fairseq

fairseq copied to clipboard

Upload a model only once & translate in any lang pair as much as possible

3

Hello team, Based on a trained multilingual Fairseq model (e.g. M2M-100), I run my translations as a service in a Docker container according to the following scheme: as an input...

molokanov50

AttributeError: 'dict' object has no attribute 'replace', when loading vq-wav2vec pretrained model as in the instructions

3

## 🐛 Bug I just run the code from the [instructions](https://github.com/pytorch/fairseq/blob/master/examples/wav2vec/README.md#vq-wav2vec) in Google Colab: ``` import torch import fairseq cp = torch.load('/path/to/vq-wav2vec.pt') model, cfg, task = fairseq.checkpoint_utils.load_model_ensemble_and_task([cp]) model = model[0]...

aviasd

Gradient overflow when running the example of Adaptive Input Representations

9

## ❓ Questions and Help #### What is your question? Got `inf` loss and gradient overflow when running the [code example of adaptive input representation](https://github.com/pytorch/fairseq/blob/main/examples/language_model/README.adaptive_inputs.md) with `--fp16`. I am trying...

ghrua

## ❓ Questions and Help It is a great contribution to release such a massive inclusion of the languages and dialects into a big translation model. Thanks :+1: #### Opposed...

giyaseddin

How to reproduce the speech data2vec result?

1

I follow the the instructions in page (https://github.com/facebookresearch/fairseq/tree/main/examples/data2vec) to reproduce the speech data2vec result. And i have got the pretrained model. Then start to use "fairseq-hydra-train \ distributed_training.distributed_port=$PORT \ task.data=/path/to/data...

yangjiabupt

KeyError: 27707 in pytorch 1.7.1 (GeForce RTX 3090)

4

## 🐛 Bug Not understand the meaning of "KeyError: 27707" , it seems to something wrong in multiple processes? I used the same command line in fairseq0.9.0 , pytorch 1.5...

Daisy-123

Error When fine-tuning from others model.bin : File "~/fairseq/fairseq/checkpoint_utils.py", line 581, in _upgrade_state_dict {"criterion_name": "CrossEntropyCriterion", "best_loss": state["best_loss"]} KeyError: 'best_loss'

3

## 🐛 Bug Hi, I am trying to do more pre-training epochs of Roberta's pre-trained model on some low-resource - English language- the model I want to use is the...

muhammed-saeed

Error while training with translation_multi_simple_epoch using custom datasets

3

#### What is your question? Hello! Some months ago, I successfully used the [previous version](https://github.com/pytorch/fairseq/tree/master/examples/translation#multilingual-translation) for multilingual translation with custom datasets. I recently noticed there's a [new one](https://github.com/pytorch/fairseq/tree/master/examples/multilingual) so I...

feralvam

[data2vec] Loss goes down and up again

10

I try to train data2vec on music data (the FMA dataset). I've made some modifications to the feature extractor ConvNet (I've made it a small ResNet essentially), and reduced the...

Tomsen1410

mBART large 50 many to many model giving incorrect long and repetitive output in Portuguese->Eng , Spanish->Eng

1

I am using hugging face facebook/mbart-large-50-many-to-many-mmt model for translating text to English. I observed that its giving vague and incorrect output in many instances. Few example below  

hardikairen27