-
I still have the problem which was also mentioned a lot on the huggingface challenge discord earlier, that pyctcdecode doesn't really like putting spaces in the transcription, e.g.:
`hetcontenenschi…
iskaj updated
9 months ago
-
Three types of docid representations are introduced in the paper "Transformer Memory as a Differentiable Search Index," namely, `Unstructured Atomic Identifiers`, `Naively Structured String Identifier…
-
大佬你好，我用https://github.com/YunwenTechnology/Unilm
提供的微博新闻摘要数据（从中随机挑选10000篇作为训练集，1000篇作为测试集）测试了下GPT2，发现rouge-1只有不到20%，而UniLM给出的结果有40.58%，请问这大概是什么原因？是GPT2的效果就是不好吗
-
按照文章的介绍，是为了更好地进行一对多生成，但事实上seq2seq模型本身就可以通过采样生成（而不是beam search确定性生成），所以原则上seq2seq模型本身就包含了一对多生成能力，文章所说的常规seq2seq不能很好地做一对多生成的断言似乎不能成立。
那么，隐变量的意义何在呢？此外，我没看到关于隐变量的正则项，那么如何保证隐变量的分布不会退化为一个one hot分布呢（即变成只有…
-
### Branch/Tag/Commit
main
### Docker Image Version
nvcr.io/nvidia/pytorch:22.09-py3
### GPU name
V100
### CUDA Driver
Driver Version: 450.80.02 CUDA Version: 11.8
### Reproduced Steps
```…
-
### Your current environment
v0.5.2. vLLM env is not an issue so I will just skip the collection process
### 🐛 Describe the bug
I am running benchmark tests and notice one potential problem. …
-
I'm getting this traceback with errors when running the CLIP captioning:
```
Traceback (most recent call last):
File "C:\Automatic1111\extensions\sd_smartprocess\smartprocess.py", line 360, in …
-
Sorry for bothering you, I just want to say your implementation is really nice, and I want to learn NLP too, but the dataset you provided with the github link doesn't look like the file that you locat…
-
Hello, tThank you for your contribution. Howeverm I notice that all mbart models exceed 2GB. Do you have any plan to fix this issue?
-
Hello! I'm trying to use PrefixTuning with T5 model. After reading source codes in seq2seq, I figure that generally speaking, prefix is added to the BART model by using the parameter _past_key_values…