-
[paper](https://arxiv.org/abs/1906.04341)
## Abstract
Large pre-trained neural networks such as BERT have had great recent success in NLP, motivating a growing body of research investigating wha…
-
Is there any way to compute surprisal for Chinese sentences? Right now, the Chinese characters are processed in a weird way and the output does not match the number of Chinese characters in the input.…
-
### 💻 系统环境
macOS
### 📦 部署环境
Docker
### 🌐 浏览器
Chrome
### 🐛 问题描述
对于列表的渲染居然不一样！
…
-
OpenAI recently announced they've added logprob access to the Chat completion endpoint: https://twitter.com/OpenAIDevs/status/1735730662362189872
Would be useful to add support for computing surpri…
-
Lan et al compute their effect size by taking the difference in surprisals over a critical region of grammatical and ungrammatical sentences with & without fillers, and then seeing if the difference o…
-
OpenAI no longer provides log probs for the prompt, making it impossible to use as a probability over a string function. It does, however, continue to provide logprobs over its own completions.
E.g…
-
现有的gpt2-chinese-cluecorpussmall预训练模型，token是字符级别，请问下：想获取词汇级别的生成概率，应该如何计算呢~
-
您好，我是这个领域的初学者。当我在使用这个代码时，遇到了如下问题，希望可以得到你们的解决与回复：
1.srilm安装并不像代码里直接写的pip就可以使用，不知道是否有相关的教程。
2.由于是该领域的初学者，在使用gpt2计算后续的surprisal和entropy时，代码显示的是与1.2类似。由于缺乏相关的经验实在很难实现，不知道可否对其进行补充，以方便使用。
我知道我的问题可能…
-
### System Info
I was considering the BLIP2 model for getting the probability distribution of each token in the caption given an image. So basically if the words in a caption are w1,w2,w3,…wt then …
-
In `model.Transformer.forward`, the following line says it'll only compute the logits for the last position in h:
```python
output = self.output(h[:, -1, :]) # only compute last logits
```
I'm in…