Spaces:
Sleeping
Sleeping
PolPC13 commited on
Commit ·
3b0eab7
1
Parent(s): 134ea6c
Removed query image tool.
Browse files- tools/new_tools.py +1 -67
tools/new_tools.py
CHANGED
|
@@ -38,73 +38,7 @@ class WikiSearchTool(Tool):
|
|
| 38 |
return f"wiki_search failed {e}"
|
| 39 |
|
| 40 |
|
| 41 |
-
|
| 42 |
-
name = "query_image"
|
| 43 |
-
description = (
|
| 44 |
-
"Analyze an image with a VLM given a natural language query and an image URL. "
|
| 45 |
-
"Args: query (str), image_url (str). Returns: textual response."
|
| 46 |
-
)
|
| 47 |
-
inputs = {
|
| 48 |
-
"query": {
|
| 49 |
-
"type": "string",
|
| 50 |
-
"description": "Question or instruction about the image."
|
| 51 |
-
},
|
| 52 |
-
"image_url": {
|
| 53 |
-
"type": "string",
|
| 54 |
-
"description": "Public URL of the image to analyze."
|
| 55 |
-
},
|
| 56 |
-
}
|
| 57 |
-
output_type = "string"
|
| 58 |
-
|
| 59 |
-
def forward(self, query: str, image_url: str) -> str:
|
| 60 |
-
try:
|
| 61 |
-
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
| 62 |
-
"Qwen/Qwen2.5-VL-3B-Instruct", torch_dtype="auto", device_map="auto"
|
| 63 |
-
)
|
| 64 |
-
|
| 65 |
-
processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-3B-Instruct")
|
| 66 |
-
|
| 67 |
-
messages = [
|
| 68 |
-
{
|
| 69 |
-
"role": "user",
|
| 70 |
-
"content": [
|
| 71 |
-
{
|
| 72 |
-
"type": "image",
|
| 73 |
-
"image": image_url,
|
| 74 |
-
"max_pixels": 360 * 420,
|
| 75 |
-
},
|
| 76 |
-
{"type": "text", "text": query},
|
| 77 |
-
],
|
| 78 |
-
}
|
| 79 |
-
]
|
| 80 |
-
|
| 81 |
-
text = processor.apply_chat_template(
|
| 82 |
-
messages, tokenize=False, add_generation_prompt=True
|
| 83 |
-
)
|
| 84 |
-
image_inputs, video_inputs = process_vision_info(messages)
|
| 85 |
-
inputs = processor(
|
| 86 |
-
text=[text],
|
| 87 |
-
images=image_inputs,
|
| 88 |
-
videos=video_inputs,
|
| 89 |
-
padding=True,
|
| 90 |
-
return_tensors="pt",
|
| 91 |
-
)
|
| 92 |
-
inputs = inputs.to("cuda")
|
| 93 |
-
|
| 94 |
-
generated_ids = model.generate(**inputs, max_new_tokens=128)
|
| 95 |
-
generated_ids_trimmed = [
|
| 96 |
-
out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
|
| 97 |
-
]
|
| 98 |
-
output_text = processor.batch_decode(
|
| 99 |
-
generated_ids_trimmed,
|
| 100 |
-
skip_special_tokens=True,
|
| 101 |
-
clean_up_tokenization_spaces=False,
|
| 102 |
-
)
|
| 103 |
-
# original function returned a list; keep or cast to str as you prefer
|
| 104 |
-
return output_text[0] if isinstance(output_text, list) else output_text
|
| 105 |
-
except Exception as e:
|
| 106 |
-
return f"query_image failed {e}"
|
| 107 |
-
|
| 108 |
|
| 109 |
class ReverseStringTool(Tool):
|
| 110 |
name = "reverse_string"
|
|
|
|
| 38 |
return f"wiki_search failed {e}"
|
| 39 |
|
| 40 |
|
| 41 |
+
c
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
|
| 43 |
class ReverseStringTool(Tool):
|
| 44 |
name = "reverse_string"
|