Giter Club home page Giter Club logo

magic's Introduction

Hi there 👋

I am a final-year Ph.D. student at Language Technology Lab, University of Cambridge. I am broadly interested in natural language processing (NLP) and machine learning. The majority of my research lies in text generation. Recently, I focus my research on the topic of contrastive learning and the study of its potential in language model pre-training, discourse representation learning, knowledge probing, open-ended text generation, and multi-modal text generation. Please refer to [my personal page] for the complete list of my research.

Personally, I really like pandas. The one in my icon is my favourite and her name is Hehua.

Yixuan's GitHub stats

magic's People

Contributors

gmftbygmftby avatar yhlleo avatar yxuansu avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

magic's Issues

about the performance of the Zerocap demonstrated in this paper

Thanks for your amazing work on the zero-shot captioning task. As shown in Table 1 of this paper, the zerocap's performance on COCO is as follows:

however, it seems different from the performance reported in Zerocap's paper and is shown as follows:

In this paper, did the Zerocap use different settings that resulted in this difference?
I would greatly appreciate your response.
Thank you.

is there anybody reproduce this methods with other LM/VLM?

Thanks to share this great work :)

I have tried to reproduce this methods with my own LM(gpt-3)/VLM(CLIP). However, the quality is significantly inferior to the example you provided.
스크린샷 2022-10-12 오전 9 11 19

is there anybody reproduce this methods with your own LM/VLM ?
or
is there any implementation detail when I use my own LM/VLM ?

Evaluation code for story generation

Hi, thanks for your inspiring work.

I'm wondering if the evaluation code for story generation is available to reporduce the result reported in main paper Table 3.

教程

感谢 博主开源 请问下 有视频安装教程吗 打扰了

Why is the image captioning score on MS-COCO not measured on full validation set, but only on a subset?

About the image captioning results on MS-COCO, I could reproduce exactly the same scores as presented in Table1 of the paper, by using the following result file. (https://github.com/yxuansu/MAGIC/blob/main/image_captioning/inference_result/mscoco/magic_result.json)
However, this file only contains results of 4982 images, while the number of full validation images is about 40k.
Why is the image captioning score on MS-COCO not measured on full validation set, but only on a subset?

Java ClassNotFoundException raised

Hi, I tried to evaluate result, ClassNotFoundException error raised.
How can I add SemgrexPattern class?

(magic) teang1995@devbox:~/codes/MAGIC/image_captioning/evaluation$ python cocoeval.py --result_file_path ../inference_result/flickr30k/baselines/contrastive_result.json
tokenization...
PTBTokenizer tokenized 72436 tokens at 390823.69 tokens per second.
PTBTokenizer tokenized 14999 tokens at 142902.49 tokens per second.
setting up scorers...
computing Bleu score...
{'testlen': 13000, 'reflen': 12470, 'guess': [13000, 12000, 11000, 10000], 'correct': [6192, 2110, 773, 341]}
ratio: 1.0425020048114642
Bleu_1: 0.476
Bleu_2: 0.289
Bleu_3: 0.181
Bleu_4: 0.119
computing METEOR score...
METEOR: 0.127
computing Rouge score...
ROUGE_L: 0.353
computing CIDEr score...
CIDEr: 0.089
computing SPICE score...
Exception in thread "main" java.lang.NoClassDefFoundError: edu/stanford/nlp/semgraph/semgrex/SemgrexPattern
        at edu.anu.spice.SpiceParser.<clinit>(SpiceParser.java:64)
        at edu.anu.spice.SpiceScorer.scoreBatch(SpiceScorer.java:70)
        at edu.anu.spice.SpiceScorer.main(SpiceScorer.java:60)
Caused by: java.lang.ClassNotFoundException: edu.stanford.nlp.semgraph.semgrex.SemgrexPattern
        at java.net.URLClassLoader.findClass(URLClassLoader.java:387)
        at java.lang.ClassLoader.loadClass(ClassLoader.java:418)
        at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:352)
        at java.lang.ClassLoader.loadClass(ClassLoader.java:351)
        ... 3 more
Traceback (most recent call last):
  File "cocoeval.py", line 16, in <module>
    cocoEval.evaluate()
  File "/home/teang1995/codes/MAGIC/image_captioning/evaluation/pycocoevalcap/eval.py", line 59, in evaluate
    score, scores = scorer.compute_score(gts, res)
  File "/home/teang1995/codes/MAGIC/image_captioning/evaluation/pycocoevalcap/spice/spice.py", line 69, in compute_score
    subprocess.check_call(spice_cmd, 
  File "/data1/teang1995/anaconda3/lib/python3.8/subprocess.py", line 364, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['java', '-jar', '-Xmx8G', 'spice-1.0.jar', '/home/teang1995/codes/MAGIC/image_captioning/evaluation/pycocoevalcap/spice/tmp/tmpc2vzaamg', '-cache', '/home/teang1995/codes/MAGIC/image_captioning/evaluation/pycocoevalcap/spice/cache', '-out', '/home/teang1995/codes/MAGIC/image_captioning/evaluation/pycocoevalcap/spice/tmp/tmp1rq2qr4m', '-subset', '-silent']' returned non-zero exit status 1.

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.