weifeng chen
commited on
Commit
•
56b40e2
1
Parent(s):
0e7276c
update
Browse files- README.md +7 -5
- pytorch_model.bin +1 -1
README.md
CHANGED
|
@@ -2,6 +2,7 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
# inference: false
|
| 4 |
# pipeline_tag: zero-shot-image-classification
|
|
|
|
| 5 |
|
| 6 |
# inference:
|
| 7 |
# parameters:
|
|
@@ -9,6 +10,7 @@ tags:
|
|
| 9 |
- clip
|
| 10 |
- zh
|
| 11 |
- image-text
|
|
|
|
| 12 |
---
|
| 13 |
|
| 14 |
# Model Details
|
|
@@ -31,8 +33,8 @@ from transformers import BertForSequenceClassification, BertConfig, BertTokenize
|
|
| 31 |
import numpy as np
|
| 32 |
|
| 33 |
# 加载TaiYi 中文 text encoder
|
| 34 |
-
text_tokenizer = BertTokenizer.from_pretrained("IDEA-CCNL/TaiYi-CLIP-Roberta-Chinese")
|
| 35 |
-
text_encoder = BertForSequenceClassification.from_pretrained("IDEA-CCNL/TaiYi-CLIP-Roberta-Chinese").eval()
|
| 36 |
text = text_tokenizer(["一只猫", "一只狗",'两只猫', '两只老虎','一只老虎'], return_tensors='pt', padding=True)['input_ids']
|
| 37 |
|
| 38 |
# 加载CLIP的image encoder
|
|
@@ -59,14 +61,14 @@ with torch.no_grad():
|
|
| 59 |
### Zero-Shot Classification
|
| 60 |
| model | dataset | Top1 | Top5 |
|
| 61 |
| ---- | ---- | ---- | ---- |
|
| 62 |
-
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese |
|
| 63 |
|
| 64 |
-
### Text-to-Image Retrieval
|
| 65 |
|
| 66 |
| model | dataset | Top1 | Top5 | Top10 |
|
| 67 |
| ---- | ---- | ---- | ---- | ---- |
|
| 68 |
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese | COCO-CN | 25.47 % | 51.70% | 63.07% |
|
| 69 |
-
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese | wukong50k |
|
| 70 |
|
| 71 |
|
| 72 |
# Citation
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
# inference: false
|
| 4 |
# pipeline_tag: zero-shot-image-classification
|
| 5 |
+
pipeline_tag: feature-extraction
|
| 6 |
|
| 7 |
# inference:
|
| 8 |
# parameters:
|
|
|
|
| 10 |
- clip
|
| 11 |
- zh
|
| 12 |
- image-text
|
| 13 |
+
- feature-extraction
|
| 14 |
---
|
| 15 |
|
| 16 |
# Model Details
|
|
|
|
| 33 |
import numpy as np
|
| 34 |
|
| 35 |
# 加载TaiYi 中文 text encoder
|
| 36 |
+
text_tokenizer = BertTokenizer.from_pretrained("IDEA-CCNL/TaiYi-CLIP-Roberta-102M-Chinese")
|
| 37 |
+
text_encoder = BertForSequenceClassification.from_pretrained("IDEA-CCNL/TaiYi-CLIP-Roberta-102M-Chinese").eval()
|
| 38 |
text = text_tokenizer(["一只猫", "一只狗",'两只猫', '两只老虎','一只老虎'], return_tensors='pt', padding=True)['input_ids']
|
| 39 |
|
| 40 |
# 加载CLIP的image encoder
|
|
|
|
| 61 |
### Zero-Shot Classification
|
| 62 |
| model | dataset | Top1 | Top5 |
|
| 63 |
| ---- | ---- | ---- | ---- |
|
| 64 |
+
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese | ImageNet1k-CN | 41.00% | 69.19% |
|
| 65 |
|
| 66 |
+
### Zero-Shot Text-to-Image Retrieval
|
| 67 |
|
| 68 |
| model | dataset | Top1 | Top5 | Top10 |
|
| 69 |
| ---- | ---- | ---- | ---- | ---- |
|
| 70 |
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese | COCO-CN | 25.47 % | 51.70% | 63.07% |
|
| 71 |
+
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese | wukong50k | 48.67 % | 81.77% | 90.09% |
|
| 72 |
|
| 73 |
|
| 74 |
# Citation
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 410713709
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53ec5505ee1ce25f970c5ce488bbd49b5727c36faa2132de0f2cf82dddbf3e37
|
| 3 |
size 410713709
|