Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
import os
|
2 |
import torch
|
3 |
-
from transformers import
|
4 |
-
from PIL import Image
|
5 |
import gradio as gr
|
|
|
6 |
import base64
|
7 |
import io
|
8 |
|
@@ -17,8 +17,8 @@ bnb_config = BitsAndBytesConfig(
|
|
17 |
bnb_4bit_compute_dtype=torch.float16
|
18 |
)
|
19 |
|
20 |
-
# Load model
|
21 |
-
model =
|
22 |
"ContactDoctor/Bio-Medical-MultiModal-Llama-3-8B-V1",
|
23 |
quantization_config=bnb_config,
|
24 |
device_map="auto",
|
@@ -37,38 +37,40 @@ tokenizer = AutoTokenizer.from_pretrained(
|
|
37 |
|
38 |
def analyze_input(image_data, question):
|
39 |
try:
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
if image_data is not None:
|
42 |
-
|
43 |
-
|
44 |
-
|
|
|
|
|
|
|
|
|
45 |
|
46 |
# Tokenize input
|
47 |
-
|
48 |
-
input_ids =
|
49 |
-
|
50 |
-
# Calculate target size (for generation length)
|
51 |
-
tgt_size = input_ids.size(1) + 256 # original length + max new tokens
|
52 |
-
|
53 |
-
# Prepare model inputs
|
54 |
-
model_inputs = {
|
55 |
-
"input_ids": input_ids,
|
56 |
-
"pixel_values": None, # Set to None for text-only queries
|
57 |
-
"tgt_sizes": [tgt_size] # Add target size for generation
|
58 |
-
}
|
59 |
-
|
60 |
# Generate response
|
61 |
outputs = model.generate(
|
62 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
)
|
64 |
-
|
65 |
-
# Decode
|
66 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
67 |
-
|
68 |
-
# Remove the prompt from the response
|
69 |
-
if prompt in response:
|
70 |
-
response = response[len(prompt):].strip()
|
71 |
-
|
72 |
return {
|
73 |
"status": "success",
|
74 |
"response": response
|
@@ -88,7 +90,7 @@ demo = gr.Interface(
|
|
88 |
],
|
89 |
outputs=gr.JSON(label="Analysis"),
|
90 |
title="Medical Query Analysis",
|
91 |
-
description="Ask medical questions. For now, please focus on text-based queries
|
92 |
flagging_mode="never"
|
93 |
)
|
94 |
|
@@ -97,4 +99,4 @@ demo.launch(
|
|
97 |
share=True,
|
98 |
server_name="0.0.0.0",
|
99 |
server_port=7860
|
100 |
-
)
|
|
|
1 |
import os
|
2 |
import torch
|
3 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
|
|
4 |
import gradio as gr
|
5 |
+
from PIL import Image
|
6 |
import base64
|
7 |
import io
|
8 |
|
|
|
17 |
bnb_4bit_compute_dtype=torch.float16
|
18 |
)
|
19 |
|
20 |
+
# Load model for causal language modeling
|
21 |
+
model = AutoModelForCausalLM.from_pretrained(
|
22 |
"ContactDoctor/Bio-Medical-MultiModal-Llama-3-8B-V1",
|
23 |
quantization_config=bnb_config,
|
24 |
device_map="auto",
|
|
|
37 |
|
38 |
def analyze_input(image_data, question):
|
39 |
try:
|
40 |
+
if not question.strip():
|
41 |
+
return {
|
42 |
+
"status": "error",
|
43 |
+
"message": "Question is required."
|
44 |
+
}
|
45 |
+
|
46 |
+
# Handle the input image (if any)
|
47 |
if image_data is not None:
|
48 |
+
return {
|
49 |
+
"status": "error",
|
50 |
+
"message": "Image support is not implemented yet."
|
51 |
+
}
|
52 |
+
|
53 |
+
# Prepare prompt for text-only input
|
54 |
+
prompt = f"Medical question: {question}\nAnswer: "
|
55 |
|
56 |
# Tokenize input
|
57 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
58 |
+
input_ids = inputs.input_ids.to(model.device)
|
59 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
60 |
# Generate response
|
61 |
outputs = model.generate(
|
62 |
+
input_ids=input_ids,
|
63 |
+
max_length=256, # Limit the length of the generated text
|
64 |
+
eos_token_id=tokenizer.eos_token_id, # Ensure generation stops correctly
|
65 |
+
pad_token_id=tokenizer.pad_token_id,
|
66 |
+
temperature=0.7, # Control randomness
|
67 |
+
top_p=0.9, # Nucleus sampling
|
68 |
+
top_k=50 # Top-k sampling
|
69 |
)
|
70 |
+
|
71 |
+
# Decode response
|
72 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
73 |
+
|
|
|
|
|
|
|
|
|
74 |
return {
|
75 |
"status": "success",
|
76 |
"response": response
|
|
|
90 |
],
|
91 |
outputs=gr.JSON(label="Analysis"),
|
92 |
title="Medical Query Analysis",
|
93 |
+
description="Ask medical questions. For now, please focus on text-based queries.",
|
94 |
flagging_mode="never"
|
95 |
)
|
96 |
|
|
|
99 |
share=True,
|
100 |
server_name="0.0.0.0",
|
101 |
server_port=7860
|
102 |
+
)
|