Update README.md
Browse files
README.md
CHANGED
|
@@ -48,6 +48,68 @@ colbert-ai==0.2.21 # for colbert retriever
|
|
| 48 |
flash-attn==2.7.4.post1 # for DSE retrievers to run with flash attention
|
| 49 |
```
|
| 50 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
|
| 52 |
## 💾Citation
|
| 53 |
If you use any datasets or models from this organization in your research, please cite our work as follows:
|
|
|
|
| 48 |
flash-attn==2.7.4.post1 # for DSE retrievers to run with flash attention
|
| 49 |
```
|
| 50 |
|
| 51 |
+
## How to use these checkpoints
|
| 52 |
+
|
| 53 |
+
We standardize codes for all retrievers in two python files
|
| 54 |
+
|
| 55 |
+
- **For text retrievers**: refer to [`text_wrapper.py`](https://huggingface.co/MMDocIR/MMDocIR_Retrievers/blob/main/text_wrapper.py)
|
| 56 |
+
- **For vision retrievers**: refer to [`vision_wrapper.py`](https://huggingface.co/MMDocIR/MMDocIR_Retrievers/blob/main/vision_wrapper.py)
|
| 57 |
+
|
| 58 |
+
If you want to encode [MMDocIR_Evaluation_Dataset](https://huggingface.co/datasets/MMDocIR/MMDocIR_Evaluation_Dataset) with these retrievers, you can refer to code [MMDocIR](https://github.com/MMDocRAG/MMDocIR/tree/main)/[encode.py](https://github.com/MMDocRAG/MMDocIR/blob/main/encode.py) and [inference command](https://github.com/MMDocRAG/MMDocIR?tab=readme-ov-file#3-inference-command).
|
| 59 |
+
|
| 60 |
+
If you want to encode your own queries/pages/layouts with these retrievers, some simple demo codes are:
|
| 61 |
+
|
| 62 |
+
- **For text retrievers**:
|
| 63 |
+
|
| 64 |
+
```python
|
| 65 |
+
From text_wrapper import DPR, BGE, GTE, E5, ColBERTReranker, Contriever
|
| 66 |
+
|
| 67 |
+
retriever = E5()
|
| 68 |
+
query = ['how much protein should a child consume', 'What is the CDC requirements for women?']
|
| 69 |
+
passage = [
|
| 70 |
+
"As a general guideline, the CDC's average requirement of protein for women ages 19 to 70 is 46 grams per day. But, as you can see from this chart, you'll need to increase that if you're expecting or training for a marathon. Check out the chart below to see how much protein you should be eating each day.",
|
| 71 |
+
"Definition of summit for English Language Learners. : 1 the highest point of a mountain : the top of a mountain. : 2 the highest level. : 3 a meeting or series of meetings between the leaders of two or more governments.",
|
| 72 |
+
"Definition of summit for English Language Learners. : 1 the highest point of a mountain : the top of a mountain. : 2 the highest level. : 3 a meeting or series of meetings between the leaders of two or more governments."
|
| 73 |
+
]
|
| 74 |
+
query_embeds = retriever.embed_queries(query)
|
| 75 |
+
passage_embeds = retriever.embed_quotes(passage)
|
| 76 |
+
scores = retriever.score(query, passage)
|
| 77 |
+
print(scores)
|
| 78 |
+
```
|
| 79 |
+
|
| 80 |
+
- **For image retrievers**:
|
| 81 |
+
|
| 82 |
+
```python
|
| 83 |
+
From vision_wrapper import DSE, ColQwen2Retriever, ColPaliRetriever
|
| 84 |
+
|
| 85 |
+
retriever = DSE(model_name="checkpoint/dse-phi3-v1", bs=2)
|
| 86 |
+
|
| 87 |
+
query = ['how much protein should a child consume', 'What is the CDC requirements for women?']
|
| 88 |
+
prefix = "/home/user/xxx"
|
| 89 |
+
images = [
|
| 90 |
+
"0704.0418_1.jpg",
|
| 91 |
+
"0704.0418_2.jpg",
|
| 92 |
+
"0704.0418_3.jpg",
|
| 93 |
+
"0705.1104_0.jpg",
|
| 94 |
+
"0705.1104_1.jpg",
|
| 95 |
+
"0705.1104_2.jpg",
|
| 96 |
+
"0704.0418_1.jpg",
|
| 97 |
+
"0704.0418_2.jpg",
|
| 98 |
+
"0704.0418_3.jpg",
|
| 99 |
+
"0705.1104_0.jpg",
|
| 100 |
+
"0705.1104_1.jpg",
|
| 101 |
+
"0705.1104_2.jpg",
|
| 102 |
+
]
|
| 103 |
+
images = [Image.open(prefix+x) for x in images]
|
| 104 |
+
q_embeds = retriever.embed_queries(queries)
|
| 105 |
+
img_embeds = retriever.embed_quotes(images)
|
| 106 |
+
|
| 107 |
+
scores = retriever.score(q_embeds, img_embeds)
|
| 108 |
+
print(scores)
|
| 109 |
+
```
|
| 110 |
+
|
| 111 |
+
|
| 112 |
+
|
| 113 |
|
| 114 |
## 💾Citation
|
| 115 |
If you use any datasets or models from this organization in your research, please cite our work as follows:
|