, or any other kind of input that you
|
30 |
+
want the model to generate a response for
|
31 |
+
:type instruction: str
|
32 |
+
:return: The function `predict` returns a string that represents the generated text from the model.
|
33 |
+
"""
|
34 |
+
sequences = pipeline(
|
35 |
+
instruction,
|
36 |
+
max_length=500,
|
37 |
+
do_sample=True,
|
38 |
+
top_k=10,
|
39 |
+
num_return_sequences=1,
|
40 |
+
eos_token_id=tokenizer.eos_token_id,
|
41 |
+
)
|
42 |
+
for seq in sequences:
|
43 |
+
result = f"Result: {seq['generated_text']}"
|
44 |
+
gc.collect()
|
45 |
+
torch.cuda.empty_cache()
|
46 |
+
return result
|
47 |
+
|
48 |
+
gr.Interface(
|
49 |
+
predict,
|
50 |
+
inputs=[
|
51 |
+
gr.inputs.Textbox(lines=2, default=instruction, label="Instruction"),
|
52 |
+
],
|
53 |
+
outputs=[gr.outputs.Textbox(label="Output")],
|
54 |
+
title= "XGen"
|
55 |
+
).launch()
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
transformers
|
2 |
+
einops
|
3 |
+
accelerate
|
4 |
+
xformers
|
5 |
+
gradio
|