Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import gradio as gr
|
3 |
+
import spaces
|
4 |
+
import torch
|
5 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
6 |
+
|
7 |
+
model_id = "futurehouse/ether0"
|
8 |
+
|
9 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
10 |
+
model = AutoModelForCausalLM.from_pretrained(
|
11 |
+
model_id,
|
12 |
+
device_map="auto",
|
13 |
+
torch_dtype=torch.float32
|
14 |
+
)
|
15 |
+
|
16 |
+
@spaces.GPU
|
17 |
+
def chat_fn(prompt, max_tokens=512):
|
18 |
+
max_tokens = min(int(max_tokens), 32_000)
|
19 |
+
messages = [{"role": "user", "content": prompt}]
|
20 |
+
chat_prompt = tokenizer.apply_chat_template(
|
21 |
+
messages, tokenize=False, add_generation_prompt=True
|
22 |
+
)
|
23 |
+
inputs = tokenizer(chat_prompt, return_tensors="pt").to(model.device)
|
24 |
+
|
25 |
+
# Generate with proper parameters
|
26 |
+
outputs = model.generate(
|
27 |
+
**inputs,
|
28 |
+
max_new_tokens=max_tokens,
|
29 |
+
do_sample=True,
|
30 |
+
temperature=0.1,
|
31 |
+
pad_token_id=tokenizer.eos_token_id
|
32 |
+
)
|
33 |
+
|
34 |
+
# Decode only the new tokens (not the input)
|
35 |
+
generated_text = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
|
36 |
+
return generated_text
|
37 |
+
|
38 |
+
gr.Interface(
|
39 |
+
fn=chat_fn,
|
40 |
+
inputs=[
|
41 |
+
gr.Textbox(label="prompt"),
|
42 |
+
gr.Number(label="max_tokens", value=512, precision=0)
|
43 |
+
],
|
44 |
+
outputs="text",
|
45 |
+
title="Ether0"
|
46 |
+
).launch(ssr_mode=False)
|