From 0c9346be5820adacff269834caffb2573756b5ec Mon Sep 17 00:00:00 2001
From: Jael Gu <mengjia.gu@zilliz.com>
Date: Tue, 19 Jul 2022 11:32:25 +0800
Subject: [PATCH] Add README for towhee.models.clip (#1582)

Signed-off-by: Jael Gu <mengjia.gu@zilliz.com>
---
 towhee/models/clip/README.md | 63 +++++++++++++++++++++++++++++++++++-
 1 file changed, 62 insertions(+), 1 deletion(-)

diff --git a/towhee/models/clip/README.md b/towhee/models/clip/README.md
index beb6eb315b..4e5bae3b28 100644
--- a/towhee/models/clip/README.md
+++ b/towhee/models/clip/README.md
@@ -1,3 +1,64 @@
 # CLIP
 
-The original codes, weights, and other files for CLIP are from the [official implementation](https://github.com/openai/CLIP).
+CLIP in Towhee is built on top of the [official implementation](https://github.com/openai/CLIP).
+
+Available model names:
+- clip_vit_b16
+- clip_vit_b32 (support multilingual)
+- clip_resnet_r50
+- clip_resnet_r101
+
+## Code Example
+
+- Create model
+```python
+from towhee.models import clip
+
+# Create CLIP model with parameters
+model = clip.create_model(
+    embed_dim=512, image_resolution=4,
+    vision_layers=12, vision_width=768, vision_patch_size=2,
+    context_length=77, vocab_size=49408, transformer_width=512,
+    transformer_heads=8, transformer_layers=12
+    )
+
+# Create CLIP model with model name (no-pretrain)
+model = clip.create_model(model_name='clip_vit_b32', pretrained=False)
+
+# Load pretrained model with model name
+model = clip.create_model(model_name='clip_vit_b32', pretrained=True)
+```
+
+- Encode image
+```python
+import torch
+
+dummy_img = torch.rand(1, 3, 224, 224)
+img_features = model.encode_image(dummy_img)
+```
+
+- Encode text
+```python
+# Tokenized input
+text = torch.randint(high=49408, size=(1, 77), dtype=torch.int32)
+text_features = model.encode_text(text)
+
+# String input
+text = ['test']
+text_features = model.encode_text(text)
+
+# Multilingual only for supported models
+text_chinese = ['测试']
+text_features = model.encode_text(text, multilingual=True)
+```
+
+- Calculate similarities
+```python
+img = torch.rand(1, 3, 224, 224)
+text = ['test']
+logits_per_img, logits_per_text = model(img, text)
+
+# Multilingual only for supported models
+text = ['测试']
+logits_per_img, logits_per_text = model(img, text, multilingual=True)
+```
\ No newline at end of file