From eec76042f438226429d9f1b545a49a483ce33abb Mon Sep 17 00:00:00 2001
From: raghavanone <115454562+raghavanone@users.noreply.github.com>
Date: Tue, 28 Feb 2023 20:24:08 +0530
Subject: [PATCH] Fix the issue of blip model returning loss even when the
 label is not provided.  (#21811)

* Fix the issue of blip model returning loss even when the label is not provoided

* Fix ruff failure

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks
---
 src/transformers/models/blip/modeling_blip.py | 8 +-------
 tests/models/blip/test_modeling_blip.py       | 3 ++-
 2 files changed, 3 insertions(+), 8 deletions(-)

diff --git a/src/transformers/models/blip/modeling_blip.py b/src/transformers/models/blip/modeling_blip.py
index 7f1b3412b6..939bf26cc1 100644
--- a/src/transformers/models/blip/modeling_blip.py
+++ b/src/transformers/models/blip/modeling_blip.py
@@ -990,7 +990,7 @@ class BlipForConditionalGeneration(BlipPreTrainedModel):
 
         >>> outputs = model(**inputs)
         ```"""
-        batch_size = pixel_values.shape[0]
+
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         vision_outputs = self.vision_model(
@@ -1002,12 +1002,6 @@ class BlipForConditionalGeneration(BlipPreTrainedModel):
 
         image_embeds = vision_outputs[0]
 
-        if input_ids is None:
-            input_ids = torch.LongTensor([[self.decoder_input_ids] * batch_size]).to(image_embeds.device)
-
-        if labels is None:
-            labels = input_ids.masked_fill(input_ids == self.decoder_pad_token_id, -100)
-
         outputs = self.text_decoder(
             input_ids=input_ids,
             attention_mask=attention_mask,
diff --git a/tests/models/blip/test_modeling_blip.py b/tests/models/blip/test_modeling_blip.py
index 3a847aaff0..9a3dd0e96e 100644
--- a/tests/models/blip/test_modeling_blip.py
+++ b/tests/models/blip/test_modeling_blip.py
@@ -24,7 +24,7 @@ import numpy as np
 import requests
 
 from transformers import BlipConfig, BlipTextConfig, BlipVisionConfig
-from transformers.testing_utils import require_torch, require_vision, slow, torch_device
+from transformers.testing_utils import require_torch, require_torch_gpu, require_vision, slow, torch_device
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_configuration_common import ConfigTester
@@ -1111,6 +1111,7 @@ class BlipModelIntegrationTest(unittest.TestCase):
             [30522, 1037, 3861, 1997, 1037, 2450, 3564, 2006, 1996, 3509, 2007, 2014, 3899, 102],
         )
 
+    @require_torch_gpu
     def test_inference_image_captioning_fp16(self):
         model = BlipForConditionalGeneration.from_pretrained(
             "Salesforce/blip-image-captioning-base", torch_dtype=torch.float16