Vblagoje
Rank:Average Model Cost: $0.0000
Number of Runs: 90,920
Models by this creator
bert-english-uncased-finetuned-pos
$-/run
64.2K
Huggingface
dpr-ctx_encoder-single-lfqa-wiki
$-/run
6.0K
Huggingface
dpr-question_encoder-single-lfqa-wiki
dpr-question_encoder-single-lfqa-wiki
The dpr-question_encoder-single-lfqa-wiki model is a question encoder based on the DPRQuestionEncoder architecture. It uses transformer's pooler outputs as question representations. The model was trained in two stages using the DPR-scale framework. In the first stage, it was fine-tuned on question-answer pairs from the LFQA dataset. In the second stage, a new training set was created using positives, negatives, and hard negatives from a Wikipedia/Faiss index instead of the LFQA dataset. The model's performance is slightly lower than the state-of-the-art REALM-based retriever on the KILT benchmark. The model was created by Vladimir Blagojevic.
$-/run
5.2K
Huggingface
dpr-question_encoder-single-lfqa-base
$-/run
516
Huggingface
bert-english-uncased-finetuned-chunk
$-/run
31
Huggingface
pipeline-generator
$-/run
20
Huggingface
pipeline-generator-v2
$-/run
19
Huggingface
retribert-base-uncased
$-/run
19
Huggingface
dpr-ctx_encoder-single-lfqa-base
$-/run
13
Huggingface