vblagoje commited on
Commit
57dcaa1
1 Parent(s): 34cbc85

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +29 -0
README.md ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: en
3
+ datasets:
4
+ - vblagoje/lfqa
5
+ license: mit
6
+ ---
7
+
8
+ ## Introduction
9
+ The context/passage encoder model based on [DPRContextEncoder](https://huggingface.co/docs/transformers/master/en/model_doc/dpr#transformers.DPRContextEncoder) architecture. It uses the transformer's pooler outputs as context/passage representations.
10
+
11
+ ## Training
12
+ We trained vblagoje/dpr-ctx_encoder-single-lfqa-base using FAIR's dpr-scale starting with PAQ based pretrained checkpoint and fine-tuned the retriever on the question-answer pairs from the LFQA dataset. As dpr-scale requires DPR formatted training set input with positive, negative, and hard negative samples - we created a training file with an answer being positive, negatives being question unrelated answers, while hard negative samples were chosen from answers on questions between 0.55 and 0.65 of cosine similarity.
13
+
14
+ ## Performance
15
+ LFQA DPR-based retriever (vblagoje/dpr-question_encoder-single-lfqa-base and vblagoje/dpr-ctx_encoder-single-lfqa-base) had a score of 6.69 for R-precision and 14.5 for Recall@5 on KILT benchmark.
16
+
17
+ ## Usage
18
+
19
+
20
+ ```python
21
+ from transformers import DPRContextEncoder, DPRContextEncoderTokenizer
22
+ model = DPRQuestionEncoder.from_pretrained("vblagoje/dpr-question_encoder-single-lfqa-base").to(device)
23
+ tokenizer = AutoTokenizer.from_pretrained("vblagoje/dpr-question_encoder-single-lfqa-base")
24
+ input_ids = tokenizer("Why do airplanes leave contrails in the sky?", return_tensors="pt")["input_ids"]
25
+ embeddings = model(input_ids).pooler_output
26
+ ```
27
+
28
+ ## Author
29
+ - Vladimir Blagojevic: `dovlex [at] gmail.com` [Twitter](https://twitter.com/vladblagoje) | [LinkedIn](https://www.linkedin.com/in/blagojevicvladimir/)