revert README.md

README.md

@@ -3692,11 +3692,16 @@ The `GME` models support three types of input: **text**, **image**, and **image-
 |[`gme-Qwen2-VL-7B`](https://huggingface.co/Alibaba-NLP/gme-Qwen2-VL-7B-Instruct) | 8.29B | 32768 | 3584 | 67.48 | 69.73 | 67.44 |
 
 ## Usage 
+**Use with sentence_transformers**
 
+The `encode` function accept `str` or `dict` with key(s) in `{'text', 'image', 'prompt'}`.
 
-**Transformers**
+**Do not pass `prompt` as the argument to `encode`**, pass as the input as a `dict` with a `prompt` key.
 
 ```python
+from sentence_transformers import SentenceTransformer
+
+
 t2i_prompt = 'Find an image that matches the given text.'
 texts = [
     "The Tesla Cybertruck is a battery electric pickup truck built by Tesla, Inc. since 2023.",
@@ -3708,41 +3713,33 @@ images = [
 ]
 
 
-gme = AutoModel.from_pretrained(
-    model_path,
-    torch_dtype="float16", device_map='cuda', trust_remote_code=True
-)
-
+gme_st = SentenceTransformer("Alibaba-NLP/gme-Qwen2-VL-2B-Instruct")
 
 # Single-modal embedding
-e_text = gme.get_text_embeddings(texts=texts)
-e_image = gme.get_image_embeddings(images=images)
+e_text = gme_st.encode(texts, convert_to_tensor=True)
+e_image = gme_st.encode([dict(image=i) for i in images], convert_to_tensor=True)
 print('Single-modal', (e_text @ e_image.T).tolist())
-## Single-modal [[0.359619140625, 0.0655517578125], [0.04180908203125, 0.374755859375]]
+## Single-modal [[0.356201171875, 0.06536865234375], [0.041717529296875, 0.37890625]]
 
 # How to set embedding instruction
-e_query = gme.get_text_embeddings(texts=texts, instruction=t2i_prompt)
-# If is_query=False, we always use the default instruction.
-e_corpus = gme.get_image_embeddings(images=images, is_query=False)
+e_query = gme_st.encode([dict(text=t, prompt=t2i_prompt) for t in texts], convert_to_tensor=True)
+# If no prompt, we always use the default instruction.
+e_corpus = gme_st.encode([dict(image=i) for i in images], convert_to_tensor=True)
 print('Single-modal with instruction', (e_query @ e_corpus.T).tolist())
-## Single-modal with instruction [[0.429931640625, 0.11505126953125], [0.049835205078125, 0.409423828125]]
+## Single-modal with instruction [[0.425537109375, 0.1158447265625], [0.049835205078125, 0.413818359375]]
 
 # Fused-modal embedding
-e_fused = gme.get_fused_embeddings(texts=texts, images=images)
+e_fused = gme_st.encode([dict(text=t, image=i) for t, i in zip(texts, images)], convert_to_tensor=True)
 print('Fused-modal', (e_fused @ e_fused.T).tolist())
-## Fused-modal [[1.0, 0.05511474609375], [0.05511474609375, 1.0]]
+## Fused-modal [[0.99951171875, 0.0556640625], [0.0556640625, 0.99951171875]]
 ```
 
 
-**sentence_transformers**
-
-The `encode` function accept `str` or `dict` with key(s) in `{'text', 'image', 'prompt'}`.
-
-**Do not pass `prompt` as the argument to `encode`**, pass as the input as a `dict` with a `prompt` key.
+**Use with custom code**
 
 ```python
-from sentence_transformers import SentenceTransformer
-
+# You can find the script gme_inference.py in https://huggingface.co/Alibaba-NLP/gme-Qwen2-VL-2B-Instruct/blob/main/gme_inference.py
+from gme_inference import GmeQwen2VL
 
 t2i_prompt = 'Find an image that matches the given text.'
 texts = [
@@ -3755,29 +3752,27 @@ images = [
 ]
 
 
-gme_st = SentenceTransformer("Alibaba-NLP/gme-Qwen2-VL-2B-Instruct")
+gme = GmeQwen2VL("Alibaba-NLP/gme-Qwen2-VL-2B-Instruct")
 
 # Single-modal embedding
-e_text = gme_st.encode(texts, convert_to_tensor=True)
-e_image = gme_st.encode([dict(image=i) for i in images], convert_to_tensor=True)
+e_text = gme.get_text_embeddings(texts=texts)
+e_image = gme.get_image_embeddings(images=images)
 print('Single-modal', (e_text @ e_image.T).tolist())
-## Single-modal [[0.356201171875, 0.06536865234375], [0.041717529296875, 0.37890625]]
+## [[0.359619140625, 0.0655517578125], [0.04180908203125, 0.374755859375]]
 
 # How to set embedding instruction
-e_query = gme_st.encode([dict(text=t, prompt=t2i_prompt) for t in texts], convert_to_tensor=True)
-# If no prompt, we always use the default instruction.
-e_corpus = gme_st.encode([dict(image=i) for i in images], convert_to_tensor=True)
+e_query = gme.get_text_embeddings(texts=texts, instruction=t2i_prompt)
+# If is_query=False, we always use the default instruction.
+e_corpus = gme.get_image_embeddings(images=images, is_query=False)
 print('Single-modal with instruction', (e_query @ e_corpus.T).tolist())
-## Single-modal with instruction [[0.425537109375, 0.1158447265625], [0.049835205078125, 0.413818359375]]
+## [[0.429931640625, 0.11505126953125], [0.049835205078125, 0.409423828125]]
 
 # Fused-modal embedding
-e_fused = gme_st.encode([dict(text=t, image=i) for t, i in zip(texts, images)], convert_to_tensor=True)
+e_fused = gme.get_fused_embeddings(texts=texts, images=images)
 print('Fused-modal', (e_fused @ e_fused.T).tolist())
-## Fused-modal [[0.99951171875, 0.0556640625], [0.0556640625, 0.99951171875]]
+## [[1.0, 0.05511474609375], [0.05511474609375, 1.0]]
 ```
 
-
-
 ## Evaluation
 
 We validated the performance on our universal multimodal retrieval benchmark (**UMRB**, see [Release UMRB](https://huggingface.co/Alibaba-NLP/gme-Qwen2-VL-7B-Instruct/discussions/2)) among others.