From 527dc04e46350ba4ef57a0bddbe34763cbda831b Mon Sep 17 00:00:00 2001
From: Fanli Lin <fanli.lin@intel.com>
Date: Tue, 3 Dec 2024 03:40:20 +0800
Subject: [PATCH] [docs] add the missing import for Image and bug fix (#34776)

* add the missing import for Image lib

* add more devices in comment

* bug fix
---
 docs/source/en/tasks/video_text_to_text.md | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/docs/source/en/tasks/video_text_to_text.md b/docs/source/en/tasks/video_text_to_text.md
index fcc1c86e8b..3929f7994b 100644
--- a/docs/source/en/tasks/video_text_to_text.md
+++ b/docs/source/en/tasks/video_text_to_text.md
@@ -47,7 +47,7 @@ model_id = "llava-hf/llava-interleave-qwen-0.5b-hf"
 processor = LlavaProcessor.from_pretrained(model_id)
 
 model = LlavaForConditionalGeneration.from_pretrained(model_id, torch_dtype=torch.float16)
-model.to("cuda")
+model.to("cuda") # can also be xpu, mps, npu etc. depending on your hardware accelerator
 ```
 
 Some models directly consume the `<video>` token, and others accept `<image>` tokens equal to the number of sampled frames. This model handles videos in the latter fashion. We will write a simple utility to handle image tokens, and another utility to get a video from a url and sample frames from it. 
@@ -56,6 +56,7 @@ Some models directly consume the `<video>` token, and others accept `<image>` to
 import uuid
 import requests
 import cv2
+from PIL import Image
 
 def replace_video_with_images(text, frames):
   return text.replace("<video>", "<image>" * frames)
@@ -82,7 +83,7 @@ def sample_frames(url, num_frames):
         if i % interval == 0:
             frames.append(pil_img)
     video.release()
-    return frames
+    return frames[:num_frames]
 ```
 
 Let's get our inputs. We will sample frames and concatenate them.
@@ -127,7 +128,7 @@ This model has a prompt template that looks like following. First, we'll put all
 user_prompt = "Are these two cats in these two videos doing the same thing?"
 toks = "<image>" * 12
 prompt = "<|im_start|>user"+ toks + f"\n{user_prompt}<|im_end|><|im_start|>assistant"
-inputs = processor(prompt, images=videos).to(model.device, model.dtype)
+inputs = processor(text=prompt, images=videos, return_tensors="pt").to(model.device, model.dtype)
 ```
 
 We can now call [`~GenerationMixin.generate`] for inference. The model outputs the question in our input and answer, so we only take the text after the prompt and `assistant` part from the model output.