Update my_model/KBVQA.py
Browse files- my_model/KBVQA.py +4 -1
my_model/KBVQA.py
CHANGED
@@ -34,6 +34,7 @@ class KBVQA:
|
|
34 |
kbvqa_model (Optional[AutoModelForCausalLM]): The fine-tuned language model for KBVQA.
|
35 |
bnb_config (BitsAndBytesConfig): Configuration for BitsAndBytes optimized model.
|
36 |
access_token (str): Access token for Hugging Face API.
|
|
|
37 |
|
38 |
Methods:
|
39 |
create_bnb_config: Creates a BitsAndBytes configuration based on the quantization setting.
|
@@ -66,6 +67,7 @@ class KBVQA:
|
|
66 |
self.kbvqa_model: Optional[AutoModelForCausalLM] = None
|
67 |
self.bnb_config: BitsAndBytesConfig = self.create_bnb_config()
|
68 |
self.access_token: str = config.HUGGINGFACE_TOKEN
|
|
|
69 |
|
70 |
|
71 |
def create_bnb_config(self) -> BitsAndBytesConfig:
|
@@ -227,8 +229,9 @@ class KBVQA:
|
|
227 |
|
228 |
prompt = self.format_prompt(question, caption=caption, objects=detected_objects_str)
|
229 |
num_tokens = len(self.kbvqa_tokenizer.tokenize(prompt))
|
|
|
230 |
if num_tokens > self.max_context_window:
|
231 |
-
st.
|
232 |
return
|
233 |
|
234 |
model_inputs = self.kbvqa_tokenizer(prompt, add_special_tokens=False, return_tensors="pt").to('cuda')
|
|
|
34 |
kbvqa_model (Optional[AutoModelForCausalLM]): The fine-tuned language model for KBVQA.
|
35 |
bnb_config (BitsAndBytesConfig): Configuration for BitsAndBytes optimized model.
|
36 |
access_token (str): Access token for Hugging Face API.
|
37 |
+
current_prompt_length (int): Prompt length.
|
38 |
|
39 |
Methods:
|
40 |
create_bnb_config: Creates a BitsAndBytes configuration based on the quantization setting.
|
|
|
67 |
self.kbvqa_model: Optional[AutoModelForCausalLM] = None
|
68 |
self.bnb_config: BitsAndBytesConfig = self.create_bnb_config()
|
69 |
self.access_token: str = config.HUGGINGFACE_TOKEN
|
70 |
+
self.current_prompt_length = None
|
71 |
|
72 |
|
73 |
def create_bnb_config(self) -> BitsAndBytesConfig:
|
|
|
229 |
|
230 |
prompt = self.format_prompt(question, caption=caption, objects=detected_objects_str)
|
231 |
num_tokens = len(self.kbvqa_tokenizer.tokenize(prompt))
|
232 |
+
self.current_prompt_length = num_tokens
|
233 |
if num_tokens > self.max_context_window:
|
234 |
+
st.warning(f"Prompt too long with {num_tokens} tokens, consider increasing the confidence threshold for the object detector")
|
235 |
return
|
236 |
|
237 |
model_inputs = self.kbvqa_tokenizer(prompt, add_special_tokens=False, return_tensors="pt").to('cuda')
|