Make min and max configurable on VQA

models/blip_vqa.py

@@ -34,7 +34,7 @@ class BLIP_VQA(nn.Module):
         self.text_decoder = BertLMHeadModel(config=decoder_config)          
 
 
-    def forward(self, image, question, answer=None, n=None, weights=None, train=True, inference='rank', k_test=128):
+    def forward(self, image, question, answer=None, n=None, weights=None, train=True, maxa_len=10, mina_len=1, inference='rank', k_test=128):
         
         image_embeds = self.visual_encoder(image) 
         image_atts = torch.ones(image_embeds.size()[:-1],dtype=torch.long).to(image.device)
@@ -97,8 +97,8 @@ class BLIP_VQA(nn.Module):
                 bos_ids = torch.full((image.size(0),1),fill_value=self.tokenizer.bos_token_id,device=image.device)
                 
                 outputs = self.text_decoder.generate(input_ids=bos_ids,
-                                                     max_length=10,
-                                                     min_length=1,
+                                                     max_length=maxa_len,
+                                                     min_length=mina_len,
                                                      num_beams=num_beams,
                                                      eos_token_id=self.tokenizer.sep_token_id,
                                                      pad_token_id=self.tokenizer.pad_token_id,