Spaces:
Running
Running
vidhanm
commited on
Commit
·
a4644a0
1
Parent(s):
aecdd6d
nonetype error
Browse files
app.py
CHANGED
@@ -29,12 +29,10 @@ else:
|
|
29 |
print(f"Using device: {device}")
|
30 |
|
31 |
# --- Configuration for model components ---
|
32 |
-
# The main model ID for weights and overall config
|
33 |
model_id_for_weights = "lusxvr/nanoVLM-222M"
|
34 |
-
# The ID for the vision backbone's image processor configuration
|
35 |
image_processor_id = "openai/clip-vit-base-patch32"
|
36 |
-
#
|
37 |
-
tokenizer_id = "
|
38 |
|
39 |
image_processor = None
|
40 |
tokenizer = None
|
@@ -49,7 +47,7 @@ if VisionLanguageModel:
|
|
49 |
print(f"Attempting to load GPT2TokenizerFast from: {tokenizer_id}")
|
50 |
tokenizer = GPT2TokenizerFast.from_pretrained(tokenizer_id, trust_remote_code=True)
|
51 |
if tokenizer.pad_token is None:
|
52 |
-
tokenizer.pad_token = tokenizer.eos_token
|
53 |
print("Set tokenizer pad_token to eos_token.")
|
54 |
print("GPT2TokenizerFast loaded.")
|
55 |
|
@@ -64,7 +62,7 @@ if VisionLanguageModel:
|
|
64 |
except Exception as e:
|
65 |
print(f"Error loading model or processor components: {e}")
|
66 |
import traceback
|
67 |
-
traceback.print_exc()
|
68 |
image_processor = None
|
69 |
tokenizer = None
|
70 |
model = None
|
@@ -140,7 +138,6 @@ def generate_text_for_image(image_input, prompt_input):
|
|
140 |
|
141 |
description = "Interactive demo for lusxvr/nanoVLM-222M."
|
142 |
example_image_url = "http://images.cocodataset.org/val2017/000000039769.jpg"
|
143 |
-
# gradio_cache_dir = os.environ.get("GRADIO_TEMP_DIR", "/tmp/gradio_tmp") # Not used for now
|
144 |
|
145 |
iface = gr.Interface(
|
146 |
fn=generate_text_for_image,
|
@@ -155,8 +152,7 @@ iface = gr.Interface(
|
|
155 |
[example_image_url, "a photo of a"],
|
156 |
[example_image_url, "Describe the image in detail."],
|
157 |
],
|
158 |
-
cache_examples=True, #
|
159 |
-
# examples_cache_folder=gradio_cache_dir, # Removed due to potential Gradio version issue
|
160 |
allow_flagging="never"
|
161 |
)
|
162 |
|
|
|
29 |
print(f"Using device: {device}")
|
30 |
|
31 |
# --- Configuration for model components ---
|
|
|
32 |
model_id_for_weights = "lusxvr/nanoVLM-222M"
|
|
|
33 |
image_processor_id = "openai/clip-vit-base-patch32"
|
34 |
+
# Load the tokenizer from its original source to ensure all files are present
|
35 |
+
tokenizer_id = "gpt2" # Changed from "lusxvr/nanoVLM-222M"
|
36 |
|
37 |
image_processor = None
|
38 |
tokenizer = None
|
|
|
47 |
print(f"Attempting to load GPT2TokenizerFast from: {tokenizer_id}")
|
48 |
tokenizer = GPT2TokenizerFast.from_pretrained(tokenizer_id, trust_remote_code=True)
|
49 |
if tokenizer.pad_token is None:
|
50 |
+
tokenizer.pad_token = tokenizer.eos_token # Important for GPT-2
|
51 |
print("Set tokenizer pad_token to eos_token.")
|
52 |
print("GPT2TokenizerFast loaded.")
|
53 |
|
|
|
62 |
except Exception as e:
|
63 |
print(f"Error loading model or processor components: {e}")
|
64 |
import traceback
|
65 |
+
traceback.print_exc()
|
66 |
image_processor = None
|
67 |
tokenizer = None
|
68 |
model = None
|
|
|
138 |
|
139 |
description = "Interactive demo for lusxvr/nanoVLM-222M."
|
140 |
example_image_url = "http://images.cocodataset.org/val2017/000000039769.jpg"
|
|
|
141 |
|
142 |
iface = gr.Interface(
|
143 |
fn=generate_text_for_image,
|
|
|
152 |
[example_image_url, "a photo of a"],
|
153 |
[example_image_url, "Describe the image in detail."],
|
154 |
],
|
155 |
+
# cache_examples=True, # Temporarily commented out to ensure Gradio starts with minimal config
|
|
|
156 |
allow_flagging="never"
|
157 |
)
|
158 |
|