Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,47 +1,30 @@
|
|
1 |
import os
|
2 |
|
3 |
-
#
|
4 |
os.environ["HF_HUB_ENABLE_XET"] = "0"
|
5 |
-
|
6 |
-
# Use the robust Rust downloader for big files
|
7 |
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
|
8 |
-
|
9 |
-
# Optional but helpful: resume and avoid symlinks on some filesystems
|
10 |
os.environ["HF_HUB_ENABLE_RESUME"] = "1"
|
11 |
os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1"
|
12 |
|
13 |
-
|
14 |
import gradio as gr
|
15 |
import torch
|
16 |
from transformers import AutoProcessor, LlavaForConditionalGeneration
|
17 |
from PIL import Image
|
18 |
|
19 |
-
|
20 |
-
# Hugging Face model identifier. See the model card for more details:
|
21 |
-
# https://huggingface.co/StarCycle/llava-dinov2-internlm2-7b-v1
|
22 |
MODEL_ID = "xtuner/llava-phi-3-mini-hf"
|
23 |
|
24 |
-
#
|
25 |
-
# available we will use it and cast the weights to half precision to
|
26 |
-
# reduce memory consumption. Otherwise we fall back to CPU.
|
27 |
if torch.cuda.is_available():
|
28 |
-
DEVICE = torch.device("cuda")
|
29 |
TORCH_DTYPE = torch.float16
|
30 |
else:
|
31 |
-
DEVICE = torch.device("cpu")
|
32 |
TORCH_DTYPE = torch.float32
|
33 |
|
34 |
|
35 |
def load_model():
|
36 |
-
"""
|
37 |
-
|
38 |
-
The model is loaded with ``trust_remote_code=True`` to allow the
|
39 |
-
repository’s custom projector and adapter classes to be registered
|
40 |
-
correctly. We specify ``device_map='auto'`` so that the
|
41 |
-
``accelerate`` library will distribute the model across the
|
42 |
-
available hardware (GPU/CPU) automatically. The ``torch_dtype``
|
43 |
-
argument ensures that the model weights are loaded in half
|
44 |
-
precision on a GPU and in full precision on a CPU.
|
45 |
"""
|
46 |
model = LlavaForConditionalGeneration.from_pretrained(
|
47 |
MODEL_ID,
|
@@ -51,93 +34,99 @@ def load_model():
|
|
51 |
low_cpu_mem_usage=True,
|
52 |
)
|
53 |
processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
return model, processor
|
55 |
|
56 |
|
57 |
-
# Load
|
58 |
-
# we only do it once. If the model fails to load (for example
|
59 |
-
# because of missing dependencies) the exception will be raised here.
|
60 |
MODEL, PROCESSOR = load_model()
|
61 |
|
62 |
|
63 |
def answer_question(image: Image.Image, question: str) -> str:
|
64 |
-
"""Generate an answer for the given question about the uploaded image.
|
65 |
-
|
66 |
-
Parameters
|
67 |
-
----------
|
68 |
-
image: PIL.Image.Image
|
69 |
-
The user‑provided image. Gradio supplies images as PIL
|
70 |
-
objects, which the LLaVA processor accepts directly.
|
71 |
-
question: str
|
72 |
-
The user’s question about the image.
|
73 |
-
|
74 |
-
Returns
|
75 |
-
-------
|
76 |
-
str
|
77 |
-
The answer generated by the model. If either the image or
|
78 |
-
question is missing, an explanatory message is returned.
|
79 |
"""
|
80 |
-
|
|
|
81 |
if image is None:
|
82 |
return "Please upload an image."
|
83 |
if not question or not question.strip():
|
84 |
return "Please enter a question about the image."
|
85 |
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
-
#
|
101 |
-
|
102 |
-
|
103 |
-
with torch.
|
104 |
generated_ids = MODEL.generate(
|
105 |
**inputs,
|
106 |
max_new_tokens=256,
|
107 |
do_sample=False,
|
108 |
)
|
109 |
|
110 |
-
|
111 |
-
# the entire conversation (e.g., ``USER: ... ASSISTANT: ...``).
|
112 |
-
output = PROCESSOR.batch_decode(
|
113 |
generated_ids,
|
114 |
skip_special_tokens=True,
|
115 |
clean_up_tokenization_spaces=True,
|
116 |
)[0]
|
117 |
|
118 |
-
|
119 |
-
# ``ASSISTANT:`` delimiter.
|
120 |
-
if "ASSISTANT:" in output:
|
121 |
-
answer = output.split("ASSISTANT:")[-1].strip()
|
122 |
-
else:
|
123 |
-
# Fallback if the delimiter is not present.
|
124 |
-
answer = output.strip()
|
125 |
-
|
126 |
-
return answer
|
127 |
|
128 |
|
129 |
def build_interface() -> gr.Interface:
|
130 |
-
"""Construct the Gradio Interface object for the app."""
|
131 |
description = (
|
132 |
"Upload an image and ask a question about it.\n\n"
|
133 |
-
"This demo uses
|
134 |
-
"
|
135 |
-
"
|
136 |
-
"the InternLM2‑Chat‑7B language model via a lightweight projector and "
|
137 |
-
"LoRA adapters. Note: inference requires a GPU with sufficient "
|
138 |
-
"memory; on a CPU the generation will be extremely slow."
|
139 |
)
|
140 |
-
|
141 |
fn=answer_question,
|
142 |
inputs=[
|
143 |
gr.Image(type="pil", label="Image"),
|
@@ -148,22 +137,16 @@ def build_interface() -> gr.Interface:
|
|
148 |
),
|
149 |
],
|
150 |
outputs=gr.Textbox(label="Answer"),
|
151 |
-
title="Visual Question Answering
|
152 |
description=description,
|
153 |
flagging_mode="never",
|
154 |
)
|
155 |
-
return iface
|
156 |
|
157 |
|
158 |
def main() -> None:
|
159 |
-
"""Launch the Gradio app."""
|
160 |
iface = build_interface()
|
161 |
-
# When running on Hugging Face Spaces the app will automatically set
|
162 |
-
# the appropriate host and port. For local development you can
|
163 |
-
# uncomment the ``server_name`` argument to make the app reachable
|
164 |
-
# from other machines on your network.
|
165 |
iface.launch()
|
166 |
|
167 |
|
168 |
if __name__ == "__main__":
|
169 |
-
main()
|
|
|
1 |
import os
|
2 |
|
3 |
+
# ---- Hub download settings (apply before any HF imports) ----
|
4 |
os.environ["HF_HUB_ENABLE_XET"] = "0"
|
5 |
+
os.environ["HF_HUB_DISABLE_XET"] = "1"
|
|
|
6 |
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
|
|
|
|
|
7 |
os.environ["HF_HUB_ENABLE_RESUME"] = "1"
|
8 |
os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1"
|
9 |
|
|
|
10 |
import gradio as gr
|
11 |
import torch
|
12 |
from transformers import AutoProcessor, LlavaForConditionalGeneration
|
13 |
from PIL import Image
|
14 |
|
15 |
+
# Use the compact HF-format LLaVA model
|
|
|
|
|
16 |
MODEL_ID = "xtuner/llava-phi-3-mini-hf"
|
17 |
|
18 |
+
# Device + dtype
|
|
|
|
|
19 |
if torch.cuda.is_available():
|
|
|
20 |
TORCH_DTYPE = torch.float16
|
21 |
else:
|
|
|
22 |
TORCH_DTYPE = torch.float32
|
23 |
|
24 |
|
25 |
def load_model():
|
26 |
+
"""
|
27 |
+
Load the LLaVA model and its processor.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
"""
|
29 |
model = LlavaForConditionalGeneration.from_pretrained(
|
30 |
MODEL_ID,
|
|
|
34 |
low_cpu_mem_usage=True,
|
35 |
)
|
36 |
processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
|
37 |
+
|
38 |
+
# ---- Robustness: ensure processor carries vision attrs expected by LLaVA ----
|
39 |
+
vcfg = getattr(model.config, "vision_config", None)
|
40 |
+
|
41 |
+
if not hasattr(processor, "patch_size") or processor.patch_size is None:
|
42 |
+
# CLIP-L/336 typically uses patch_size=14; default to 14 if missing
|
43 |
+
processor.patch_size = getattr(vcfg, "patch_size", 14)
|
44 |
+
|
45 |
+
if (
|
46 |
+
not hasattr(processor, "vision_feature_select_strategy")
|
47 |
+
or processor.vision_feature_select_strategy is None
|
48 |
+
):
|
49 |
+
processor.vision_feature_select_strategy = getattr(
|
50 |
+
model.config, "vision_feature_select_strategy", "default"
|
51 |
+
)
|
52 |
+
|
53 |
+
if (
|
54 |
+
not hasattr(processor, "num_additional_image_tokens")
|
55 |
+
or processor.num_additional_image_tokens is None
|
56 |
+
):
|
57 |
+
# CLIP ViT uses a single CLS token
|
58 |
+
processor.num_additional_image_tokens = 1
|
59 |
+
|
60 |
return model, processor
|
61 |
|
62 |
|
63 |
+
# Load once at import
|
|
|
|
|
64 |
MODEL, PROCESSOR = load_model()
|
65 |
|
66 |
|
67 |
def answer_question(image: Image.Image, question: str) -> str:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
"""
|
69 |
+
Generate an answer about the uploaded image.
|
70 |
+
"""
|
71 |
if image is None:
|
72 |
return "Please upload an image."
|
73 |
if not question or not question.strip():
|
74 |
return "Please enter a question about the image."
|
75 |
|
76 |
+
try:
|
77 |
+
# ---- Preferred: chat-template path (handles image + text cleanly) ----
|
78 |
+
conversation = [{
|
79 |
+
"role": "user",
|
80 |
+
"content": [
|
81 |
+
{"type": "image"},
|
82 |
+
{"type": "text", "text": question.strip()},
|
83 |
+
],
|
84 |
+
}]
|
85 |
+
|
86 |
+
inputs = PROCESSOR.apply_chat_template(
|
87 |
+
conversation,
|
88 |
+
add_generation_prompt=True,
|
89 |
+
tokenize=True,
|
90 |
+
return_dict=True,
|
91 |
+
return_tensors="pt",
|
92 |
+
images=[image],
|
93 |
+
)
|
94 |
+
except Exception:
|
95 |
+
# ---- Fallback: legacy prompt with <image> placeholder ----
|
96 |
+
prompt = f"USER: <image>\n{question.strip()} ASSISTANT:"
|
97 |
+
inputs = PROCESSOR(
|
98 |
+
images=image,
|
99 |
+
text=prompt,
|
100 |
+
return_tensors="pt",
|
101 |
+
)
|
102 |
|
103 |
+
# Move all tensors to the model's device
|
104 |
+
inputs = {k: (v.to(MODEL.device) if hasattr(v, "to") else v) for k, v in inputs.items()}
|
105 |
+
|
106 |
+
with torch.inference_mode():
|
107 |
generated_ids = MODEL.generate(
|
108 |
**inputs,
|
109 |
max_new_tokens=256,
|
110 |
do_sample=False,
|
111 |
)
|
112 |
|
113 |
+
text = PROCESSOR.batch_decode(
|
|
|
|
|
114 |
generated_ids,
|
115 |
skip_special_tokens=True,
|
116 |
clean_up_tokenization_spaces=True,
|
117 |
)[0]
|
118 |
|
119 |
+
return text.strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
120 |
|
121 |
|
122 |
def build_interface() -> gr.Interface:
|
|
|
123 |
description = (
|
124 |
"Upload an image and ask a question about it.\n\n"
|
125 |
+
"This demo uses **xtuner/llava-phi-3-mini-hf** (LLaVA in HF format) "
|
126 |
+
"to perform visual question answering. Note: a GPU is recommended; "
|
127 |
+
"CPU inference will be slow."
|
|
|
|
|
|
|
128 |
)
|
129 |
+
return gr.Interface(
|
130 |
fn=answer_question,
|
131 |
inputs=[
|
132 |
gr.Image(type="pil", label="Image"),
|
|
|
137 |
),
|
138 |
],
|
139 |
outputs=gr.Textbox(label="Answer"),
|
140 |
+
title="Visual Question Answering (LLaVA Phi-3 Mini)",
|
141 |
description=description,
|
142 |
flagging_mode="never",
|
143 |
)
|
|
|
144 |
|
145 |
|
146 |
def main() -> None:
|
|
|
147 |
iface = build_interface()
|
|
|
|
|
|
|
|
|
148 |
iface.launch()
|
149 |
|
150 |
|
151 |
if __name__ == "__main__":
|
152 |
+
main()
|