pipeline_tag stringclasses 48 values | library_name stringclasses 198 values | text stringlengths 1 900k | metadata stringlengths 2 438k | id stringlengths 5 122 | last_modified null | tags listlengths 1 1.84k | sha null | created_at stringlengths 25 25 | arxiv listlengths 0 201 | languages listlengths 0 1.83k | tags_str stringlengths 17 9.34k | text_str stringlengths 0 389k | text_lists listlengths 0 722 | processed_texts listlengths 1 723 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
automatic-speech-recognition | transformers |
# bp500-base100k_voxpopuli: Wav2vec 2.0 with Brazilian Portuguese (BP) Dataset
This is a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
- [CETUC](http://www02.smt.ufrj.br/~igor.quintanilha/alcaim.tar.gz): contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the [CETEN-Folha](https://www.linguateca.pt/cetenfolha/) corpus.
- [Common Voice 7.0](https://commonvoice.mozilla.org/pt): is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages. In this project, volunteers donate and validate speech using the [oficial site](https://commonvoice.mozilla.org/pt).
- [Lapsbm](https://github.com/falabrasil/gitlab-resources): "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
- [Multilingual Librispeech (MLS)](https://arxiv.org/abs/2012.03411): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like [LibriVox](https://librivox.org/). The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese [used in this work](http://www.openslr.org/94/) (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
- [Multilingual TEDx](http://www.openslr.org/100): a collection of audio recordings from TEDx talks in 8 source languages. The Portuguese set (mostly Brazilian Portuguese variant) contains 164 hours of transcribed speech.
- [Sidney](https://igormq.github.io/datasets/) (SID): contains 5,777 utterances recorded by 72 speakers (20 women) from 17 to 59 years old with fields such as place of birth, age, gender, education, and occupation;
- [VoxForge](http://www.voxforge.org/): is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively. We also made test sets for all the gathered datasets.
| Dataset | Train | Valid | Test |
|--------------------------------|-------:|------:|------:|
| CETUC | 94.0h | -- | 5.4h |
| Common Voice | 37.8h | 8.9h | 9.5h |
| LaPS BM | 0.8h | -- | 0.1h |
| MLS | 161.0h | -- | 3.7h |
| Multilingual TEDx (Portuguese) | 148.9h | -- | 1.8h |
| SID | 7.2h | -- | 1.0h |
| VoxForge | 3.9h | -- | 0.1h |
| Total | 453.6h | 8.9h | 21.6h |
The original model was fine-tuned using [fairseq](https://github.com/pytorch/fairseq). This notebook uses a converted version of the original one. The link to the original fairseq model is available [here](https://drive.google.com/file/d/10iESR5AQxuxF5F7w3wLbpc_9YMsYbY9H/view?usp=sharing).
#### Summary
| | CETUC | CV | LaPS | MLS | SID | TEDx | VF | AVG |
|----------------------|---------------|----------------|----------------|----------------|----------------|----------------|----------------|----------------|
| bp\_500-base100k_voxpopuli (demonstration below) | 0.142 | 0.201 | 0.052 | 0.224 | 0.102 | 0.317 | 0.048 | 0.155 |
| bp\_500-base100k_voxpopuli + 4-gram (demonstration below) | 0.099 | 0.149 | 0.047 | 0.192 | 0.115 | 0.371 | 0.127 | 0.157 |
#### Transcription examples
| Text | Transcription |
|------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------|
|qual o instagram dele|**qualo** **está** **gramedele**|
|o capitão foi expulso do exército porque era doido|o **capitãl** foi **exposo** do exército porque era doido|
|também por que não|também **porque** não|
|não existe tempo como o presente|não existe tempo como *o* presente|
|eu pulei para salvar rachel|eu pulei para salvar **haquel**|
|augusto cezar passos marinho|augusto **cesa** **passoesmarinho**|
## Demonstration
```python
MODEL_NAME = "lgris/bp500-base100k_voxpopuli"
```
### Imports and dependencies
```python
%%capture
!pip install torch==1.8.2+cu111 torchvision==0.9.2+cu111 torchaudio===0.8.2 -f https://download.pytorch.org/whl/lts/1.8/torch_lts.html
!pip install datasets
!pip install jiwer
!pip install transformers
!pip install soundfile
!pip install pyctcdecode
!pip install https://github.com/kpu/kenlm/archive/master.zip
```
```python
import jiwer
import torchaudio
from datasets import load_dataset, load_metric
from transformers import (
Wav2Vec2ForCTC,
Wav2Vec2Processor,
)
from pyctcdecode import build_ctcdecoder
import torch
import re
import sys
```
### Helpers
```python
chars_to_ignore_regex = '[\,\?\.\!\;\:\"]' # noqa: W605
def map_to_array(batch):
speech, _ = torchaudio.load(batch["path"])
batch["speech"] = speech.squeeze(0).numpy()
batch["sampling_rate"] = 16_000
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
batch["target"] = batch["sentence"]
return batch
```
```python
def calc_metrics(truths, hypos):
wers = []
mers = []
wils = []
for t, h in zip(truths, hypos):
try:
wers.append(jiwer.wer(t, h))
mers.append(jiwer.mer(t, h))
wils.append(jiwer.wil(t, h))
except: # Empty string?
pass
wer = sum(wers)/len(wers)
mer = sum(mers)/len(mers)
wil = sum(wils)/len(wils)
return wer, mer, wil
```
```python
def load_data(dataset):
data_files = {'test': f'{dataset}/test.csv'}
dataset = load_dataset('csv', data_files=data_files)["test"]
return dataset.map(map_to_array)
```
### Model
```python
class STT:
def __init__(self,
model_name,
device='cuda' if torch.cuda.is_available() else 'cpu',
lm=None):
self.model_name = model_name
self.model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
self.processor = Wav2Vec2Processor.from_pretrained(model_name)
self.vocab_dict = self.processor.tokenizer.get_vocab()
self.sorted_dict = {
k.lower(): v for k, v in sorted(self.vocab_dict.items(),
key=lambda item: item[1])
}
self.device = device
self.lm = lm
if self.lm:
self.lm_decoder = build_ctcdecoder(
list(self.sorted_dict.keys()),
self.lm
)
def batch_predict(self, batch):
features = self.processor(batch["speech"],
sampling_rate=batch["sampling_rate"][0],
padding=True,
return_tensors="pt")
input_values = features.input_values.to(self.device)
with torch.no_grad():
logits = self.model(input_values).logits
if self.lm:
logits = logits.cpu().numpy()
batch["predicted"] = []
for sample_logits in logits:
batch["predicted"].append(self.lm_decoder.decode(sample_logits))
else:
pred_ids = torch.argmax(logits, dim=-1)
batch["predicted"] = self.processor.batch_decode(pred_ids)
return batch
```
### Download datasets
```python
%%capture
!gdown --id 1HFECzIizf-bmkQRLiQD0QVqcGtOG5upI
!mkdir bp_dataset
!unzip bp_dataset -d bp_dataset/
```
```python
%cd bp_dataset
```
/content/bp_dataset
### Tests
```python
stt = STT(MODEL_NAME)
```
#### CETUC
```python
ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)
```
CETUC WER: 0.1419179499917191
#### Common Voice
```python
ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)
```
CV WER: 0.20079950312040154
#### LaPS
```python
ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)
```
Laps WER: 0.052780934343434324
#### MLS
```python
ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)
```
MLS WER: 0.22413887199364113
#### SID
```python
ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)
```
Sid WER: 0.1019041538671034
#### TEDx
```python
ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)
```
TEDx WER: 0.31711268778273327
#### VoxForge
```python
ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)
```
VoxForge WER: 0.04826433982683982
### Tests with LM
```python
!rm -rf ~/.cache
!gdown --id 1GJIKseP5ZkTbllQVgOL98R4yYAcIySFP # trained with wikipedia
stt = STT(MODEL_NAME, lm='pt-BR-wiki.word.4-gram.arpa')
# !gdown --id 1dLFldy7eguPtyJj5OAlI4Emnx0BpFywg # trained with bp
# stt = STT(MODEL_NAME, lm='pt-BR.word.4-gram.arpa')
```
### Cetuc
```python
ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)
```
CETUC WER: 0.099518615112877
#### Common Voice
```python
ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)
```
CV WER: 0.1488912889506362
#### LaPS
```python
ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)
```
Laps WER: 0.047080176767676764
#### MLS
```python
ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)
```
MLS WER: 0.19220291966887196
#### SID
```python
ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)
```
Sid WER: 0.11535498771650306
#### TEDx
```python
ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)
```
TEDx WER: 0.3707890073539895
#### VoxForge
```python
ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)
```
VoxForge WER: 0.12682088744588746
| {"language": "pt", "license": "apache-2.0", "tags": ["audio", "speech", "wav2vec2", "pt", "portuguese-speech-corpus", "automatic-speech-recognition", "speech", "PyTorch"], "datasets": ["common_voice", "mls", "cetuc", "lapsbm", "voxforge", "tedx", "sid"], "metrics": ["wer"]} | lgris/bp500-base100k_voxpopuli | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"audio",
"speech",
"pt",
"portuguese-speech-corpus",
"PyTorch",
"dataset:common_voice",
"dataset:mls",
"dataset:cetuc",
"dataset:lapsbm",
"dataset:voxforge",
"dataset:tedx",
"dataset:sid",
"arxiv:2012.03411",
"l... | null | 2022-03-02T23:29:05+00:00 | [
"2012.03411"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #dataset-tedx #dataset-sid #arxiv-2012.03411 #license-apache-2.0 #endpoints_compatible #region-us
| bp500-base100k\_voxpopuli: Wav2vec 2.0 with Brazilian Portuguese (BP) Dataset
=============================================================================
This is a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
* CETUC: contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the CETEN-Folha corpus.
* Common Voice 7.0: is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages. In this project, volunteers donate and validate speech using the oficial site.
* Lapsbm: "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
* Multilingual Librispeech (MLS): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like LibriVox. The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese used in this work (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
* Multilingual TEDx: a collection of audio recordings from TEDx talks in 8 source languages. The Portuguese set (mostly Brazilian Portuguese variant) contains 164 hours of transcribed speech.
* Sidney (SID): contains 5,777 utterances recorded by 72 speakers (20 women) from 17 to 59 years old with fields such as place of birth, age, gender, education, and occupation;
* VoxForge: is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively. We also made test sets for all the gathered datasets.
The original model was fine-tuned using fairseq. This notebook uses a converted version of the original one. The link to the original fairseq model is available here.
#### Summary
#### Transcription examples
Demonstration
-------------
### Imports and dependencies
### Helpers
### Model
### Download datasets
```
/content/bp_dataset
```
### Tests
#### CETUC
```
CETUC WER: 0.1419179499917191
```
#### Common Voice
```
CV WER: 0.20079950312040154
```
#### LaPS
```
Laps WER: 0.052780934343434324
```
#### MLS
```
MLS WER: 0.22413887199364113
```
#### SID
```
Sid WER: 0.1019041538671034
```
#### TEDx
```
TEDx WER: 0.31711268778273327
```
#### VoxForge
```
VoxForge WER: 0.04826433982683982
```
### Tests with LM
### Cetuc
```
CETUC WER: 0.099518615112877
```
#### Common Voice
```
CV WER: 0.1488912889506362
```
#### LaPS
```
Laps WER: 0.047080176767676764
```
#### MLS
```
MLS WER: 0.19220291966887196
```
#### SID
```
Sid WER: 0.11535498771650306
```
#### TEDx
```
TEDx WER: 0.3707890073539895
```
#### VoxForge
```
VoxForge WER: 0.12682088744588746
```
| [
"#### Summary",
"#### Transcription examples\n\n\n\nDemonstration\n-------------",
"### Imports and dependencies",
"### Helpers",
"### Model",
"### Download datasets\n\n\n\n```\n/content/bp_dataset\n\n```",
"### Tests",
"#### CETUC\n\n\n\n```\nCETUC WER: 0.1419179499917191\n\n```",
"#### Common Voic... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #dataset-tedx #dataset-sid #arxiv-2012.03411 #license-apache-2.0 #endpoints_compatible #region-us \n",
"##... |
automatic-speech-recognition | transformers |
# bp500-base10k_voxpopuli: Wav2vec 2.0 with Brazilian Portuguese (BP) Dataset
This is a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
- [CETUC](http://www02.smt.ufrj.br/~igor.quintanilha/alcaim.tar.gz): contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the [CETEN-Folha](https://www.linguateca.pt/cetenfolha/) corpus.
- [Common Voice 7.0](https://commonvoice.mozilla.org/pt): is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages. In this project, volunteers donate and validate speech using the [oficial site](https://commonvoice.mozilla.org/pt).
- [Lapsbm](https://github.com/falabrasil/gitlab-resources): "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
- [Multilingual Librispeech (MLS)](https://arxiv.org/abs/2012.03411): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like [LibriVox](https://librivox.org/). The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese [used in this work](http://www.openslr.org/94/) (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
- [Multilingual TEDx](http://www.openslr.org/100): a collection of audio recordings from TEDx talks in 8 source languages. The Portuguese set (mostly Brazilian Portuguese variant) contains 164 hours of transcribed speech.
- [Sidney](https://igormq.github.io/datasets/) (SID): contains 5,777 utterances recorded by 72 speakers (20 women) from 17 to 59 years old with fields such as place of birth, age, gender, education, and occupation;
- [VoxForge](http://www.voxforge.org/): is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively. We also made test sets for all the gathered datasets.
| Dataset | Train | Valid | Test |
|--------------------------------|-------:|------:|------:|
| CETUC | 94.0h | -- | 5.4h |
| Common Voice | 37.8h | 8.9h | 9.5h |
| LaPS BM | 0.8h | -- | 0.1h |
| MLS | 161.0h | -- | 3.7h |
| Multilingual TEDx (Portuguese) | 148.9h | -- | 1.8h |
| SID | 7.2h | -- | 1.0h |
| VoxForge | 3.9h | -- | 0.1h |
| Total | 453.6h | 8.9h | 21.6h |
The original model was fine-tuned using [fairseq](https://github.com/pytorch/fairseq). This notebook uses a converted version of the original one. The link to the original fairseq model is available [here](https://drive.google.com/file/d/19kkENi8uvczmw9OLSdqnjvKqBE53cl_W/view?usp=sharing).
#### Summary
| | CETUC | CV | LaPS | MLS | SID | TEDx | VF | AVG |
|----------------------|---------------|----------------|----------------|----------------|----------------|----------------|----------------|----------------|
| bp\_500-base10k_voxpopuli (demonstration below) | 0.120 | 0.249 | 0.039 | 0.227 | 0.169 | 0.349 | 0.116 | 0.181 |
| bp\_500-base10k_voxpopuli + 4-gram (demonstration below) | 0.074 | 0.174 | 0.032 | 0.182 | 0.181 | 0.349 | 0.111 | 0.157 |
#### Transcription examples
| Text | Transcription |
|------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------|
|suco de uva e água misturam bem|suco **deúva** e água **misturão** bem|
|culpa do dinheiro|**cupa** do dinheiro|
|eu amo shooters call of duty é o meu favorito|eu **omo** **shúters cofedete** é meu favorito|
|você pode explicar por que isso acontece|você pode explicar *por* que isso **ontece**|
|no futuro você desejará ter começado a investir hoje|no futuro você desejará **a** ter começado a investir hoje|
## Demonstration
```python
MODEL_NAME = "lgris/bp500-base10k_voxpopuli"
```
### Imports and dependencies
```python
%%capture
!pip install torch==1.8.2+cu111 torchvision==0.9.2+cu111 torchaudio===0.8.2 -f https://download.pytorch.org/whl/lts/1.8/torch_lts.html
!pip install datasets
!pip install jiwer
!pip install transformers
!pip install soundfile
!pip install pyctcdecode
!pip install https://github.com/kpu/kenlm/archive/master.zip
```
```python
import jiwer
import torchaudio
from datasets import load_dataset, load_metric
from transformers import (
Wav2Vec2ForCTC,
Wav2Vec2Processor,
)
from pyctcdecode import build_ctcdecoder
import torch
import re
import sys
```
### Helpers
```python
chars_to_ignore_regex = '[\,\?\.\!\;\:\"]' # noqa: W605
def map_to_array(batch):
speech, _ = torchaudio.load(batch["path"])
batch["speech"] = speech.squeeze(0).numpy()
batch["sampling_rate"] = 16_000
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
batch["target"] = batch["sentence"]
return batch
```
```python
def calc_metrics(truths, hypos):
wers = []
mers = []
wils = []
for t, h in zip(truths, hypos):
try:
wers.append(jiwer.wer(t, h))
mers.append(jiwer.mer(t, h))
wils.append(jiwer.wil(t, h))
except: # Empty string?
pass
wer = sum(wers)/len(wers)
mer = sum(mers)/len(mers)
wil = sum(wils)/len(wils)
return wer, mer, wil
```
```python
def load_data(dataset):
data_files = {'test': f'{dataset}/test.csv'}
dataset = load_dataset('csv', data_files=data_files)["test"]
return dataset.map(map_to_array)
```
### Model
```python
class STT:
def __init__(self,
model_name,
device='cuda' if torch.cuda.is_available() else 'cpu',
lm=None):
self.model_name = model_name
self.model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
self.processor = Wav2Vec2Processor.from_pretrained(model_name)
self.vocab_dict = self.processor.tokenizer.get_vocab()
self.sorted_dict = {
k.lower(): v for k, v in sorted(self.vocab_dict.items(),
key=lambda item: item[1])
}
self.device = device
self.lm = lm
if self.lm:
self.lm_decoder = build_ctcdecoder(
list(self.sorted_dict.keys()),
self.lm
)
def batch_predict(self, batch):
features = self.processor(batch["speech"],
sampling_rate=batch["sampling_rate"][0],
padding=True,
return_tensors="pt")
input_values = features.input_values.to(self.device)
with torch.no_grad():
logits = self.model(input_values).logits
if self.lm:
logits = logits.cpu().numpy()
batch["predicted"] = []
for sample_logits in logits:
batch["predicted"].append(self.lm_decoder.decode(sample_logits))
else:
pred_ids = torch.argmax(logits, dim=-1)
batch["predicted"] = self.processor.batch_decode(pred_ids)
return batch
```
### Download datasets
```python
%%capture
!gdown --id 1HFECzIizf-bmkQRLiQD0QVqcGtOG5upI
!mkdir bp_dataset
!unzip bp_dataset -d bp_dataset/
```
```python
%cd bp_dataset
```
/content/bp_dataset
### Tests
```python
stt = STT(MODEL_NAME)
```
#### CETUC
```python
ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)
```
CETUC WER: 0.12096759949218888
#### Common Voice
```python
ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)
```
CV WER: 0.24977003159495725
#### LaPS
```python
ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)
```
Laps WER: 0.039769570707070705
#### MLS
```python
ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)
```
MLS WER: 0.2269637077788063
#### SID
```python
ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)
```
Sid WER: 0.1691680138494731
#### TEDx
```python
ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)
```
TEDx WER: 0.34908555859018014
#### VoxForge
```python
ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)
```
VoxForge WER: 0.11649350649350651
### Tests with LM
```python
!rm -rf ~/.cache
!gdown --id 1GJIKseP5ZkTbllQVgOL98R4yYAcIySFP # trained with wikipedia
stt = STT(MODEL_NAME, lm='pt-BR-wiki.word.4-gram.arpa')
# !gdown --id 1dLFldy7eguPtyJj5OAlI4Emnx0BpFywg # trained with bp
# stt = STT(MODEL_NAME, lm='pt-BR.word.4-gram.arpa')
```
### Cetuc
```python
ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)
```
CETUC WER: 0.07499558425787961
#### Common Voice
```python
ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)
```
CV WER: 0.17442648452610307
#### LaPS
```python
ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)
```
Laps WER: 0.032774621212121206
#### MLS
```python
ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)
```
MLS WER: 0.18213620321569274
#### SID
```python
ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)
```
Sid WER: 0.18102544972868206
#### TEDx
```python
ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)
```
TEDx WER: 0.3491402028105601
#### VoxForge
```python
ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)
```
VoxForge WER: 0.11189529220779222
| {"language": "pt", "license": "apache-2.0", "tags": ["audio", "speech", "wav2vec2", "pt", "portuguese-speech-corpus", "automatic-speech-recognition", "speech", "PyTorch", "hf-asr-leaderboard"], "datasets": ["common_voice", "mls", "cetuc", "lapsbm", "voxforge", "tedx", "sid"], "metrics": ["wer"], "model-index": [{"name": "bp500-base10k_voxpopuli", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice", "type": "common_voice", "args": "pt"}, "metrics": [{"type": "wer", "value": 24.9, "name": "Test WER"}]}]}]} | lgris/bp500-base10k_voxpopuli | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"audio",
"speech",
"pt",
"portuguese-speech-corpus",
"PyTorch",
"hf-asr-leaderboard",
"dataset:common_voice",
"dataset:mls",
"dataset:cetuc",
"dataset:lapsbm",
"dataset:voxforge",
"dataset:tedx",
"dataset:sid",
... | null | 2022-03-02T23:29:05+00:00 | [
"2012.03411"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #dataset-tedx #dataset-sid #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| bp500-base10k\_voxpopuli: Wav2vec 2.0 with Brazilian Portuguese (BP) Dataset
============================================================================
This is a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
* CETUC: contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the CETEN-Folha corpus.
* Common Voice 7.0: is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages. In this project, volunteers donate and validate speech using the oficial site.
* Lapsbm: "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
* Multilingual Librispeech (MLS): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like LibriVox. The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese used in this work (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
* Multilingual TEDx: a collection of audio recordings from TEDx talks in 8 source languages. The Portuguese set (mostly Brazilian Portuguese variant) contains 164 hours of transcribed speech.
* Sidney (SID): contains 5,777 utterances recorded by 72 speakers (20 women) from 17 to 59 years old with fields such as place of birth, age, gender, education, and occupation;
* VoxForge: is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively. We also made test sets for all the gathered datasets.
The original model was fine-tuned using fairseq. This notebook uses a converted version of the original one. The link to the original fairseq model is available here.
#### Summary
#### Transcription examples
Demonstration
-------------
### Imports and dependencies
### Helpers
### Model
### Download datasets
```
/content/bp_dataset
```
### Tests
#### CETUC
```
CETUC WER: 0.12096759949218888
```
#### Common Voice
```
CV WER: 0.24977003159495725
```
#### LaPS
```
Laps WER: 0.039769570707070705
```
#### MLS
```
MLS WER: 0.2269637077788063
```
#### SID
```
Sid WER: 0.1691680138494731
```
#### TEDx
```
TEDx WER: 0.34908555859018014
```
#### VoxForge
```
VoxForge WER: 0.11649350649350651
```
### Tests with LM
### Cetuc
```
CETUC WER: 0.07499558425787961
```
#### Common Voice
```
CV WER: 0.17442648452610307
```
#### LaPS
```
Laps WER: 0.032774621212121206
```
#### MLS
```
MLS WER: 0.18213620321569274
```
#### SID
```
Sid WER: 0.18102544972868206
```
#### TEDx
```
TEDx WER: 0.3491402028105601
```
#### VoxForge
```
VoxForge WER: 0.11189529220779222
```
| [
"#### Summary",
"#### Transcription examples\n\n\n\nDemonstration\n-------------",
"### Imports and dependencies",
"### Helpers",
"### Model",
"### Download datasets\n\n\n\n```\n/content/bp_dataset\n\n```",
"### Tests",
"#### CETUC\n\n\n\n```\nCETUC WER: 0.12096759949218888\n\n```",
"#### Common Voi... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #dataset-tedx #dataset-sid #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints... |
automatic-speech-recognition | transformers |
# bp500-xlsr: Wav2vec 2.0 with Brazilian Portuguese (BP) Dataset
This is a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
- [CETUC](http://www02.smt.ufrj.br/~igor.quintanilha/alcaim.tar.gz): contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the [CETEN-Folha](https://www.linguateca.pt/cetenfolha/) corpus;
- [Common Voice 7.0](https://commonvoice.mozilla.org/pt): is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages. In this project, volunteers donate and validate speech using the [oficial site](https://commonvoice.mozilla.org/pt);
- [Lapsbm](https://github.com/falabrasil/gitlab-resources): "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control;
- [Multilingual Librispeech (MLS)](https://arxiv.org/abs/2012.03411): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like [LibriVox](https://librivox.org/). The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese [used in this work](http://www.openslr.org/94/) (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers;
- [VoxForge](http://www.voxforge.org/): is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively. We also made test sets for all the gathered datasets.
| Dataset | Train | Valid | Test |
|--------------------------------|-------:|------:|------:|
| CETUC | 93.9h | -- | 5.4h |
| Common Voice | 37.6h | 8.9h | 9.5h |
| LaPS BM | 0.8h | -- | 0.1h |
| MLS | 161.0h | -- | 3.7h |
| Multilingual TEDx (Portuguese) | 144.2h | -- | 1.8h |
| SID | 5.0h | -- | 1.0h |
| VoxForge | 2.8h | -- | 0.1h |
| Total | 437.2h | 8.9h | 21.6h |
The original model was fine-tuned using [fairseq](https://github.com/pytorch/fairseq). This notebook uses a converted version of the original one. The link to the original fairseq model is available [here](https://drive.google.com/file/d/1J8aR1ltDLQFe-dVrGuyxoRm2uyJjCWgf/view?usp=sharing).
#### Summary
| | CETUC | CV | LaPS | MLS | SID | TEDx | VF | AVG |
|----------------------|---------------|----------------|----------------|----------------|----------------|----------------|----------------|----------------|
| bp\_500 (demonstration below) | 0.051 | 0.136 | 0.032 | 0.118 | 0.095 | 0.248 | 0.082 | 0.108 |
| bp\_500 + 4-gram (demonstration below) | 0.032 | 0.097 | 0.022 | 0.114 | 0.125 | 0.246 | 0.065 | 0.100 |
#### Transcription examples
| Text | Transcription |
|------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------|
|não há um departamento de mediadores independente das federações e das agremiações|não há um **dearamento** de mediadores independente das federações e das **agrebiações**|
|mas que bodega|**masque** bodega|
|a cortina abriu o show começou|a cortina abriu o **chô** começou|
|por sorte havia uma passadeira|**busote avinhoa** **passadeiro**|
|estou maravilhada está tudo pronto|**stou** estou maravilhada está tudo pronto|
## Demonstration
```python
MODEL_NAME = "lgris/bp500-xlsr"
```
### Imports and dependencies
```python
%%capture
!pip install torch==1.8.2+cu111 torchvision==0.9.2+cu111 torchaudio===0.8.2 -f https://download.pytorch.org/whl/lts/1.8/torch_lts.html
!pip install datasets
!pip install jiwer
!pip install transformers
!pip install soundfile
!pip install pyctcdecode
!pip install https://github.com/kpu/kenlm/archive/master.zip
```
```python
import jiwer
import torchaudio
from datasets import load_dataset, load_metric
from transformers import (
Wav2Vec2ForCTC,
Wav2Vec2Processor,
)
from pyctcdecode import build_ctcdecoder
import torch
import re
import sys
```
### Helpers
```python
chars_to_ignore_regex = '[\,\?\.\!\;\:\"]' # noqa: W605
def map_to_array(batch):
speech, _ = torchaudio.load(batch["path"])
batch["speech"] = speech.squeeze(0).numpy()
batch["sampling_rate"] = 16_000
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
batch["target"] = batch["sentence"]
return batch
```
```python
def calc_metrics(truths, hypos):
wers = []
mers = []
wils = []
for t, h in zip(truths, hypos):
try:
wers.append(jiwer.wer(t, h))
mers.append(jiwer.mer(t, h))
wils.append(jiwer.wil(t, h))
except: # Empty string?
pass
wer = sum(wers)/len(wers)
mer = sum(mers)/len(mers)
wil = sum(wils)/len(wils)
return wer, mer, wil
```
```python
def load_data(dataset):
data_files = {'test': f'{dataset}/test.csv'}
dataset = load_dataset('csv', data_files=data_files)["test"]
return dataset.map(map_to_array)
```
### Model
```python
class STT:
def __init__(self,
model_name,
device='cuda' if torch.cuda.is_available() else 'cpu',
lm=None):
self.model_name = model_name
self.model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
self.processor = Wav2Vec2Processor.from_pretrained(model_name)
self.vocab_dict = self.processor.tokenizer.get_vocab()
self.sorted_dict = {
k.lower(): v for k, v in sorted(self.vocab_dict.items(),
key=lambda item: item[1])
}
self.device = device
self.lm = lm
if self.lm:
self.lm_decoder = build_ctcdecoder(
list(self.sorted_dict.keys()),
self.lm
)
def batch_predict(self, batch):
features = self.processor(batch["speech"],
sampling_rate=batch["sampling_rate"][0],
padding=True,
return_tensors="pt")
input_values = features.input_values.to(self.device)
attention_mask = features.attention_mask.to(self.device)
with torch.no_grad():
logits = self.model(input_values, attention_mask=attention_mask).logits
if self.lm:
logits = logits.cpu().numpy()
batch["predicted"] = []
for sample_logits in logits:
batch["predicted"].append(self.lm_decoder.decode(sample_logits))
else:
pred_ids = torch.argmax(logits, dim=-1)
batch["predicted"] = self.processor.batch_decode(pred_ids)
return batch
```
### Download datasets
```python
%%capture
!gdown --id 1HFECzIizf-bmkQRLiQD0QVqcGtOG5upI
!mkdir bp_dataset
!unzip bp_dataset -d bp_dataset/
```
```python
%cd bp_dataset
```
/content/bp_dataset
### Tests
```python
stt = STT(MODEL_NAME)
```
#### CETUC
```python
ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)
```
CETUC WER: 0.05159097808687998
#### Common Voice
```python
ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)
```
CV WER: 0.13659981509705973
#### LaPS
```python
ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)
```
Laps WER: 0.03196969696969697
#### MLS
```python
ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)
```
MLS WER: 0.1178481066463896
#### SID
```python
ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)
```
Sid WER: 0.09544588416964224
#### TEDx
```python
ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)
```
TEDx WER: 0.24868046340420813
#### VoxForge
```python
ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)
```
VoxForge WER: 0.08246076839826841
### Tests with LM
```python
!rm -rf ~/.cache
!gdown --id 1GJIKseP5ZkTbllQVgOL98R4yYAcIySFP # trained with wikipedia
stt = STT(MODEL_NAME, lm='pt-BR-wiki.word.4-gram.arpa')
# !gdown --id 1dLFldy7eguPtyJj5OAlI4Emnx0BpFywg # trained with bp
# stt = STT(MODEL_NAME, lm='pt-BR.word.4-gram.arpa')
```
### Cetuc
```python
ds = load_data('cetuc_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CETUC WER:", wer)
```
CETUC WER: 0.03222801788375573
#### Common Voice
```python
ds = load_data('commonvoice_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("CV WER:", wer)
```
CV WER: 0.09713866021093655
#### LaPS
```python
ds = load_data('lapsbm_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Laps WER:", wer)
```
Laps WER: 0.022310606060606065
#### MLS
```python
ds = load_data('mls_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("MLS WER:", wer)
```
MLS WER: 0.11408590958696524
#### SID
```python
ds = load_data('sid_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("Sid WER:", wer)
```
Sid WER: 0.12502797252979136
#### TEDx
```python
ds = load_data('tedx_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("TEDx WER:", wer)
```
TEDx WER: 0.24603179403904793
#### VoxForge
```python
ds = load_data('voxforge_dataset')
result = ds.map(stt.batch_predict, batched=True, batch_size=8)
wer, mer, wil = calc_metrics(result["sentence"], result["predicted"])
print("VoxForge WER:", wer)
```
VoxForge WER: 0.06542207792207791
| {"language": "pt", "license": "apache-2.0", "tags": ["audio", "speech", "wav2vec2", "pt", "portuguese-speech-corpus", "automatic-speech-recognition", "speech", "PyTorch", "hf-asr-leaderboard"], "datasets": ["common_voice", "mls", "cetuc", "lapsbm", "voxforge", "tedx", "sid"], "metrics": ["wer"], "model-index": [{"name": "bp400-xlsr", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice", "type": "common_voice", "args": "pt"}, "metrics": [{"type": "wer", "value": 13.6, "name": "Test WER"}]}]}]} | lgris/bp500-xlsr | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"audio",
"speech",
"pt",
"portuguese-speech-corpus",
"PyTorch",
"hf-asr-leaderboard",
"dataset:common_voice",
"dataset:mls",
"dataset:cetuc",
"dataset:lapsbm",
"dataset:voxforge",
"dataset:tedx",
"dataset:sid",
... | null | 2022-03-02T23:29:05+00:00 | [
"2012.03411"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #dataset-tedx #dataset-sid #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| bp500-xlsr: Wav2vec 2.0 with Brazilian Portuguese (BP) Dataset
==============================================================
This is a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
* CETUC: contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the CETEN-Folha corpus;
* Common Voice 7.0: is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages. In this project, volunteers donate and validate speech using the oficial site;
* Lapsbm: "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control;
* Multilingual Librispeech (MLS): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like LibriVox. The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese used in this work (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers;
* VoxForge: is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively. We also made test sets for all the gathered datasets.
The original model was fine-tuned using fairseq. This notebook uses a converted version of the original one. The link to the original fairseq model is available here.
#### Summary
#### Transcription examples
Demonstration
-------------
### Imports and dependencies
### Helpers
### Model
### Download datasets
```
/content/bp_dataset
```
### Tests
#### CETUC
```
CETUC WER: 0.05159097808687998
```
#### Common Voice
```
CV WER: 0.13659981509705973
```
#### LaPS
```
Laps WER: 0.03196969696969697
```
#### MLS
```
MLS WER: 0.1178481066463896
```
#### SID
```
Sid WER: 0.09544588416964224
```
#### TEDx
```
TEDx WER: 0.24868046340420813
```
#### VoxForge
```
VoxForge WER: 0.08246076839826841
```
### Tests with LM
### Cetuc
```
CETUC WER: 0.03222801788375573
```
#### Common Voice
```
CV WER: 0.09713866021093655
```
#### LaPS
```
Laps WER: 0.022310606060606065
```
#### MLS
```
MLS WER: 0.11408590958696524
```
#### SID
```
Sid WER: 0.12502797252979136
```
#### TEDx
```
TEDx WER: 0.24603179403904793
```
#### VoxForge
```
VoxForge WER: 0.06542207792207791
```
| [
"#### Summary",
"#### Transcription examples\n\n\n\nDemonstration\n-------------",
"### Imports and dependencies",
"### Helpers",
"### Model",
"### Download datasets\n\n\n\n```\n/content/bp_dataset\n\n```",
"### Tests",
"#### CETUC\n\n\n\n```\nCETUC WER: 0.05159097808687998\n\n```",
"#### Common Voi... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #dataset-tedx #dataset-sid #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints... |
automatic-speech-recognition | transformers |
# bp_400h_xlsr2_300M | {"language": ["pt"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "mozilla-foundation/common_voice_7_0", "pt", "hf-asr-leaderboard"], "model-index": [{"name": "bp_400h_xlsr2_300M", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 7", "type": "mozilla-foundation/common_voice_7_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 10.83, "name": "Test WER"}, {"type": "cer", "value": 3.11, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 22.48, "name": "Test WER"}, {"type": "cer", "value": 9.33, "name": "Test CER"}]}]}]} | lgris/bp_400h_xlsr2_300M | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"mozilla-foundation/common_voice_7_0",
"pt",
"hf-asr-leaderboard",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #mozilla-foundation/common_voice_7_0 #pt #hf-asr-leaderboard #license-apache-2.0 #model-index #endpoints_compatible #region-us
|
# bp_400h_xlsr2_300M | [
"# bp_400h_xlsr2_300M"
] | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #mozilla-foundation/common_voice_7_0 #pt #hf-asr-leaderboard #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"# bp_400h_xlsr2_300M"
] |
feature-extraction | transformers |
# DistilXLSR-53 for BP
[DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese](https://github.com/s3prl/s3prl/tree/master/s3prl/upstream/distiller)
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
**Note**: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model **speech recognition**, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more in-detail explanation of how to fine-tune the model.
Paper: [DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT](https://arxiv.org/abs/2110.01900)
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
**Note 2**: The XLSR-53 model was distilled using [Brazilian Portuguese Datasets](https://huggingface.co/lgris/bp400-xlsr) for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the [original work](https://arxiv.org/abs/2110.01900)).
**Abstract**
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more information on how to fine-tune the model.
| {"language": "pt", "license": "apache-2.0", "tags": ["speech"]} | lgris/distilxlsr_bp_12-16 | null | [
"transformers",
"pytorch",
"wav2vec2",
"feature-extraction",
"speech",
"pt",
"arxiv:2110.01900",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2110.01900"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us
|
# DistilXLSR-53 for BP
DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model.
Paper: DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
Note 2: The XLSR-53 model was distilled using Brazilian Portuguese Datasets for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the original work).
Abstract
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See this blog for more information on how to fine-tune the model.
| [
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.\n\nNote: This model does not have a tokenizer as it was pretrain... | [
"TAGS\n#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us \n",
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech a... |
feature-extraction | transformers |
# DistilXLSR-53 for BP
[DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese](https://github.com/s3prl/s3prl/tree/master/s3prl/upstream/distiller)
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
**Note**: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model **speech recognition**, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more in-detail explanation of how to fine-tune the model.
Paper: [DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT](https://arxiv.org/abs/2110.01900)
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
**Note 2**: The XLSR-53 model was distilled using [Brazilian Portuguese Datasets](https://huggingface.co/lgris/bp400-xlsr) for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the [original work](https://arxiv.org/abs/2110.01900)).
**Abstract**
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more information on how to fine-tune the model.
| {"language": "pt", "license": "apache-2.0", "tags": ["speech"]} | lgris/distilxlsr_bp_16-24 | null | [
"transformers",
"pytorch",
"wav2vec2",
"feature-extraction",
"speech",
"pt",
"arxiv:2110.01900",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2110.01900"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us
|
# DistilXLSR-53 for BP
DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model.
Paper: DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
Note 2: The XLSR-53 model was distilled using Brazilian Portuguese Datasets for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the original work).
Abstract
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See this blog for more information on how to fine-tune the model.
| [
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.\n\nNote: This model does not have a tokenizer as it was pretrain... | [
"TAGS\n#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us \n",
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech a... |
feature-extraction | transformers |
# DistilXLSR-53 for BP
[DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese](https://github.com/s3prl/s3prl/tree/master/s3prl/upstream/distiller)
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
**Note**: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model **speech recognition**, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more in-detail explanation of how to fine-tune the model.
Paper: [DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT](https://arxiv.org/abs/2110.01900)
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
**Note 2**: The XLSR-53 model was distilled using [Brazilian Portuguese Datasets](https://huggingface.co/lgris/bp400-xlsr) for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the [original work](https://arxiv.org/abs/2110.01900)).
**Abstract**
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more information on how to fine-tune the model.
| {"language": "pt", "license": "apache-2.0", "tags": ["speech"]} | lgris/distilxlsr_bp_4-12 | null | [
"transformers",
"pytorch",
"wav2vec2",
"feature-extraction",
"speech",
"pt",
"arxiv:2110.01900",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2110.01900"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us
|
# DistilXLSR-53 for BP
DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model.
Paper: DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
Note 2: The XLSR-53 model was distilled using Brazilian Portuguese Datasets for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the original work).
Abstract
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See this blog for more information on how to fine-tune the model.
| [
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.\n\nNote: This model does not have a tokenizer as it was pretrain... | [
"TAGS\n#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us \n",
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech a... |
feature-extraction | transformers |
# DistilXLSR-53 for BP
[DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese](https://github.com/s3prl/s3prl/tree/master/s3prl/upstream/distiller)
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
**Note**: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model **speech recognition**, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more in-detail explanation of how to fine-tune the model.
Paper: [DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT](https://arxiv.org/abs/2110.01900)
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
**Note 2**: The XLSR-53 model was distilled using [Brazilian Portuguese Datasets](https://huggingface.co/lgris/bp400-xlsr) for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the [original work](https://arxiv.org/abs/2110.01900)).
**Abstract**
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more information on how to fine-tune the model.
| {"language": "pt", "license": "apache-2.0", "tags": ["speech"]} | lgris/distilxlsr_bp_8-12-24 | null | [
"transformers",
"pytorch",
"wav2vec2",
"feature-extraction",
"speech",
"pt",
"arxiv:2110.01900",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2110.01900"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us
|
# DistilXLSR-53 for BP
DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model.
Paper: DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
Note 2: The XLSR-53 model was distilled using Brazilian Portuguese Datasets for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the original work).
Abstract
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See this blog for more information on how to fine-tune the model.
| [
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.\n\nNote: This model does not have a tokenizer as it was pretrain... | [
"TAGS\n#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us \n",
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech a... |
feature-extraction | transformers |
# DistilXLSR-53 for BP
[DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese](https://github.com/s3prl/s3prl/tree/master/s3prl/upstream/distiller)
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
**Note**: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model **speech recognition**, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more in-detail explanation of how to fine-tune the model.
Paper: [DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT](https://arxiv.org/abs/2110.01900)
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
**Note 2**: The XLSR-53 model was distilled using [Brazilian Portuguese Datasets](https://huggingface.co/lgris/bp400-xlsr) for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the [original work](https://arxiv.org/abs/2110.01900)).
**Abstract**
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more information on how to fine-tune the model.
| {"language": "pt", "license": "apache-2.0", "tags": ["speech"]} | lgris/distilxlsr_bp_8-12 | null | [
"transformers",
"pytorch",
"wav2vec2",
"feature-extraction",
"speech",
"pt",
"arxiv:2110.01900",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2110.01900"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us
|
# DistilXLSR-53 for BP
DistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model.
Paper: DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT
Authors: Heng-Jui Chang, Shu-wen Yang, Hung-yi Lee
Note 2: The XLSR-53 model was distilled using Brazilian Portuguese Datasets for test purposes. The dataset is quite small to perform such task (the performance might not be so good as the original work).
Abstract
Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.
# Usage
See this blog for more information on how to fine-tune the model.
| [
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.\n\nNote: This model does not have a tokenizer as it was pretrain... | [
"TAGS\n#transformers #pytorch #wav2vec2 #feature-extraction #speech #pt #arxiv-2110.01900 #license-apache-2.0 #endpoints_compatible #region-us \n",
"# DistilXLSR-53 for BP\nDistilXLSR-53 for BP: DistilHuBERT applied to Wav2vec XLSR-53 for Brazilian Portuguese\n\nThe base model pretrained on 16kHz sampled speech a... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# sew-tiny-portuguese-cv
This model is a fine-tuned version of [lgris/sew-tiny-pt](https://huggingface.co/lgris/sew-tiny-pt) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5110
- Wer: 0.2842
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- training_steps: 40000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:------:|:-----:|:---------------:|:------:|
| No log | 4.92 | 1000 | 0.8468 | 0.6494 |
| 3.4638 | 9.85 | 2000 | 0.4978 | 0.3815 |
| 3.4638 | 14.78 | 3000 | 0.4734 | 0.3417 |
| 0.9904 | 19.7 | 4000 | 0.4577 | 0.3344 |
| 0.9904 | 24.63 | 5000 | 0.4376 | 0.3170 |
| 0.8849 | 29.55 | 6000 | 0.4225 | 0.3118 |
| 0.8849 | 34.48 | 7000 | 0.4354 | 0.3080 |
| 0.819 | 39.41 | 8000 | 0.4434 | 0.3004 |
| 0.819 | 44.33 | 9000 | 0.4710 | 0.3132 |
| 0.7706 | 49.26 | 10000 | 0.4497 | 0.3064 |
| 0.7706 | 54.19 | 11000 | 0.4598 | 0.3100 |
| 0.7264 | 59.11 | 12000 | 0.4271 | 0.3013 |
| 0.7264 | 64.04 | 13000 | 0.4333 | 0.2959 |
| 0.6909 | 68.96 | 14000 | 0.4554 | 0.3019 |
| 0.6909 | 73.89 | 15000 | 0.4444 | 0.2888 |
| 0.6614 | 78.81 | 16000 | 0.4734 | 0.3081 |
| 0.6614 | 83.74 | 17000 | 0.4820 | 0.3058 |
| 0.6379 | 88.67 | 18000 | 0.4416 | 0.2950 |
| 0.6379 | 93.59 | 19000 | 0.4614 | 0.2974 |
| 0.6055 | 98.52 | 20000 | 0.4812 | 0.3018 |
| 0.6055 | 103.45 | 21000 | 0.4700 | 0.3018 |
| 0.5823 | 108.37 | 22000 | 0.4726 | 0.2999 |
| 0.5823 | 113.3 | 23000 | 0.4979 | 0.2887 |
| 0.5597 | 118.23 | 24000 | 0.4813 | 0.2980 |
| 0.5597 | 123.15 | 25000 | 0.4968 | 0.2972 |
| 0.542 | 128.08 | 26000 | 0.5331 | 0.3059 |
| 0.542 | 133.0 | 27000 | 0.5046 | 0.2978 |
| 0.5185 | 137.93 | 28000 | 0.4882 | 0.2922 |
| 0.5185 | 142.85 | 29000 | 0.4945 | 0.2938 |
| 0.499 | 147.78 | 30000 | 0.4971 | 0.2913 |
| 0.499 | 152.71 | 31000 | 0.4948 | 0.2873 |
| 0.4811 | 157.63 | 32000 | 0.4924 | 0.2918 |
| 0.4811 | 162.56 | 33000 | 0.5128 | 0.2911 |
| 0.4679 | 167.49 | 34000 | 0.5098 | 0.2892 |
| 0.4679 | 172.41 | 35000 | 0.4966 | 0.2863 |
| 0.456 | 177.34 | 36000 | 0.5033 | 0.2839 |
| 0.456 | 182.27 | 37000 | 0.5114 | 0.2875 |
| 0.4453 | 187.19 | 38000 | 0.5154 | 0.2859 |
| 0.4453 | 192.12 | 39000 | 0.5102 | 0.2847 |
| 0.4366 | 197.04 | 40000 | 0.5110 | 0.2842 |
### Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.17.1.dev0
- Tokenizers 0.11.0
| {"language": ["pt"], "license": "apache-2.0", "tags": ["generated_from_trainer", "hf-asr-leaderboard", "pt", "robust-speech-event"], "datasets": ["common_voice"], "model-index": [{"name": "sew-tiny-portuguese-cv", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 6", "type": "common_voice", "args": "pt"}, "metrics": [{"type": "wer", "value": 30.02, "name": "Test WER"}, {"type": "cer", "value": 10.34, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 56.46, "name": "Test WER"}, {"type": "cer", "value": 22.94, "name": "Test CER"}, {"type": "wer", "value": 57.17, "name": "Test WER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Test Data", "type": "speech-recognition-community-v2/eval_data", "args": "pt"}, "metrics": [{"type": "wer", "value": 61.3, "name": "Test WER"}]}]}]} | lgris/sew-tiny-portuguese-cv | null | [
"transformers",
"pytorch",
"sew",
"automatic-speech-recognition",
"generated_from_trainer",
"hf-asr-leaderboard",
"pt",
"robust-speech-event",
"dataset:common_voice",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #sew #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #pt #robust-speech-event #dataset-common_voice #license-apache-2.0 #model-index #endpoints_compatible #region-us
| sew-tiny-portuguese-cv
======================
This model is a fine-tuned version of lgris/sew-tiny-pt on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.5110
* Wer: 0.2842
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 4
* total\_train\_batch\_size: 32
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 1000
* training\_steps: 40000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.16.0.dev0
* Pytorch 1.10.1+cu102
* Datasets 1.17.1.dev0
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 32\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #sew #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #pt #robust-speech-event #dataset-common_voice #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were used during train... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# sew-tiny-portuguese-cv7
This model is a fine-tuned version of [lgris/sew-tiny-pt](https://huggingface.co/lgris/sew-tiny-pt) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4232
- Wer: 0.2745
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- training_steps: 40000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:------:|:-----:|:---------------:|:------:|
| No log | 2.6 | 1000 | 1.0034 | 0.7308 |
| 4.1307 | 5.19 | 2000 | 0.6274 | 0.4721 |
| 4.1307 | 7.79 | 3000 | 0.5541 | 0.4130 |
| 1.3117 | 10.39 | 4000 | 0.5302 | 0.3880 |
| 1.3117 | 12.99 | 5000 | 0.5082 | 0.3644 |
| 1.2047 | 15.58 | 6000 | 0.4818 | 0.3539 |
| 1.2047 | 18.18 | 7000 | 0.4822 | 0.3477 |
| 1.14 | 20.78 | 8000 | 0.4781 | 0.3428 |
| 1.14 | 23.38 | 9000 | 0.4840 | 0.3401 |
| 1.0818 | 25.97 | 10000 | 0.4613 | 0.3251 |
| 1.0818 | 28.57 | 11000 | 0.4569 | 0.3257 |
| 1.0451 | 31.17 | 12000 | 0.4494 | 0.3132 |
| 1.0451 | 33.77 | 13000 | 0.4560 | 0.3201 |
| 1.011 | 36.36 | 14000 | 0.4687 | 0.3174 |
| 1.011 | 38.96 | 15000 | 0.4397 | 0.3122 |
| 0.9785 | 41.56 | 16000 | 0.4605 | 0.3173 |
| 0.9785 | 44.16 | 17000 | 0.4380 | 0.3064 |
| 0.9458 | 46.75 | 18000 | 0.4372 | 0.3048 |
| 0.9458 | 49.35 | 19000 | 0.4426 | 0.3039 |
| 0.9126 | 51.95 | 20000 | 0.4317 | 0.2962 |
| 0.9126 | 54.54 | 21000 | 0.4345 | 0.2960 |
| 0.8926 | 57.14 | 22000 | 0.4365 | 0.2948 |
| 0.8926 | 59.74 | 23000 | 0.4306 | 0.2940 |
| 0.8654 | 62.34 | 24000 | 0.4303 | 0.2928 |
| 0.8654 | 64.93 | 25000 | 0.4351 | 0.2915 |
| 0.8373 | 67.53 | 26000 | 0.4340 | 0.2909 |
| 0.8373 | 70.13 | 27000 | 0.4279 | 0.2907 |
| 0.83 | 72.73 | 28000 | 0.4214 | 0.2867 |
| 0.83 | 75.32 | 29000 | 0.4256 | 0.2849 |
| 0.8062 | 77.92 | 30000 | 0.4281 | 0.2826 |
| 0.8062 | 80.52 | 31000 | 0.4398 | 0.2865 |
| 0.7846 | 83.12 | 32000 | 0.4218 | 0.2812 |
| 0.7846 | 85.71 | 33000 | 0.4227 | 0.2791 |
| 0.7697 | 88.31 | 34000 | 0.4200 | 0.2767 |
| 0.7697 | 90.91 | 35000 | 0.4285 | 0.2791 |
| 0.7539 | 93.51 | 36000 | 0.4238 | 0.2777 |
| 0.7539 | 96.1 | 37000 | 0.4288 | 0.2757 |
| 0.7413 | 98.7 | 38000 | 0.4205 | 0.2748 |
| 0.7413 | 101.3 | 39000 | 0.4241 | 0.2761 |
| 0.7348 | 103.89 | 40000 | 0.4232 | 0.2745 |
### Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.17.1.dev0
- Tokenizers 0.11.0
| {"language": ["pt"], "license": "apache-2.0", "tags": ["generated_from_trainer", "hf-asr-leaderboard", "pt", "robust-speech-event"], "datasets": ["mozilla-foundation/common_voice_7_0"], "model-index": [{"name": "sew-tiny-portuguese-cv7", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 7", "type": "mozilla-foundation/common_voice_7_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 28.9, "name": "Test WER"}, {"type": "cer", "value": 9.41, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 47.27, "name": "Test WER"}, {"type": "cer", "value": 19.62, "name": "Test CER"}, {"type": "wer", "value": 47.3, "name": "Test WER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Test Data", "type": "speech-recognition-community-v2/eval_data", "args": "pt"}, "metrics": [{"type": "wer", "value": 49.83, "name": "Test WER"}]}]}]} | lgris/sew-tiny-portuguese-cv7 | null | [
"transformers",
"pytorch",
"tensorboard",
"sew",
"automatic-speech-recognition",
"generated_from_trainer",
"hf-asr-leaderboard",
"pt",
"robust-speech-event",
"dataset:mozilla-foundation/common_voice_7_0",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #tensorboard #sew #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| sew-tiny-portuguese-cv7
=======================
This model is a fine-tuned version of lgris/sew-tiny-pt on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.4232
* Wer: 0.2745
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 16
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 32
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 1000
* training\_steps: 40000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.16.0.dev0
* Pytorch 1.10.1+cu102
* Datasets 1.17.1.dev0
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 32\n* optimizer: Adam with betas=(0.9,0.999) and epsilon... | [
"TAGS\n#transformers #pytorch #tensorboard #sew #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hy... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# sew-tiny-portuguese-cv8
This model is a fine-tuned version of [lgris/sew-tiny-pt](https://huggingface.co/lgris/sew-tiny-pt) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4082
- Wer: 0.3053
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- training_steps: 40000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| No log | 1.93 | 1000 | 2.9134 | 0.9767 |
| 2.9224 | 3.86 | 2000 | 2.8405 | 0.9789 |
| 2.9224 | 5.79 | 3000 | 2.8094 | 0.9800 |
| 2.8531 | 7.72 | 4000 | 2.7439 | 0.9891 |
| 2.8531 | 9.65 | 5000 | 2.7057 | 1.0159 |
| 2.7721 | 11.58 | 6000 | 2.7235 | 1.0709 |
| 2.7721 | 13.51 | 7000 | 2.5931 | 1.1035 |
| 2.6566 | 15.44 | 8000 | 2.2171 | 0.9884 |
| 2.6566 | 17.37 | 9000 | 1.2399 | 0.8081 |
| 1.9558 | 19.31 | 10000 | 0.9045 | 0.6353 |
| 1.9558 | 21.24 | 11000 | 0.7705 | 0.5533 |
| 1.4987 | 23.17 | 12000 | 0.7068 | 0.5165 |
| 1.4987 | 25.1 | 13000 | 0.6641 | 0.4718 |
| 1.3811 | 27.03 | 14000 | 0.6043 | 0.4470 |
| 1.3811 | 28.96 | 15000 | 0.5532 | 0.4268 |
| 1.2897 | 30.89 | 16000 | 0.5371 | 0.4101 |
| 1.2897 | 32.82 | 17000 | 0.5924 | 0.4150 |
| 1.225 | 34.75 | 18000 | 0.4949 | 0.3894 |
| 1.225 | 36.68 | 19000 | 0.5591 | 0.4045 |
| 1.193 | 38.61 | 20000 | 0.4927 | 0.3731 |
| 1.193 | 40.54 | 21000 | 0.4922 | 0.3712 |
| 1.1482 | 42.47 | 22000 | 0.4799 | 0.3662 |
| 1.1482 | 44.4 | 23000 | 0.4846 | 0.3648 |
| 1.1201 | 46.33 | 24000 | 0.4770 | 0.3623 |
| 1.1201 | 48.26 | 25000 | 0.4530 | 0.3426 |
| 1.0892 | 50.19 | 26000 | 0.4523 | 0.3527 |
| 1.0892 | 52.12 | 27000 | 0.4573 | 0.3443 |
| 1.0583 | 54.05 | 28000 | 0.4488 | 0.3353 |
| 1.0583 | 55.98 | 29000 | 0.4295 | 0.3285 |
| 1.0319 | 57.92 | 30000 | 0.4321 | 0.3220 |
| 1.0319 | 59.85 | 31000 | 0.4244 | 0.3236 |
| 1.0076 | 61.78 | 32000 | 0.4197 | 0.3201 |
| 1.0076 | 63.71 | 33000 | 0.4230 | 0.3208 |
| 0.9851 | 65.64 | 34000 | 0.4090 | 0.3127 |
| 0.9851 | 67.57 | 35000 | 0.4088 | 0.3133 |
| 0.9695 | 69.5 | 36000 | 0.4123 | 0.3088 |
| 0.9695 | 71.43 | 37000 | 0.4017 | 0.3090 |
| 0.9514 | 73.36 | 38000 | 0.4184 | 0.3086 |
| 0.9514 | 75.29 | 39000 | 0.4075 | 0.3043 |
| 0.944 | 77.22 | 40000 | 0.4082 | 0.3053 |
### Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.17.1.dev0
- Tokenizers 0.11.0
| {"language": ["pt"], "license": "apache-2.0", "tags": ["generated_from_trainer", "hf-asr-leaderboard", "pt", "robust-speech-event"], "datasets": ["mozilla-foundation/common_voice_8_0"], "model-index": [{"name": "sew-tiny-portuguese-cv8", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 8", "type": "mozilla-foundation/common_voice_8_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 33.71, "name": "Test WER"}, {"type": "cer", "value": 10.69, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 52.79, "name": "Test WER"}, {"type": "cer", "value": 20.98, "name": "Test CER"}, {"type": "wer", "value": 53.18, "name": "Test WER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Test Data", "type": "speech-recognition-community-v2/eval_data", "args": "pt"}, "metrics": [{"type": "wer", "value": 55.23, "name": "Test WER"}]}]}]} | lgris/sew-tiny-portuguese-cv8 | null | [
"transformers",
"pytorch",
"tensorboard",
"sew",
"automatic-speech-recognition",
"generated_from_trainer",
"hf-asr-leaderboard",
"pt",
"robust-speech-event",
"dataset:mozilla-foundation/common_voice_8_0",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #tensorboard #sew #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| sew-tiny-portuguese-cv8
=======================
This model is a fine-tuned version of lgris/sew-tiny-pt on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.4082
* Wer: 0.3053
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 4
* total\_train\_batch\_size: 32
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 1000
* training\_steps: 40000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.16.0.dev0
* Pytorch 1.10.1+cu102
* Datasets 1.17.1.dev0
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 32\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #sew #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hy... |
feature-extraction | transformers |
# SEW-tiny-pt
This is a pretrained version of [SEW tiny by ASAPP Research](https://github.com/asappresearch/sew) trained over Brazilian Portuguese audio.
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz. Note that this model should be fine-tuned on a downstream task, like Automatic Speech Recognition, Speaker Identification, Intent Classification, Emotion Recognition, etc...
Paper: [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870)
Authors: Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi
**Abstract**
This paper is a study of performance-efficiency trade-offs in pre-trained models for automatic speech recognition (ASR). We focus on wav2vec 2.0, and formalize several architecture designs that influence both the model performance and its efficiency. Putting together all our observations, we introduce SEW (Squeezed and Efficient Wav2vec), a pre-trained model architecture with significant improvements along both performance and efficiency dimensions across a variety of training setups. For example, under the 100h-960h semi-supervised setup on LibriSpeech, SEW achieves a 1.9x inference speedup compared to wav2vec 2.0, with a 13.5% relative reduction in word error rate. With a similar inference time, SEW reduces word error rate by 25-50% across different model sizes.
The original model can be found under https://github.com/asappresearch/sew#model-checkpoints .
# Usage
See [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more information on how to fine-tune the model. Note that the class `Wav2Vec2ForCTC` has to be replaced by `SEWForCTC`.
| {"language": "pt", "license": "apache-2.0", "tags": ["speech"]} | lgris/sew-tiny-pt | null | [
"transformers",
"pytorch",
"sew",
"feature-extraction",
"speech",
"pt",
"arxiv:2109.06870",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2109.06870"
] | [
"pt"
] | TAGS
#transformers #pytorch #sew #feature-extraction #speech #pt #arxiv-2109.06870 #license-apache-2.0 #endpoints_compatible #region-us
|
# SEW-tiny-pt
This is a pretrained version of SEW tiny by ASAPP Research trained over Brazilian Portuguese audio.
The base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz. Note that this model should be fine-tuned on a downstream task, like Automatic Speech Recognition, Speaker Identification, Intent Classification, Emotion Recognition, etc...
Paper: Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition
Authors: Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi
Abstract
This paper is a study of performance-efficiency trade-offs in pre-trained models for automatic speech recognition (ASR). We focus on wav2vec 2.0, and formalize several architecture designs that influence both the model performance and its efficiency. Putting together all our observations, we introduce SEW (Squeezed and Efficient Wav2vec), a pre-trained model architecture with significant improvements along both performance and efficiency dimensions across a variety of training setups. For example, under the 100h-960h semi-supervised setup on LibriSpeech, SEW achieves a 1.9x inference speedup compared to wav2vec 2.0, with a 13.5% relative reduction in word error rate. With a similar inference time, SEW reduces word error rate by 25-50% across different model sizes.
The original model can be found under URL .
# Usage
See this blog for more information on how to fine-tune the model. Note that the class 'Wav2Vec2ForCTC' has to be replaced by 'SEWForCTC'.
| [
"# SEW-tiny-pt\n\nThis is a pretrained version of SEW tiny by ASAPP Research trained over Brazilian Portuguese audio.\n\nThe base model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz. Note that this model should be fine-tuned on a downstream ... | [
"TAGS\n#transformers #pytorch #sew #feature-extraction #speech #pt #arxiv-2109.06870 #license-apache-2.0 #endpoints_compatible #region-us \n",
"# SEW-tiny-pt\n\nThis is a pretrained version of SEW tiny by ASAPP Research trained over Brazilian Portuguese audio.\n\nThe base model pretrained on 16kHz sampled speech ... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-large-xls-r-300m-pt-cv
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.3418
- Wer: 0.3581
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 5
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 10.9035 | 0.2 | 100 | 4.2750 | 1.0 |
| 3.3275 | 0.41 | 200 | 3.0334 | 1.0 |
| 3.0016 | 0.61 | 300 | 2.9494 | 1.0 |
| 2.1874 | 0.82 | 400 | 1.4355 | 0.8721 |
| 1.09 | 1.02 | 500 | 0.9987 | 0.7165 |
| 0.8251 | 1.22 | 600 | 0.7886 | 0.6406 |
| 0.6927 | 1.43 | 700 | 0.6753 | 0.5801 |
| 0.6143 | 1.63 | 800 | 0.6300 | 0.5509 |
| 0.5451 | 1.84 | 900 | 0.5586 | 0.5156 |
| 0.5003 | 2.04 | 1000 | 0.5493 | 0.5027 |
| 0.3712 | 2.24 | 1100 | 0.5271 | 0.4872 |
| 0.3486 | 2.45 | 1200 | 0.4953 | 0.4817 |
| 0.3498 | 2.65 | 1300 | 0.4619 | 0.4538 |
| 0.3112 | 2.86 | 1400 | 0.4570 | 0.4387 |
| 0.3013 | 3.06 | 1500 | 0.4437 | 0.4147 |
| 0.2136 | 3.27 | 1600 | 0.4176 | 0.4124 |
| 0.2131 | 3.47 | 1700 | 0.4281 | 0.4194 |
| 0.2099 | 3.67 | 1800 | 0.3864 | 0.3949 |
| 0.1925 | 3.88 | 1900 | 0.3926 | 0.3913 |
| 0.1709 | 4.08 | 2000 | 0.3764 | 0.3804 |
| 0.1406 | 4.29 | 2100 | 0.3787 | 0.3742 |
| 0.1342 | 4.49 | 2200 | 0.3645 | 0.3693 |
| 0.1305 | 4.69 | 2300 | 0.3463 | 0.3625 |
| 0.1298 | 4.9 | 2400 | 0.3418 | 0.3581 |
### Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.13.3
- Tokenizers 0.10.3
| {"language": ["pt"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "robust-speech-event", "pt", "hf-asr-leaderboard"], "datasets": ["common_voice"], "model-index": [{"name": "wav2vec2-large-xls-r-300m-pt-cv", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 6", "type": "common_voice", "args": "pt"}, "metrics": [{"type": "wer", "value": 24.29, "name": "Test WER"}, {"type": "cer", "value": 7.51, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 55.72, "name": "Test WER"}, {"type": "cer", "value": 21.82, "name": "Test CER"}, {"type": "wer", "value": 47.88, "name": "Test WER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Test Data", "type": "speech-recognition-community-v2/eval_data", "args": "pt"}, "metrics": [{"type": "wer", "value": 50.78, "name": "Test WER"}]}]}]} | lgris/wav2vec2-large-xls-r-300m-pt-cv | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"robust-speech-event",
"pt",
"hf-asr-leaderboard",
"dataset:common_voice",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #robust-speech-event #pt #hf-asr-leaderboard #dataset-common_voice #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-large-xls-r-300m-pt-cv
===============================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.3418
* Wer: 0.3581
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0003
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 500
* num\_epochs: 5
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.11.3
* Pytorch 1.10.0+cu111
* Datasets 1.13.3
* Tokenizers 0.10.3
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0003\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #robust-speech-event #pt #hf-asr-leaderboard #dataset-common_voice #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-large-xlsr-coraa-portuguese-cv7
This model is a fine-tuned version of [Edresson/wav2vec2-large-xlsr-coraa-portuguese](https://huggingface.co/Edresson/wav2vec2-large-xlsr-coraa-portuguese) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1777
- Wer: 0.1339
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 5000
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 0.4779 | 0.13 | 100 | 0.2620 | 0.2020 |
| 0.4505 | 0.26 | 200 | 0.2339 | 0.1998 |
| 0.4285 | 0.39 | 300 | 0.2507 | 0.2109 |
| 0.4148 | 0.52 | 400 | 0.2311 | 0.2101 |
| 0.4072 | 0.65 | 500 | 0.2278 | 0.1899 |
| 0.388 | 0.78 | 600 | 0.2193 | 0.1898 |
| 0.3952 | 0.91 | 700 | 0.2108 | 0.1901 |
| 0.3851 | 1.04 | 800 | 0.2121 | 0.1788 |
| 0.3496 | 1.17 | 900 | 0.2154 | 0.1776 |
| 0.3063 | 1.3 | 1000 | 0.2095 | 0.1730 |
| 0.3376 | 1.43 | 1100 | 0.2129 | 0.1801 |
| 0.3273 | 1.56 | 1200 | 0.2132 | 0.1776 |
| 0.3347 | 1.69 | 1300 | 0.2054 | 0.1698 |
| 0.323 | 1.82 | 1400 | 0.1986 | 0.1724 |
| 0.3079 | 1.95 | 1500 | 0.2005 | 0.1701 |
| 0.3029 | 2.08 | 1600 | 0.2159 | 0.1644 |
| 0.2694 | 2.21 | 1700 | 0.1992 | 0.1678 |
| 0.2733 | 2.34 | 1800 | 0.2032 | 0.1657 |
| 0.269 | 2.47 | 1900 | 0.2056 | 0.1592 |
| 0.2869 | 2.6 | 2000 | 0.2058 | 0.1616 |
| 0.2813 | 2.73 | 2100 | 0.1868 | 0.1584 |
| 0.2616 | 2.86 | 2200 | 0.1841 | 0.1550 |
| 0.2809 | 2.99 | 2300 | 0.1902 | 0.1577 |
| 0.2598 | 3.12 | 2400 | 0.1910 | 0.1514 |
| 0.24 | 3.25 | 2500 | 0.1971 | 0.1555 |
| 0.2481 | 3.38 | 2600 | 0.1853 | 0.1537 |
| 0.2437 | 3.51 | 2700 | 0.1897 | 0.1496 |
| 0.2384 | 3.64 | 2800 | 0.1842 | 0.1495 |
| 0.2405 | 3.77 | 2900 | 0.1884 | 0.1500 |
| 0.2372 | 3.9 | 3000 | 0.1950 | 0.1548 |
| 0.229 | 4.03 | 3100 | 0.1928 | 0.1477 |
| 0.2047 | 4.16 | 3200 | 0.1891 | 0.1472 |
| 0.2102 | 4.29 | 3300 | 0.1930 | 0.1473 |
| 0.199 | 4.42 | 3400 | 0.1914 | 0.1456 |
| 0.2121 | 4.55 | 3500 | 0.1840 | 0.1437 |
| 0.211 | 4.67 | 3600 | 0.1843 | 0.1403 |
| 0.2072 | 4.8 | 3700 | 0.1836 | 0.1428 |
| 0.2224 | 4.93 | 3800 | 0.1747 | 0.1412 |
| 0.1974 | 5.06 | 3900 | 0.1813 | 0.1416 |
| 0.1895 | 5.19 | 4000 | 0.1869 | 0.1406 |
| 0.1763 | 5.32 | 4100 | 0.1830 | 0.1394 |
| 0.2001 | 5.45 | 4200 | 0.1775 | 0.1394 |
| 0.1909 | 5.58 | 4300 | 0.1806 | 0.1373 |
| 0.1812 | 5.71 | 4400 | 0.1784 | 0.1359 |
| 0.1737 | 5.84 | 4500 | 0.1778 | 0.1353 |
| 0.1915 | 5.97 | 4600 | 0.1777 | 0.1349 |
| 0.1921 | 6.1 | 4700 | 0.1784 | 0.1359 |
| 0.1805 | 6.23 | 4800 | 0.1757 | 0.1348 |
| 0.1742 | 6.36 | 4900 | 0.1771 | 0.1341 |
| 0.1709 | 6.49 | 5000 | 0.1777 | 0.1339 |
### Framework versions
- Transformers 4.16.1
- Pytorch 1.10.0+cu111
- Datasets 1.18.2
- Tokenizers 0.11.0
| {"license": "apache-2.0", "tags": ["generated_from_trainer", "pt", "robust-speech-event"], "datasets": ["mozilla-foundation/common_voice_7_0"], "model-index": [{"name": "wav2vec2-large-xlsr-coraa-portuguese-cv7", "results": []}]} | lgris/wav2vec2-large-xlsr-coraa-portuguese-cv7 | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"pt",
"robust-speech-event",
"dataset:mozilla-foundation/common_voice_7_0",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #endpoints_compatible #region-us
| wav2vec2-large-xlsr-coraa-portuguese-cv7
========================================
This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.1777
* Wer: 0.1339
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 5000
### Training results
### Framework versions
* Transformers 4.16.1
* Pytorch 1.10.0+cu111
* Datasets 1.18.2
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were used duri... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-large-xlsr-coraa-portuguese-cv8
This model is a fine-tuned version of [Edresson/wav2vec2-large-xlsr-coraa-portuguese](https://huggingface.co/Edresson/wav2vec2-large-xlsr-coraa-portuguese) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1626
- Wer: 0.1365
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 5000
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 0.5614 | 0.1 | 100 | 0.2542 | 0.1986 |
| 0.5181 | 0.19 | 200 | 0.2740 | 0.2146 |
| 0.5056 | 0.29 | 300 | 0.2472 | 0.2068 |
| 0.4747 | 0.39 | 400 | 0.2464 | 0.2166 |
| 0.4627 | 0.48 | 500 | 0.2277 | 0.2041 |
| 0.4403 | 0.58 | 600 | 0.2245 | 0.1977 |
| 0.4413 | 0.68 | 700 | 0.2156 | 0.1968 |
| 0.437 | 0.77 | 800 | 0.2102 | 0.1919 |
| 0.4305 | 0.87 | 900 | 0.2130 | 0.1864 |
| 0.4324 | 0.97 | 1000 | 0.2144 | 0.1902 |
| 0.4217 | 1.06 | 1100 | 0.2230 | 0.1891 |
| 0.3823 | 1.16 | 1200 | 0.2033 | 0.1774 |
| 0.3641 | 1.25 | 1300 | 0.2143 | 0.1830 |
| 0.3707 | 1.35 | 1400 | 0.2034 | 0.1793 |
| 0.3767 | 1.45 | 1500 | 0.2029 | 0.1823 |
| 0.3483 | 1.54 | 1600 | 0.1999 | 0.1740 |
| 0.3577 | 1.64 | 1700 | 0.1928 | 0.1728 |
| 0.3667 | 1.74 | 1800 | 0.1898 | 0.1726 |
| 0.3283 | 1.83 | 1900 | 0.1920 | 0.1688 |
| 0.3571 | 1.93 | 2000 | 0.1904 | 0.1649 |
| 0.3467 | 2.03 | 2100 | 0.1994 | 0.1648 |
| 0.3145 | 2.12 | 2200 | 0.1940 | 0.1682 |
| 0.3186 | 2.22 | 2300 | 0.1879 | 0.1571 |
| 0.3058 | 2.32 | 2400 | 0.1975 | 0.1678 |
| 0.3096 | 2.41 | 2500 | 0.1877 | 0.1589 |
| 0.2964 | 2.51 | 2600 | 0.1862 | 0.1568 |
| 0.3068 | 2.61 | 2700 | 0.1809 | 0.1588 |
| 0.3036 | 2.7 | 2800 | 0.1769 | 0.1573 |
| 0.3084 | 2.8 | 2900 | 0.1836 | 0.1524 |
| 0.3109 | 2.9 | 3000 | 0.1807 | 0.1519 |
| 0.2969 | 2.99 | 3100 | 0.1851 | 0.1516 |
| 0.2698 | 3.09 | 3200 | 0.1737 | 0.1490 |
| 0.2703 | 3.19 | 3300 | 0.1759 | 0.1457 |
| 0.2759 | 3.28 | 3400 | 0.1778 | 0.1471 |
| 0.2728 | 3.38 | 3500 | 0.1717 | 0.1462 |
| 0.2398 | 3.47 | 3600 | 0.1767 | 0.1451 |
| 0.256 | 3.57 | 3700 | 0.1742 | 0.1410 |
| 0.2712 | 3.67 | 3800 | 0.1674 | 0.1414 |
| 0.2648 | 3.76 | 3900 | 0.1717 | 0.1423 |
| 0.2576 | 3.86 | 4000 | 0.1672 | 0.1403 |
| 0.2504 | 3.96 | 4100 | 0.1683 | 0.1381 |
| 0.2406 | 4.05 | 4200 | 0.1685 | 0.1399 |
| 0.2403 | 4.15 | 4300 | 0.1656 | 0.1381 |
| 0.2233 | 4.25 | 4400 | 0.1687 | 0.1371 |
| 0.2546 | 4.34 | 4500 | 0.1642 | 0.1377 |
| 0.2431 | 4.44 | 4600 | 0.1655 | 0.1372 |
| 0.2337 | 4.54 | 4700 | 0.1625 | 0.1370 |
| 0.2607 | 4.63 | 4800 | 0.1618 | 0.1363 |
| 0.2292 | 4.73 | 4900 | 0.1622 | 0.1366 |
| 0.2232 | 4.83 | 5000 | 0.1626 | 0.1365 |
### Framework versions
- Transformers 4.16.2
- Pytorch 1.10.0+cu111
- Datasets 1.18.2
- Tokenizers 0.11.0
| {"license": "apache-2.0", "tags": ["generated_from_trainer"], "datasets": ["mozilla-foundation/common_voice_8_0"], "model-index": [{"name": "wav2vec2-large-xlsr-coraa-portuguese-cv8", "results": []}]} | lgris/wav2vec2-large-xlsr-coraa-portuguese-cv8 | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"dataset:mozilla-foundation/common_voice_8_0",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #endpoints_compatible #region-us
| wav2vec2-large-xlsr-coraa-portuguese-cv8
========================================
This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.1626
* Wer: 0.1365
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 5000
### Training results
### Framework versions
* Transformers 4.16.2
* Pytorch 1.10.0+cu111
* Datasets 1.18.2
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learn... |
automatic-speech-recognition | transformers |
# Wav2vec 2.0 With Open Brazilian Portuguese Datasets v2
This a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
- [CETUC](http://www02.smt.ufrj.br/~igor.quintanilha/alcaim.tar.gz): contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the [CETEN-Folha](https://www.linguateca.pt/cetenfolha/) corpus.
- [Multilingual Librispeech (MLS)](https://arxiv.org/abs/2012.03411): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like [LibriVox](https://librivox.org/). The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese [used in this work](http://www.openslr.org/94/) (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
- [VoxForge](http://www.voxforge.org/): is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
- [Common Voice 6.1](https://commonvoice.mozilla.org/pt): is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages to train ASR models. In this project, volunteers donate and validate speech using the [oficial site](https://commonvoice.mozilla.org/pt). The set in Portuguese (mostly Brazilian variant) used in this work is the 6.1 version (pt_63h_2020-12-11) that contains about 50 validated hours and 1,120 unique speakers.
- [Lapsbm](https://github.com/falabrasil/gitlab-resources): "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively.
The original model was fine-tuned using [fairseq](https://github.com/pytorch/fairseq). This notebook uses a converted version of the original one.
__NOTE: The common voice test reports 10% of WER, however, this model was trained using all the validated instances of Common Voice, except the instances of the test set. This means that some speakers of the train set can be present on the test set.__
## Imports and dependencies
```python
%%capture
!pip install datasets
!pip install jiwer
!pip install torchaudio
!pip install transformers
!pip install soundfile
```
```python
import torchaudio
from datasets import load_dataset, load_metric
from transformers import (
Wav2Vec2ForCTC,
Wav2Vec2Processor,
)
import torch
import re
import sys
```
## Preparation
```python
chars_to_ignore_regex = '[\,\?\.\!\;\:\"]' # noqa: W605
wer = load_metric("wer")
device = "cuda"
```
```python
model_name = 'lgris/wav2vec2-large-xlsr-open-brazilian-portuguese-v2'
model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
processor = Wav2Vec2Processor.from_pretrained(model_name)
```
```python
def map_to_pred(batch):
features = processor(batch["speech"], sampling_rate=batch["sampling_rate"][0], padding=True, return_tensors="pt")
input_values = features.input_values.to(device)
attention_mask = features.attention_mask.to(device)
with torch.no_grad():
logits = model(input_values, attention_mask=attention_mask).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["predicted"] = processor.batch_decode(pred_ids)
batch["predicted"] = [pred.lower() for pred in batch["predicted"]]
batch["target"] = batch["sentence"]
return batch
```
## Tests
### Test against Common Voice (In-domain)
```python
dataset = load_dataset("common_voice", "pt", split="test", data_dir="./cv-corpus-6.1-2020-12-11")
resampler = torchaudio.transforms.Resample(orig_freq=48_000, new_freq=16_000)
def map_to_array(batch):
speech, _ = torchaudio.load(batch["path"])
batch["speech"] = resampler.forward(speech.squeeze(0)).numpy()
batch["sampling_rate"] = resampler.new_freq
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
return batch
```
```python
ds = dataset.map(map_to_array)
result = ds.map(map_to_pred, batched=True, batch_size=1, remove_columns=list(ds.features.keys()))
print(wer.compute(predictions=result["predicted"], references=result["target"]))
for pred, target in zip(result["predicted"][:10], result["target"][:10]):
print(pred, "|", target)
```
**Result**: 10.69%
### Test against [TEDx](http://www.openslr.org/100/) (Out-of-domain)
```python
!gdown --id 1HJEnvthaGYwcV_whHEywgH2daIN4bQna
!tar -xf tedx.tar.gz
```
```python
dataset = load_dataset('csv', data_files={'test': 'test.csv'})['test']
def map_to_array(batch):
speech, _ = torchaudio.load(batch["path"])
batch["speech"] = speech.squeeze(0).numpy()
batch["sampling_rate"] = resampler.new_freq
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
return batch
```
```python
ds = dataset.map(map_to_array)
result = ds.map(map_to_pred, batched=True, batch_size=1, remove_columns=list(ds.features.keys()))
print(wer.compute(predictions=result["predicted"], references=result["target"]))
for pred, target in zip(result["predicted"][:10], result["target"][:10]):
print(pred, "|", target)
```
**Result**: 34.53% | {"language": "pt", "license": "apache-2.0", "tags": ["audio", "speech", "wav2vec2", "pt", "portuguese-speech-corpus", "automatic-speech-recognition", "speech", "PyTorch", "hf-asr-leaderboard"], "datasets": ["common_voice", "mls", "cetuc", "lapsbm", "voxforge"], "metrics": ["wer"], "model-index": [{"name": "wav2vec2-large-xlsr-open-brazilian-portuguese-v2", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice", "type": "common_voice", "args": "pt"}, "metrics": [{"type": "wer", "value": 10.69, "name": "Test WER"}]}]}]} | lgris/wav2vec2-large-xlsr-open-brazilian-portuguese-v2 | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"audio",
"speech",
"pt",
"portuguese-speech-corpus",
"PyTorch",
"hf-asr-leaderboard",
"dataset:common_voice",
"dataset:mls",
"dataset:cetuc",
"dataset:lapsbm",
"dataset:voxforge",
"arxiv:2012.03411",
"license:apac... | null | 2022-03-02T23:29:05+00:00 | [
"2012.03411"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints_compatible #region-us
|
# Wav2vec 2.0 With Open Brazilian Portuguese Datasets v2
This a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
- CETUC: contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the CETEN-Folha corpus.
- Multilingual Librispeech (MLS): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like LibriVox. The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese used in this work (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
- VoxForge: is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
- Common Voice 6.1: is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages to train ASR models. In this project, volunteers donate and validate speech using the oficial site. The set in Portuguese (mostly Brazilian variant) used in this work is the 6.1 version (pt_63h_2020-12-11) that contains about 50 validated hours and 1,120 unique speakers.
- Lapsbm: "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively.
The original model was fine-tuned using fairseq. This notebook uses a converted version of the original one.
__NOTE: The common voice test reports 10% of WER, however, this model was trained using all the validated instances of Common Voice, except the instances of the test set. This means that some speakers of the train set can be present on the test set.__
## Imports and dependencies
## Preparation
## Tests
### Test against Common Voice (In-domain)
Result: 10.69%
### Test against TEDx (Out-of-domain)
Result: 34.53% | [
"# Wav2vec 2.0 With Open Brazilian Portuguese Datasets v2\n\nThis a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:\n\n- CETUC: contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing ap... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",... |
automatic-speech-recognition | transformers |
# Wav2vec 2.0 With Open Brazilian Portuguese Datasets
This a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
- [CETUC](http://www02.smt.ufrj.br/~igor.quintanilha/alcaim.tar.gz): contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the [CETEN-Folha](https://www.linguateca.pt/cetenfolha/) corpus.
- [Multilingual Librispeech (MLS)](https://arxiv.org/abs/2012.03411): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like [LibriVox](https://librivox.org/). The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese [used in this work](http://www.openslr.org/94/) (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
- [VoxForge](http://www.voxforge.org/): is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
- [Common Voice 6.1](https://commonvoice.mozilla.org/pt) (_only train_): is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages to train ASR models. In this project, volunteers donate and validate speech using the [oficial site](https://commonvoice.mozilla.org/pt). The set in Portuguese (mostly Brazilian variant) used in this work is the 6.1 version (pt_63h_2020-12-11) that contains about 50 validated hours and 1,120 unique speakers.
- [Lapsbm](https://github.com/falabrasil/gitlab-resources): "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively.
The original model was fine-tuned using [fairseq](https://github.com/pytorch/fairseq). This notebook uses a converted version of the original one. The link to the original fairseq model is available [here](https://drive.google.com/drive/folders/1XTKIUB4kp3oYOavwH97wq8IPFsxP5sNz?usp=sharing).
This model was trained in 80k updates.
#### Datasets in number of instances and number of frames
The following image shows the overall distribution of the dataset:
#### Transcription examples
| Text | Transcription |
|------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------|
| É comum os usuários confundirem software livre com software livre | É comum os __usuares__ __confunder em__ __softwerlivr__ com __softwerlivre__ |
| Ele fez tanto ghostwriting que ele começa a se sentir como um fantasma também | Ele fez tanto __golstraitn__ que ele __começou__ a se sentir como um fantasma também |
| Arnold apresentou um gráfico mostrando quantas cegonhas ele havia contado nos últimos dez anos | Arnold apresentou um gráfico mostrando quantas __segonhas__ ele havia contado nos últimos dez anos |
| Mais cedo ou mais tarde eles descobrirão como ler esses hieróglifos | Mais __sedo__ ou mais tarde eles descobriram como __de__ esses __ierogrôficos__ |
| Viver juntos compartilhar objetivos e ter um bom relacionamento | __E ver__ juntos __signafica__ viver juntos ou __fartlhar__ objetivos ter um bom __relacionamentoo__ |
| Da mesma forma uma patente pode impedir que concorrentes desenvolvam produtos similares | Da mesma forma uma patente pode impedir que concorrentes __desenvolva__ produtos similares |
| Duas mulheres e uma menina levantam com troféus | Duas mulheres e uma menina levantam com __trofés__ |
| Esse acrobata de circo deve ter um sistema vestibular bem treinado pensou o espectador | Esse acrobata de __cirko__ deve ter um sistema vestibular __bemtreinado__ pensou o espectador |
| Durante a exposição o tribunal pode fazer quaisquer perguntas ou esclarecimentos que considere apropriados | Durante a exposição o tribunal pode fazer quaisquer perguntas ou esclarecimentos que considere __apropriado__ |
## Imports and dependencies
```python
%%capture
!pip install datasets
!pip install jiwer
!pip install torchaudio
!pip install transformers
!pip install soundfile
```
```python
import torchaudio
from datasets import load_dataset, load_metric
from transformers import (
Wav2Vec2ForCTC,
Wav2Vec2Processor,
)
import torch
import re
import sys
```
## Preparation
```python
chars_to_ignore_regex = '[\,\?\.\!\;\:\"]' # noqa: W605
wer = load_metric("wer")
device = "cuda"
```
```python
model_name = 'lgris/wav2vec2-large-xlsr-open-brazilian-portuguese'
model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
processor = Wav2Vec2Processor.from_pretrained(model_name)
```
```python
def map_to_pred(batch):
features = processor(batch["speech"], sampling_rate=batch["sampling_rate"][0], padding=True, return_tensors="pt")
input_values = features.input_values.to(device)
attention_mask = features.attention_mask.to(device)
with torch.no_grad():
logits = model(input_values, attention_mask=attention_mask).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["predicted"] = processor.batch_decode(pred_ids)
batch["predicted"] = [pred.lower() for pred in batch["predicted"]]
batch["target"] = batch["sentence"]
return batch
```
## Tests
### Test against Common Voice (In-domain)
```python
dataset = load_dataset("common_voice", "pt", split="test", data_dir="./cv-corpus-6.1-2020-12-11")
resampler = torchaudio.transforms.Resample(orig_freq=48_000, new_freq=16_000)
def map_to_array(batch):
speech, _ = torchaudio.load(batch["path"])
batch["speech"] = resampler.forward(speech.squeeze(0)).numpy()
batch["sampling_rate"] = resampler.new_freq
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
return batch
```
```python
ds = dataset.map(map_to_array)
result = ds.map(map_to_pred, batched=True, batch_size=1, remove_columns=list(ds.features.keys()))
print(wer.compute(predictions=result["predicted"], references=result["target"]))
for pred, target in zip(result["predicted"][:10], result["target"][:10]):
print(pred, "|", target)
```
0.12905054857823264
nem o varanin os altros influmindo os de teterno um bombederster | nem o radar nem os outros instrumentos detectaram o bombardeiro stealth
pedir dinheiro é emprestado das pessoas do aldeia | pedir dinheiro emprestado às pessoas da aldeia
oito | oito
teno calcos | trancá-los
realizaram a investigação para resolver o problema | realizar uma investigação para resolver o problema
iotube ainda é a melhor plataforma de vídeos | o youtube ainda é a melhor plataforma de vídeos
menina e menino beijando nas sombras | menina e menino beijando nas sombras
eu sou o senhor | eu sou o senhor
duas metcas sentam-se para baixo randes jornais | duas mulheres que sentam-se para baixo lendo jornais
eu originalmente esperava | eu originalmente esperava
**Result**: 12.90%
### Test against [TEDx](http://www.openslr.org/100/) (Out-of-domain)
```python
!gdown --id 1HJEnvthaGYwcV_whHEywgH2daIN4bQna
!tar -xf tedx.tar.gz
```
```python
dataset = load_dataset('csv', data_files={'test': 'tedx/test.csv'})['test']
def map_to_array(batch):
speech, _ = torchaudio.load(batch["path"])
batch["speech"] = speech.squeeze(0).numpy()
batch["sampling_rate"] = resampler.new_freq
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
return batch
```
```python
ds = dataset.map(map_to_array)
result = ds.map(map_to_pred, batched=True, batch_size=1, remove_columns=list(ds.features.keys()))
print(wer.compute(predictions=result["predicted"], references=result["target"]))
for pred, target in zip(result["predicted"][:10], result["target"][:10]):
print(pred, "|", target)
```
0.35215851987208774
com isso a gente vê que essa rede de pactuação de de deparcerias nos remete a um raciocínio lógico que ao que a gente crê que é a prevenção | com isso a gente vê que essa rede de pactuação de parcerias nos remete a um raciocínio lógico que é o que a gente crê que é a prevenção
ente vai para o resultado | e aí a gente vai pro resultado
curiosidade hé o que eu descobri desde que comecei a fazer pesquisa lá no ensino médio | e a curiosidade é algo que descobri desde que comecei a fazer pesquisa lá no ensino médio
val des quemesho | há vários caminhos
que é uma opcissão por comer soldado | que é uma obsessão por comer saudável
isso é tão é forte algoltão universal que existem dados que mostram que setenta e cinco por cento das reuniões são dominadas pela voz masculina | e isso é tão forte é algo tão universal que existem dados que mostram que das reuniões são dominadas pela voz masculina
não era exatamente isso não estávamos deveto | e não era exatamente isso que nós estávamos a ver
durante meci do médio ofiz pesquisa estudei numa escola que chamam a fundação liberate ficava relativamente próximo daqui | durante o ensino médio eu fiz pesquisa estudei numa escola que se chama fundação liberato que fica relativamente próxima daqui
oito anos atrás eu fui apresentado por uma doença que até então eu não conhecia e que é bem provável que a maior parte de nós todos aqui não conheçamos | oito anos atrás fui apresentado para uma doença que até então eu não conhecia e que é bem provável que a maior parte de nós todos aqui não conheçamos
o terceiro é o museu do ripiopeco | o terceiro é o museu do hip hop
**Result**: 35.21% | {"language": "pt", "license": "apache-2.0", "tags": ["audio", "speech", "wav2vec2", "pt", "portuguese-speech-corpus", "automatic-speech-recognition", "speech", "PyTorch", "hf-asr-leaderboard"], "datasets": ["common_voice", "mls", "cetuc", "lapsbm", "voxforge"], "metrics": ["wer"]} | lgris/wav2vec2-large-xlsr-open-brazilian-portuguese | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"audio",
"speech",
"pt",
"portuguese-speech-corpus",
"PyTorch",
"hf-asr-leaderboard",
"dataset:common_voice",
"dataset:mls",
"dataset:cetuc",
"dataset:lapsbm",
"dataset:voxforge",
"arxiv:2012.03411",
"license:apac... | null | 2022-03-02T23:29:05+00:00 | [
"2012.03411"
] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints_compatible #has_space #region-us
| Wav2vec 2.0 With Open Brazilian Portuguese Datasets
===================================================
This a the demonstration of a fine-tuned Wav2vec model for Brazilian Portuguese using the following datasets:
* CETUC: contains approximately 145 hours of Brazilian Portuguese speech distributed among 50 male and 50 female speakers, each pronouncing approximately 1,000 phonetically balanced sentences selected from the CETEN-Folha corpus.
* Multilingual Librispeech (MLS): a massive dataset available in many languages. The MLS is based on audiobook recordings in public domain like LibriVox. The dataset contains a total of 6k hours of transcribed data in many languages. The set in Portuguese used in this work (mostly Brazilian variant) has approximately 284 hours of speech, obtained from 55 audiobooks read by 62 speakers.
* VoxForge: is a project with the goal to build open datasets for acoustic models. The corpus contains approximately 100 speakers and 4,130 utterances of Brazilian Portuguese, with sample rates varying from 16kHz to 44.1kHz.
* Common Voice 6.1 (*only train*): is a project proposed by Mozilla Foundation with the goal to create a wide open dataset in different languages to train ASR models. In this project, volunteers donate and validate speech using the oficial site. The set in Portuguese (mostly Brazilian variant) used in this work is the 6.1 version (pt\_63h\_2020-12-11) that contains about 50 validated hours and 1,120 unique speakers.
* Lapsbm: "Falabrasil - UFPA" is a dataset used by the Fala Brasil group to benchmark ASR systems in Brazilian Portuguese. Contains 35 speakers (10 females), each one pronouncing 20 unique sentences, totalling 700 utterances in Brazilian Portuguese. The audios were recorded in 22.05 kHz without environment control.
These datasets were combined to build a larger Brazilian Portuguese dataset. All data was used for training except Common Voice dev/test sets, that were used for validation/test respectively.
The original model was fine-tuned using fairseq. This notebook uses a converted version of the original one. The link to the original fairseq model is available here.
This model was trained in 80k updates.
#### Datasets in number of instances and number of frames
The following image shows the overall distribution of the dataset:
!datasets
#### Transcription examples
Imports and dependencies
------------------------
Preparation
-----------
Tests
-----
### Test against Common Voice (In-domain)
```
0.12905054857823264
nem o varanin os altros influmindo os de teterno um bombederster | nem o radar nem os outros instrumentos detectaram o bombardeiro stealth
pedir dinheiro é emprestado das pessoas do aldeia | pedir dinheiro emprestado às pessoas da aldeia
oito | oito
teno calcos | trancá-los
realizaram a investigação para resolver o problema | realizar uma investigação para resolver o problema
iotube ainda é a melhor plataforma de vídeos | o youtube ainda é a melhor plataforma de vídeos
menina e menino beijando nas sombras | menina e menino beijando nas sombras
eu sou o senhor | eu sou o senhor
duas metcas sentam-se para baixo randes jornais | duas mulheres que sentam-se para baixo lendo jornais
eu originalmente esperava | eu originalmente esperava
```
Result: 12.90%
### Test against TEDx (Out-of-domain)
```
0.35215851987208774
com isso a gente vê que essa rede de pactuação de de deparcerias nos remete a um raciocínio lógico que ao que a gente crê que é a prevenção | com isso a gente vê que essa rede de pactuação de parcerias nos remete a um raciocínio lógico que é o que a gente crê que é a prevenção
ente vai para o resultado | e aí a gente vai pro resultado
curiosidade hé o que eu descobri desde que comecei a fazer pesquisa lá no ensino médio | e a curiosidade é algo que descobri desde que comecei a fazer pesquisa lá no ensino médio
val des quemesho | há vários caminhos
que é uma opcissão por comer soldado | que é uma obsessão por comer saudável
isso é tão é forte algoltão universal que existem dados que mostram que setenta e cinco por cento das reuniões são dominadas pela voz masculina | e isso é tão forte é algo tão universal que existem dados que mostram que das reuniões são dominadas pela voz masculina
não era exatamente isso não estávamos deveto | e não era exatamente isso que nós estávamos a ver
durante meci do médio ofiz pesquisa estudei numa escola que chamam a fundação liberate ficava relativamente próximo daqui | durante o ensino médio eu fiz pesquisa estudei numa escola que se chama fundação liberato que fica relativamente próxima daqui
oito anos atrás eu fui apresentado por uma doença que até então eu não conhecia e que é bem provável que a maior parte de nós todos aqui não conheçamos | oito anos atrás fui apresentado para uma doença que até então eu não conhecia e que é bem provável que a maior parte de nós todos aqui não conheçamos
o terceiro é o museu do ripiopeco | o terceiro é o museu do hip hop
```
Result: 35.21%
| [
"#### Datasets in number of instances and number of frames\n\n\nThe following image shows the overall distribution of the dataset:\n\n\n!datasets",
"#### Transcription examples\n\n\n\nImports and dependencies\n------------------------\n\n\nPreparation\n-----------\n\n\nTests\n-----",
"### Test against Common Vo... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #audio #speech #pt #portuguese-speech-corpus #PyTorch #hf-asr-leaderboard #dataset-common_voice #dataset-mls #dataset-cetuc #dataset-lapsbm #dataset-voxforge #arxiv-2012.03411 #license-apache-2.0 #model-index #endpoints_compatible #has_space #reg... |
automatic-speech-recognition | transformers | <!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-xls-r-1b-cv8
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - PT dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2007
- Wer: 0.1838
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 7.5e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 2000
- num_epochs: 30.0
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 2.1172 | 0.32 | 500 | 1.2852 | 0.9783 |
| 1.4152 | 0.64 | 1000 | 0.6434 | 0.6105 |
| 1.4342 | 0.96 | 1500 | 0.4844 | 0.3989 |
| 1.4657 | 1.29 | 2000 | 0.5080 | 0.4490 |
| 1.4961 | 1.61 | 2500 | 0.4764 | 0.4264 |
| 1.4515 | 1.93 | 3000 | 0.4519 | 0.4068 |
| 1.3924 | 2.25 | 3500 | 0.4472 | 0.4132 |
| 1.4524 | 2.57 | 4000 | 0.4455 | 0.3939 |
| 1.4328 | 2.89 | 4500 | 0.4369 | 0.4069 |
| 1.3456 | 3.22 | 5000 | 0.4234 | 0.3774 |
| 1.3725 | 3.54 | 5500 | 0.4387 | 0.3789 |
| 1.3812 | 3.86 | 6000 | 0.4298 | 0.3825 |
| 1.3282 | 4.18 | 6500 | 0.4025 | 0.3703 |
| 1.3326 | 4.5 | 7000 | 0.3917 | 0.3502 |
| 1.3028 | 4.82 | 7500 | 0.3889 | 0.3582 |
| 1.293 | 5.14 | 8000 | 0.3859 | 0.3496 |
| 1.321 | 5.47 | 8500 | 0.3875 | 0.3576 |
| 1.3165 | 5.79 | 9000 | 0.3927 | 0.3589 |
| 1.2701 | 6.11 | 9500 | 0.4058 | 0.3621 |
| 1.2718 | 6.43 | 10000 | 0.4211 | 0.3916 |
| 1.2683 | 6.75 | 10500 | 0.3968 | 0.3620 |
| 1.2643 | 7.07 | 11000 | 0.4128 | 0.3848 |
| 1.2485 | 7.4 | 11500 | 0.3849 | 0.3727 |
| 1.2608 | 7.72 | 12000 | 0.3770 | 0.3474 |
| 1.2388 | 8.04 | 12500 | 0.3774 | 0.3574 |
| 1.2524 | 8.36 | 13000 | 0.3789 | 0.3550 |
| 1.2458 | 8.68 | 13500 | 0.3770 | 0.3410 |
| 1.2505 | 9.0 | 14000 | 0.3638 | 0.3403 |
| 1.2254 | 9.32 | 14500 | 0.3770 | 0.3509 |
| 1.2459 | 9.65 | 15000 | 0.3592 | 0.3349 |
| 1.2049 | 9.97 | 15500 | 0.3600 | 0.3428 |
| 1.2097 | 10.29 | 16000 | 0.3626 | 0.3347 |
| 1.1988 | 10.61 | 16500 | 0.3740 | 0.3269 |
| 1.1671 | 10.93 | 17000 | 0.3548 | 0.3245 |
| 1.1532 | 11.25 | 17500 | 0.3394 | 0.3140 |
| 1.1459 | 11.58 | 18000 | 0.3349 | 0.3156 |
| 1.1511 | 11.9 | 18500 | 0.3272 | 0.3110 |
| 1.1465 | 12.22 | 19000 | 0.3348 | 0.3084 |
| 1.1426 | 12.54 | 19500 | 0.3193 | 0.3027 |
| 1.1278 | 12.86 | 20000 | 0.3318 | 0.3021 |
| 1.149 | 13.18 | 20500 | 0.3169 | 0.2947 |
| 1.114 | 13.5 | 21000 | 0.3224 | 0.2986 |
| 1.1249 | 13.83 | 21500 | 0.3227 | 0.2921 |
| 1.0968 | 14.15 | 22000 | 0.3033 | 0.2878 |
| 1.0851 | 14.47 | 22500 | 0.2996 | 0.2863 |
| 1.0985 | 14.79 | 23000 | 0.3011 | 0.2843 |
| 1.0808 | 15.11 | 23500 | 0.2932 | 0.2759 |
| 1.069 | 15.43 | 24000 | 0.2919 | 0.2750 |
| 1.0602 | 15.76 | 24500 | 0.2959 | 0.2713 |
| 1.0369 | 16.08 | 25000 | 0.2931 | 0.2754 |
| 1.0573 | 16.4 | 25500 | 0.2920 | 0.2722 |
| 1.051 | 16.72 | 26000 | 0.2855 | 0.2632 |
| 1.0279 | 17.04 | 26500 | 0.2850 | 0.2649 |
| 1.0496 | 17.36 | 27000 | 0.2817 | 0.2585 |
| 1.0516 | 17.68 | 27500 | 0.2961 | 0.2635 |
| 1.0244 | 18.01 | 28000 | 0.2781 | 0.2589 |
| 1.0099 | 18.33 | 28500 | 0.2783 | 0.2565 |
| 1.0016 | 18.65 | 29000 | 0.2719 | 0.2537 |
| 1.0157 | 18.97 | 29500 | 0.2621 | 0.2449 |
| 0.9572 | 19.29 | 30000 | 0.2582 | 0.2427 |
| 0.9802 | 19.61 | 30500 | 0.2707 | 0.2468 |
| 0.9577 | 19.94 | 31000 | 0.2563 | 0.2389 |
| 0.9562 | 20.26 | 31500 | 0.2592 | 0.2382 |
| 0.962 | 20.58 | 32000 | 0.2539 | 0.2341 |
| 0.9541 | 20.9 | 32500 | 0.2505 | 0.2288 |
| 0.9587 | 21.22 | 33000 | 0.2486 | 0.2302 |
| 0.9146 | 21.54 | 33500 | 0.2461 | 0.2269 |
| 0.9215 | 21.86 | 34000 | 0.2387 | 0.2228 |
| 0.9105 | 22.19 | 34500 | 0.2405 | 0.2222 |
| 0.8949 | 22.51 | 35000 | 0.2316 | 0.2191 |
| 0.9153 | 22.83 | 35500 | 0.2358 | 0.2180 |
| 0.8907 | 23.15 | 36000 | 0.2369 | 0.2168 |
| 0.8973 | 23.47 | 36500 | 0.2323 | 0.2120 |
| 0.8878 | 23.79 | 37000 | 0.2293 | 0.2104 |
| 0.8818 | 24.12 | 37500 | 0.2302 | 0.2132 |
| 0.8919 | 24.44 | 38000 | 0.2262 | 0.2083 |
| 0.8473 | 24.76 | 38500 | 0.2257 | 0.2040 |
| 0.8516 | 25.08 | 39000 | 0.2246 | 0.2031 |
| 0.8451 | 25.4 | 39500 | 0.2198 | 0.2000 |
| 0.8288 | 25.72 | 40000 | 0.2199 | 0.1990 |
| 0.8465 | 26.05 | 40500 | 0.2165 | 0.1972 |
| 0.8305 | 26.37 | 41000 | 0.2128 | 0.1957 |
| 0.8202 | 26.69 | 41500 | 0.2127 | 0.1937 |
| 0.8223 | 27.01 | 42000 | 0.2100 | 0.1934 |
| 0.8322 | 27.33 | 42500 | 0.2076 | 0.1905 |
| 0.8139 | 27.65 | 43000 | 0.2054 | 0.1880 |
| 0.8299 | 27.97 | 43500 | 0.2026 | 0.1868 |
| 0.7937 | 28.3 | 44000 | 0.2045 | 0.1872 |
| 0.7972 | 28.62 | 44500 | 0.2025 | 0.1861 |
| 0.809 | 28.94 | 45000 | 0.2026 | 0.1858 |
| 0.813 | 29.26 | 45500 | 0.2013 | 0.1838 |
| 0.7718 | 29.58 | 46000 | 0.2010 | 0.1837 |
| 0.7929 | 29.9 | 46500 | 0.2008 | 0.1840 |
### Framework versions
- Transformers 4.17.0.dev0
- Pytorch 1.10.2+cu102
- Datasets 1.18.3.dev0
- Tokenizers 0.11.0
| {"language": ["pt"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "hf-asr-leaderboard", "mozilla-foundation/common_voice_8_0", "pt", "robust-speech-event"], "datasets": ["mozilla-foundation/common_voice_8_0"], "model-index": [{"name": "wav2vec2-xls-r-1b-cv8", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 8", "type": "mozilla-foundation/common_voice_8_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 17.7, "name": "Test WER"}, {"type": "cer", "value": 5.21, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 45.68, "name": "Test WER"}, {"type": "cer", "value": 18.67, "name": "Test CER"}, {"type": "wer", "value": 45.29, "name": "Test WER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Test Data", "type": "speech-recognition-community-v2/eval_data", "args": "pt"}, "metrics": [{"type": "wer", "value": 48.03, "name": "Test WER"}]}]}]} | lgris/wav2vec2-xls-r-1b-cv8 | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"hf-asr-leaderboard",
"mozilla-foundation/common_voice_8_0",
"pt",
"robust-speech-event",
"dataset:mozilla-foundation/common_voice_8_0",
"license:apache-2.0",
"model-index",
"endpoints_compatible... | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #mozilla-foundation/common_voice_8_0 #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-xls-r-1b-cv8
=====================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the MOZILLA-FOUNDATION/COMMON\_VOICE\_8\_0 - PT dataset.
It achieves the following results on the evaluation set:
* Loss: 0.2007
* Wer: 0.1838
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 7.5e-05
* train\_batch\_size: 4
* eval\_batch\_size: 4
* seed: 42
* gradient\_accumulation\_steps: 4
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 2000
* num\_epochs: 30.0
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.17.0.dev0
* Pytorch 1.10.2+cu102
* Datasets 1.18.3.dev0
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 7.5e-05\n* train\\_batch\\_size: 4\n* eval\\_batch\\_size: 4\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #mozilla-foundation/common_voice_8_0 #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperpar... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-xls-r-1b-portuguese-CORAA-3
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on [CORAA dataset](https://github.com/nilc-nlp/CORAA).
It achieves the following results on the evaluation set:
- Loss: 1.0029
- Wer: 0.6020
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 5000
- training_steps: 30000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 2.0169 | 0.21 | 5000 | 1.9582 | 0.9283 |
| 1.8561 | 0.42 | 10000 | 1.6144 | 0.8554 |
| 1.6823 | 0.63 | 15000 | 1.4165 | 0.7710 |
| 1.52 | 0.84 | 20000 | 1.2441 | 0.7289 |
| 1.3757 | 1.05 | 25000 | 1.1061 | 0.6491 |
| 1.2377 | 1.26 | 30000 | 1.0029 | 0.6020 |
### Framework versions
- Transformers 4.17.0.dev0
- Pytorch 1.10.2+cu102
- Datasets 1.18.3.dev0
- Tokenizers 0.11.0
| {"language": ["pt"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "pt", "robust-speech-event", "hf-asr-leaderboard"], "model-index": [{"name": "wav2vec2-xls-r-1b-portuguese-CORAA-3", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 8", "type": "mozilla-foundation/common_voice_8_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 71.67, "name": "Test WER"}, {"type": "cer", "value": 30.64, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 7", "type": "mozilla-foundation/common_voice_7_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 68.18, "name": "Test WER"}, {"type": "cer", "value": 28.34, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 56.76, "name": "Test WER"}, {"type": "cer", "value": 23.7, "name": "Test CER"}]}]}]} | lgris/wav2vec2-xls-r-1b-portuguese-CORAA-3 | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"pt",
"robust-speech-event",
"hf-asr-leaderboard",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #generated_from_trainer #pt #robust-speech-event #hf-asr-leaderboard #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-xls-r-1b-portuguese-CORAA-3
====================================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on CORAA dataset.
It achieves the following results on the evaluation set:
* Loss: 1.0029
* Wer: 0.6020
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0003
* train\_batch\_size: 4
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 4
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 5000
* training\_steps: 30000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.17.0.dev0
* Pytorch 1.10.2+cu102
* Datasets 1.18.3.dev0
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0003\n* train\\_batch\\_size: 4\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #generated_from_trainer #pt #robust-speech-event #hf-asr-leaderboard #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learn... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-xls-r-300m-gn-cv8-3
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.9517
- Wer: 0.8542
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 5000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:------:|:----:|:---------------:|:------:|
| 19.9125 | 5.54 | 100 | 5.4279 | 1.0 |
| 3.8031 | 11.11 | 200 | 3.3070 | 1.0 |
| 3.3783 | 16.65 | 300 | 3.2450 | 1.0 |
| 3.3472 | 22.22 | 400 | 3.2424 | 1.0 |
| 3.2714 | 27.76 | 500 | 3.1100 | 1.0 |
| 3.2367 | 33.32 | 600 | 3.1091 | 1.0 |
| 3.1968 | 38.86 | 700 | 3.1013 | 1.0 |
| 3.2004 | 44.43 | 800 | 3.1173 | 1.0 |
| 3.1656 | 49.97 | 900 | 3.0682 | 1.0 |
| 3.1563 | 55.54 | 1000 | 3.0457 | 1.0 |
| 3.1356 | 61.11 | 1100 | 3.0139 | 1.0 |
| 3.086 | 66.65 | 1200 | 2.8108 | 1.0 |
| 2.954 | 72.22 | 1300 | 2.3238 | 1.0 |
| 2.6125 | 77.76 | 1400 | 1.6461 | 1.0 |
| 2.3296 | 83.32 | 1500 | 1.2834 | 0.9744 |
| 2.1345 | 88.86 | 1600 | 1.1091 | 0.9693 |
| 2.0346 | 94.43 | 1700 | 1.0273 | 0.9233 |
| 1.9611 | 99.97 | 1800 | 0.9642 | 0.9182 |
| 1.9066 | 105.54 | 1900 | 0.9590 | 0.9105 |
| 1.8178 | 111.11 | 2000 | 0.9679 | 0.9028 |
| 1.7799 | 116.65 | 2100 | 0.9007 | 0.8619 |
| 1.7726 | 122.22 | 2200 | 0.9689 | 0.8951 |
| 1.7389 | 127.76 | 2300 | 0.8876 | 0.8593 |
| 1.7151 | 133.32 | 2400 | 0.8716 | 0.8542 |
| 1.6842 | 138.86 | 2500 | 0.9536 | 0.8772 |
| 1.6449 | 144.43 | 2600 | 0.9296 | 0.8542 |
| 1.5978 | 149.97 | 2700 | 0.8895 | 0.8440 |
| 1.6515 | 155.54 | 2800 | 0.9162 | 0.8568 |
| 1.6586 | 161.11 | 2900 | 0.9039 | 0.8568 |
| 1.5966 | 166.65 | 3000 | 0.8627 | 0.8542 |
| 1.5695 | 172.22 | 3100 | 0.9549 | 0.8824 |
| 1.5699 | 177.76 | 3200 | 0.9332 | 0.8517 |
| 1.5297 | 183.32 | 3300 | 0.9163 | 0.8338 |
| 1.5367 | 188.86 | 3400 | 0.8822 | 0.8312 |
| 1.5586 | 194.43 | 3500 | 0.9217 | 0.8363 |
| 1.5429 | 199.97 | 3600 | 0.9564 | 0.8568 |
| 1.5273 | 205.54 | 3700 | 0.9508 | 0.8542 |
| 1.5043 | 211.11 | 3800 | 0.9374 | 0.8542 |
| 1.4724 | 216.65 | 3900 | 0.9622 | 0.8619 |
| 1.4794 | 222.22 | 4000 | 0.9550 | 0.8363 |
| 1.4843 | 227.76 | 4100 | 0.9577 | 0.8465 |
| 1.4781 | 233.32 | 4200 | 0.9543 | 0.8440 |
| 1.4507 | 238.86 | 4300 | 0.9553 | 0.8491 |
| 1.4997 | 244.43 | 4400 | 0.9728 | 0.8491 |
| 1.4371 | 249.97 | 4500 | 0.9543 | 0.8670 |
| 1.4825 | 255.54 | 4600 | 0.9636 | 0.8619 |
| 1.4187 | 261.11 | 4700 | 0.9609 | 0.8440 |
| 1.4363 | 266.65 | 4800 | 0.9567 | 0.8593 |
| 1.4463 | 272.22 | 4900 | 0.9581 | 0.8542 |
| 1.4117 | 277.76 | 5000 | 0.9517 | 0.8542 |
### Framework versions
- Transformers 4.16.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.1
- Tokenizers 0.11.0
| {"language": ["gn"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "gn", "robust-speech-event", "hf-asr-leaderboard"], "datasets": ["mozilla-foundation/common_voice_8_0"], "model-index": [{"name": "wav2vec2-xls-r-300m-gn-cv8-3", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 8.0", "type": "mozilla-foundation/common_voice_8_0", "args": "gn"}, "metrics": [{"type": "wer", "value": 76.68, "name": "Test WER"}]}]}]} | lgris/wav2vec2-xls-r-300m-gn-cv8-3 | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"gn",
"robust-speech-event",
"hf-asr-leaderboard",
"dataset:mozilla-foundation/common_voice_8_0",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"gn"
] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-xls-r-300m-gn-cv8-3
============================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.9517
* Wer: 0.8542
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 5000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.16.0
* Pytorch 1.10.0+cu111
* Datasets 1.18.1
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe followi... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-xls-r-300m-gn-cv8-4
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 1.5805
- Wer: 0.7545
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 13000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:------:|:-----:|:---------------:|:------:|
| 9.2216 | 16.65 | 300 | 3.2771 | 1.0 |
| 3.1804 | 33.32 | 600 | 2.2869 | 1.0 |
| 1.5856 | 49.97 | 900 | 0.9573 | 0.8772 |
| 1.0299 | 66.65 | 1200 | 0.9044 | 0.8082 |
| 0.8916 | 83.32 | 1500 | 0.9478 | 0.8056 |
| 0.8451 | 99.97 | 1800 | 0.8814 | 0.8107 |
| 0.7649 | 116.65 | 2100 | 0.9897 | 0.7826 |
| 0.7185 | 133.32 | 2400 | 0.9988 | 0.7621 |
| 0.6595 | 149.97 | 2700 | 1.0607 | 0.7749 |
| 0.6211 | 166.65 | 3000 | 1.1826 | 0.7877 |
| 0.59 | 183.32 | 3300 | 1.1060 | 0.7826 |
| 0.5383 | 199.97 | 3600 | 1.1826 | 0.7852 |
| 0.5205 | 216.65 | 3900 | 1.2148 | 0.8261 |
| 0.4786 | 233.32 | 4200 | 1.2710 | 0.7928 |
| 0.4482 | 249.97 | 4500 | 1.1943 | 0.7980 |
| 0.4149 | 266.65 | 4800 | 1.2449 | 0.8031 |
| 0.3904 | 283.32 | 5100 | 1.3100 | 0.7928 |
| 0.3619 | 299.97 | 5400 | 1.3125 | 0.7596 |
| 0.3496 | 316.65 | 5700 | 1.3699 | 0.7877 |
| 0.3277 | 333.32 | 6000 | 1.4344 | 0.8031 |
| 0.2958 | 349.97 | 6300 | 1.4093 | 0.7980 |
| 0.2883 | 366.65 | 6600 | 1.3296 | 0.7570 |
| 0.2598 | 383.32 | 6900 | 1.4026 | 0.7980 |
| 0.2564 | 399.97 | 7200 | 1.4847 | 0.8031 |
| 0.2408 | 416.65 | 7500 | 1.4896 | 0.8107 |
| 0.2266 | 433.32 | 7800 | 1.4232 | 0.7698 |
| 0.224 | 449.97 | 8100 | 1.5560 | 0.7903 |
| 0.2038 | 466.65 | 8400 | 1.5355 | 0.7724 |
| 0.1948 | 483.32 | 8700 | 1.4624 | 0.7621 |
| 0.1995 | 499.97 | 9000 | 1.5808 | 0.7724 |
| 0.1864 | 516.65 | 9300 | 1.5653 | 0.7698 |
| 0.18 | 533.32 | 9600 | 1.4868 | 0.7494 |
| 0.1689 | 549.97 | 9900 | 1.5379 | 0.7749 |
| 0.1624 | 566.65 | 10200 | 1.5936 | 0.7749 |
| 0.1537 | 583.32 | 10500 | 1.6436 | 0.7801 |
| 0.1455 | 599.97 | 10800 | 1.6401 | 0.7673 |
| 0.1437 | 616.65 | 11100 | 1.6069 | 0.7673 |
| 0.1452 | 633.32 | 11400 | 1.6041 | 0.7519 |
| 0.139 | 649.97 | 11700 | 1.5758 | 0.7545 |
| 0.1299 | 666.65 | 12000 | 1.5559 | 0.7545 |
| 0.127 | 683.32 | 12300 | 1.5776 | 0.7596 |
| 0.1264 | 699.97 | 12600 | 1.5790 | 0.7519 |
| 0.1209 | 716.65 | 12900 | 1.5805 | 0.7545 |
### Framework versions
- Transformers 4.16.1
- Pytorch 1.10.0+cu111
- Datasets 1.18.2
- Tokenizers 0.11.0
| {"language": ["gn"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "gn", "robust-speech-event", "hf-asr-leaderboard"], "datasets": ["mozilla-foundation/common_voice_8_0"], "model-index": [{"name": "wav2vec2-xls-r-300m-gn-cv8-4", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 8.0", "type": "mozilla-foundation/common_voice_8_0", "args": "gn"}, "metrics": [{"type": "wer", "value": 68.45, "name": "Test WER"}]}]}]} | lgris/wav2vec2-xls-r-300m-gn-cv8-4 | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"gn",
"robust-speech-event",
"hf-asr-leaderboard",
"dataset:mozilla-foundation/common_voice_8_0",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"gn"
] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-xls-r-300m-gn-cv8-4
============================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 1.5805
* Wer: 0.7545
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 13000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.16.1
* Pytorch 1.10.0+cu111
* Datasets 1.18.2
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe followi... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-xls-r-300m-gn-cv8
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.9392
- Wer: 0.7033
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 5000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:------:|:----:|:---------------:|:------:|
| 20.0601 | 5.54 | 100 | 5.1622 | 1.0 |
| 3.7052 | 11.11 | 200 | 3.2869 | 1.0 |
| 3.3275 | 16.65 | 300 | 3.2162 | 1.0 |
| 3.2984 | 22.22 | 400 | 3.1638 | 1.0 |
| 3.1111 | 27.76 | 500 | 2.5541 | 1.0 |
| 2.238 | 33.32 | 600 | 1.2198 | 0.9616 |
| 1.5284 | 38.86 | 700 | 0.9571 | 0.8593 |
| 1.2735 | 44.43 | 800 | 0.8719 | 0.8363 |
| 1.1269 | 49.97 | 900 | 0.8334 | 0.7954 |
| 1.0427 | 55.54 | 1000 | 0.7700 | 0.7749 |
| 1.0152 | 61.11 | 1100 | 0.7747 | 0.7877 |
| 0.943 | 66.65 | 1200 | 0.7151 | 0.7442 |
| 0.9132 | 72.22 | 1300 | 0.7224 | 0.7289 |
| 0.8397 | 77.76 | 1400 | 0.7354 | 0.7059 |
| 0.8577 | 83.32 | 1500 | 0.7285 | 0.7263 |
| 0.7931 | 88.86 | 1600 | 0.7863 | 0.7084 |
| 0.7995 | 94.43 | 1700 | 0.7562 | 0.6880 |
| 0.799 | 99.97 | 1800 | 0.7905 | 0.7059 |
| 0.7373 | 105.54 | 1900 | 0.7791 | 0.7161 |
| 0.749 | 111.11 | 2000 | 0.8125 | 0.7161 |
| 0.6925 | 116.65 | 2100 | 0.7722 | 0.6905 |
| 0.7034 | 122.22 | 2200 | 0.8989 | 0.7136 |
| 0.6745 | 127.76 | 2300 | 0.8270 | 0.6982 |
| 0.6837 | 133.32 | 2400 | 0.8569 | 0.7161 |
| 0.6689 | 138.86 | 2500 | 0.8339 | 0.6982 |
| 0.6471 | 144.43 | 2600 | 0.8441 | 0.7110 |
| 0.615 | 149.97 | 2700 | 0.9038 | 0.7212 |
| 0.6477 | 155.54 | 2800 | 0.9089 | 0.7059 |
| 0.6047 | 161.11 | 2900 | 0.9149 | 0.7059 |
| 0.5613 | 166.65 | 3000 | 0.8582 | 0.7263 |
| 0.6017 | 172.22 | 3100 | 0.8787 | 0.7084 |
| 0.5546 | 177.76 | 3200 | 0.8753 | 0.6957 |
| 0.5747 | 183.32 | 3300 | 0.9167 | 0.7212 |
| 0.5535 | 188.86 | 3400 | 0.8448 | 0.6905 |
| 0.5331 | 194.43 | 3500 | 0.8644 | 0.7161 |
| 0.5428 | 199.97 | 3600 | 0.8730 | 0.7033 |
| 0.5219 | 205.54 | 3700 | 0.9047 | 0.6982 |
| 0.5158 | 211.11 | 3800 | 0.8706 | 0.7033 |
| 0.5107 | 216.65 | 3900 | 0.9139 | 0.7084 |
| 0.4903 | 222.22 | 4000 | 0.9456 | 0.7315 |
| 0.4772 | 227.76 | 4100 | 0.9475 | 0.7161 |
| 0.4713 | 233.32 | 4200 | 0.9237 | 0.7059 |
| 0.4743 | 238.86 | 4300 | 0.9305 | 0.6957 |
| 0.4705 | 244.43 | 4400 | 0.9561 | 0.7110 |
| 0.4908 | 249.97 | 4500 | 0.9389 | 0.7084 |
| 0.4717 | 255.54 | 4600 | 0.9234 | 0.6982 |
| 0.4462 | 261.11 | 4700 | 0.9323 | 0.6957 |
| 0.4556 | 266.65 | 4800 | 0.9432 | 0.7033 |
| 0.4691 | 272.22 | 4900 | 0.9389 | 0.7059 |
| 0.4601 | 277.76 | 5000 | 0.9392 | 0.7033 |
### Framework versions
- Transformers 4.16.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.1
- Tokenizers 0.11.0
| {"language": ["gn"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "gn", "robust-speech-event", "hf-asr-leaderboard"], "datasets": ["mozilla-foundation/common_voice_8_0"], "model-index": [{"name": "wav2vec2-xls-r-300m-gn-cv8", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 8", "type": "mozilla-foundation/common_voice_8_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 69.05, "name": "Test WER"}, {"type": "cer", "value": 14.7, "name": "Test CER"}, {"type": "wer", "value": 69.05, "name": "Test WER"}]}]}]} | lgris/wav2vec2-xls-r-300m-gn-cv8 | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"gn",
"robust-speech-event",
"hf-asr-leaderboard",
"dataset:mozilla-foundation/common_voice_8_0",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"gn"
] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-xls-r-300m-gn-cv8
==========================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.9392
* Wer: 0.7033
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 5000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.16.0
* Pytorch 1.10.0+cu111
* Datasets 1.18.1
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_8_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe followi... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-xls-r-gn-cv7
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 1.7197
- Wer: 0.7434
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 13000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:------:|:-----:|:---------------:|:------:|
| 3.4669 | 6.24 | 100 | 3.3003 | 1.0 |
| 3.3214 | 12.48 | 200 | 3.2090 | 1.0 |
| 3.1619 | 18.73 | 300 | 2.6322 | 1.0 |
| 1.751 | 24.97 | 400 | 1.4089 | 0.9803 |
| 0.7997 | 31.24 | 500 | 0.9996 | 0.9211 |
| 0.4996 | 37.48 | 600 | 0.9879 | 0.8553 |
| 0.3677 | 43.73 | 700 | 0.9543 | 0.8289 |
| 0.2851 | 49.97 | 800 | 1.0627 | 0.8487 |
| 0.2556 | 56.24 | 900 | 1.0933 | 0.8355 |
| 0.2268 | 62.48 | 1000 | 0.9191 | 0.8026 |
| 0.1914 | 68.73 | 1100 | 0.9582 | 0.7961 |
| 0.1749 | 74.97 | 1200 | 1.0502 | 0.8092 |
| 0.157 | 81.24 | 1300 | 0.9998 | 0.7632 |
| 0.1505 | 87.48 | 1400 | 1.0076 | 0.7303 |
| 0.1278 | 93.73 | 1500 | 0.9321 | 0.75 |
| 0.1078 | 99.97 | 1600 | 1.0383 | 0.7697 |
| 0.1156 | 106.24 | 1700 | 1.0302 | 0.7763 |
| 0.1107 | 112.48 | 1800 | 1.0419 | 0.7763 |
| 0.091 | 118.73 | 1900 | 1.0694 | 0.75 |
| 0.0829 | 124.97 | 2000 | 1.0257 | 0.7829 |
| 0.0865 | 131.24 | 2100 | 1.2108 | 0.7368 |
| 0.0907 | 137.48 | 2200 | 1.0458 | 0.7697 |
| 0.0897 | 143.73 | 2300 | 1.1504 | 0.7895 |
| 0.0766 | 149.97 | 2400 | 1.1663 | 0.7237 |
| 0.0659 | 156.24 | 2500 | 1.1320 | 0.7632 |
| 0.0699 | 162.48 | 2600 | 1.2586 | 0.7434 |
| 0.0613 | 168.73 | 2700 | 1.1815 | 0.8158 |
| 0.0598 | 174.97 | 2800 | 1.3299 | 0.75 |
| 0.0577 | 181.24 | 2900 | 1.2035 | 0.7171 |
| 0.0576 | 187.48 | 3000 | 1.2134 | 0.7434 |
| 0.0518 | 193.73 | 3100 | 1.3406 | 0.7566 |
| 0.0524 | 199.97 | 3200 | 1.4251 | 0.75 |
| 0.0467 | 206.24 | 3300 | 1.3533 | 0.7697 |
| 0.0428 | 212.48 | 3400 | 1.2463 | 0.7368 |
| 0.0453 | 218.73 | 3500 | 1.4532 | 0.7566 |
| 0.0473 | 224.97 | 3600 | 1.3152 | 0.7434 |
| 0.0451 | 231.24 | 3700 | 1.2232 | 0.7368 |
| 0.0361 | 237.48 | 3800 | 1.2938 | 0.7171 |
| 0.045 | 243.73 | 3900 | 1.4148 | 0.7434 |
| 0.0422 | 249.97 | 4000 | 1.3786 | 0.7961 |
| 0.036 | 256.24 | 4100 | 1.4488 | 0.7697 |
| 0.0352 | 262.48 | 4200 | 1.2294 | 0.6776 |
| 0.0326 | 268.73 | 4300 | 1.2796 | 0.6974 |
| 0.034 | 274.97 | 4400 | 1.3805 | 0.7303 |
| 0.0305 | 281.24 | 4500 | 1.4994 | 0.7237 |
| 0.0325 | 287.48 | 4600 | 1.4330 | 0.6908 |
| 0.0338 | 293.73 | 4700 | 1.3091 | 0.7368 |
| 0.0306 | 299.97 | 4800 | 1.2174 | 0.7171 |
| 0.0299 | 306.24 | 4900 | 1.3527 | 0.7763 |
| 0.0287 | 312.48 | 5000 | 1.3651 | 0.7368 |
| 0.0274 | 318.73 | 5100 | 1.4337 | 0.7368 |
| 0.0258 | 324.97 | 5200 | 1.3831 | 0.6908 |
| 0.022 | 331.24 | 5300 | 1.3556 | 0.6974 |
| 0.021 | 337.48 | 5400 | 1.3836 | 0.7237 |
| 0.0241 | 343.73 | 5500 | 1.4352 | 0.7039 |
| 0.0229 | 349.97 | 5600 | 1.3904 | 0.7105 |
| 0.026 | 356.24 | 5700 | 1.4131 | 0.7171 |
| 0.021 | 362.48 | 5800 | 1.5426 | 0.6974 |
| 0.0191 | 368.73 | 5900 | 1.5960 | 0.7632 |
| 0.0227 | 374.97 | 6000 | 1.6240 | 0.7368 |
| 0.0204 | 381.24 | 6100 | 1.4301 | 0.7105 |
| 0.0175 | 387.48 | 6200 | 1.5554 | 0.75 |
| 0.0183 | 393.73 | 6300 | 1.6044 | 0.7697 |
| 0.0183 | 399.97 | 6400 | 1.5963 | 0.7368 |
| 0.016 | 406.24 | 6500 | 1.5679 | 0.7829 |
| 0.0178 | 412.48 | 6600 | 1.5928 | 0.7697 |
| 0.014 | 418.73 | 6700 | 1.7000 | 0.7632 |
| 0.0182 | 424.97 | 6800 | 1.5340 | 0.75 |
| 0.0148 | 431.24 | 6900 | 1.9274 | 0.7368 |
| 0.0148 | 437.48 | 7000 | 1.6437 | 0.7697 |
| 0.0173 | 443.73 | 7100 | 1.5468 | 0.75 |
| 0.0109 | 449.97 | 7200 | 1.6083 | 0.75 |
| 0.0167 | 456.24 | 7300 | 1.6732 | 0.75 |
| 0.0139 | 462.48 | 7400 | 1.5097 | 0.7237 |
| 0.013 | 468.73 | 7500 | 1.5947 | 0.7171 |
| 0.0128 | 474.97 | 7600 | 1.6260 | 0.7105 |
| 0.0166 | 481.24 | 7700 | 1.5756 | 0.7237 |
| 0.0127 | 487.48 | 7800 | 1.4506 | 0.6908 |
| 0.013 | 493.73 | 7900 | 1.4882 | 0.7368 |
| 0.0125 | 499.97 | 8000 | 1.5589 | 0.7829 |
| 0.0141 | 506.24 | 8100 | 1.6328 | 0.7434 |
| 0.0115 | 512.48 | 8200 | 1.6586 | 0.7434 |
| 0.0117 | 518.73 | 8300 | 1.6043 | 0.7105 |
| 0.009 | 524.97 | 8400 | 1.6508 | 0.7237 |
| 0.0108 | 531.24 | 8500 | 1.4507 | 0.6974 |
| 0.011 | 537.48 | 8600 | 1.5942 | 0.7434 |
| 0.009 | 543.73 | 8700 | 1.8121 | 0.7697 |
| 0.0112 | 549.97 | 8800 | 1.6923 | 0.7697 |
| 0.0073 | 556.24 | 8900 | 1.7096 | 0.7368 |
| 0.0098 | 562.48 | 9000 | 1.7052 | 0.7829 |
| 0.0088 | 568.73 | 9100 | 1.6956 | 0.7566 |
| 0.0099 | 574.97 | 9200 | 1.4909 | 0.7171 |
| 0.0075 | 581.24 | 9300 | 1.6307 | 0.7697 |
| 0.0077 | 587.48 | 9400 | 1.6196 | 0.7961 |
| 0.0088 | 593.73 | 9500 | 1.6119 | 0.7566 |
| 0.0085 | 599.97 | 9600 | 1.4512 | 0.7368 |
| 0.0086 | 606.24 | 9700 | 1.5992 | 0.7237 |
| 0.0109 | 612.48 | 9800 | 1.4706 | 0.7368 |
| 0.0098 | 618.73 | 9900 | 1.3824 | 0.7171 |
| 0.0091 | 624.97 | 10000 | 1.4776 | 0.6974 |
| 0.0072 | 631.24 | 10100 | 1.4896 | 0.7039 |
| 0.0087 | 637.48 | 10200 | 1.5467 | 0.7368 |
| 0.007 | 643.73 | 10300 | 1.5493 | 0.75 |
| 0.0076 | 649.97 | 10400 | 1.5706 | 0.7303 |
| 0.0085 | 656.24 | 10500 | 1.5748 | 0.7237 |
| 0.0075 | 662.48 | 10600 | 1.5081 | 0.7105 |
| 0.0068 | 668.73 | 10700 | 1.4967 | 0.6842 |
| 0.0117 | 674.97 | 10800 | 1.4986 | 0.7105 |
| 0.0054 | 681.24 | 10900 | 1.5587 | 0.7303 |
| 0.0059 | 687.48 | 11000 | 1.5886 | 0.7171 |
| 0.0071 | 693.73 | 11100 | 1.5746 | 0.7171 |
| 0.0048 | 699.97 | 11200 | 1.6166 | 0.7237 |
| 0.0048 | 706.24 | 11300 | 1.6098 | 0.7237 |
| 0.0056 | 712.48 | 11400 | 1.5834 | 0.7237 |
| 0.0048 | 718.73 | 11500 | 1.5653 | 0.7171 |
| 0.0045 | 724.97 | 11600 | 1.6252 | 0.7237 |
| 0.0068 | 731.24 | 11700 | 1.6794 | 0.7171 |
| 0.0044 | 737.48 | 11800 | 1.6881 | 0.7039 |
| 0.008 | 743.73 | 11900 | 1.7393 | 0.75 |
| 0.0045 | 749.97 | 12000 | 1.6869 | 0.7237 |
| 0.0047 | 756.24 | 12100 | 1.7105 | 0.7303 |
| 0.0057 | 762.48 | 12200 | 1.7439 | 0.7303 |
| 0.004 | 768.73 | 12300 | 1.7871 | 0.7434 |
| 0.0061 | 774.97 | 12400 | 1.7812 | 0.7303 |
| 0.005 | 781.24 | 12500 | 1.7410 | 0.7434 |
| 0.0056 | 787.48 | 12600 | 1.7220 | 0.7303 |
| 0.0064 | 793.73 | 12700 | 1.7141 | 0.7434 |
| 0.0042 | 799.97 | 12800 | 1.7139 | 0.7368 |
| 0.0049 | 806.24 | 12900 | 1.7211 | 0.7434 |
| 0.0044 | 812.48 | 13000 | 1.7197 | 0.7434 |
### Framework versions
- Transformers 4.15.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.0
- Tokenizers 0.10.3
| {"language": ["gn"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "gn", "robust-speech-event", "hf-asr-leaderboard"], "datasets": ["mozilla-foundation/common_voice_7_0"], "model-index": [{"name": "wav2vec2-xls-r-gn-cv7", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 7", "type": "mozilla-foundation/common_voice_7_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 73.02, "name": "Validation WER"}, {"type": "cer", "value": 17.79, "name": "Validation CER"}, {"type": "wer", "value": 62.65, "name": "Test WER"}]}]}]} | lgris/wav2vec2-xls-r-gn-cv7 | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"gn",
"robust-speech-event",
"hf-asr-leaderboard",
"dataset:mozilla-foundation/common_voice_7_0",
"license:apache-2.0",
"model-index",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"gn"
] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-xls-r-gn-cv7
=====================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 1.7197
* Wer: 0.7434
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 13000
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.15.0
* Pytorch 1.10.0+cu111
* Datasets 1.18.0
* Tokenizers 0.10.3
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #gn #robust-speech-event #hf-asr-leaderboard #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe followi... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-xls-r-pt-cv7-from-bp400h
This model is a fine-tuned version of [lgris/bp_400h_xlsr2_300M](https://huggingface.co/lgris/bp_400h_xlsr2_300M) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1535
- Wer: 0.1254
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 5000
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 0.4991 | 0.13 | 100 | 0.1774 | 0.1464 |
| 0.4655 | 0.26 | 200 | 0.1884 | 0.1568 |
| 0.4689 | 0.39 | 300 | 0.2282 | 0.1672 |
| 0.4662 | 0.52 | 400 | 0.1997 | 0.1584 |
| 0.4592 | 0.65 | 500 | 0.1989 | 0.1663 |
| 0.4533 | 0.78 | 600 | 0.2004 | 0.1698 |
| 0.4391 | 0.91 | 700 | 0.1888 | 0.1642 |
| 0.4655 | 1.04 | 800 | 0.1921 | 0.1624 |
| 0.4138 | 1.17 | 900 | 0.1950 | 0.1602 |
| 0.374 | 1.3 | 1000 | 0.2077 | 0.1658 |
| 0.4064 | 1.43 | 1100 | 0.1945 | 0.1596 |
| 0.3922 | 1.56 | 1200 | 0.2069 | 0.1665 |
| 0.4226 | 1.69 | 1300 | 0.1962 | 0.1573 |
| 0.3974 | 1.82 | 1400 | 0.1919 | 0.1553 |
| 0.3631 | 1.95 | 1500 | 0.1854 | 0.1573 |
| 0.3797 | 2.08 | 1600 | 0.1902 | 0.1550 |
| 0.3287 | 2.21 | 1700 | 0.1926 | 0.1598 |
| 0.3568 | 2.34 | 1800 | 0.1888 | 0.1534 |
| 0.3415 | 2.47 | 1900 | 0.1834 | 0.1502 |
| 0.3545 | 2.6 | 2000 | 0.1906 | 0.1560 |
| 0.3344 | 2.73 | 2100 | 0.1804 | 0.1524 |
| 0.3308 | 2.86 | 2200 | 0.1741 | 0.1485 |
| 0.344 | 2.99 | 2300 | 0.1787 | 0.1455 |
| 0.309 | 3.12 | 2400 | 0.1773 | 0.1448 |
| 0.312 | 3.25 | 2500 | 0.1738 | 0.1440 |
| 0.3066 | 3.38 | 2600 | 0.1727 | 0.1417 |
| 0.2999 | 3.51 | 2700 | 0.1692 | 0.1436 |
| 0.2985 | 3.64 | 2800 | 0.1732 | 0.1430 |
| 0.3058 | 3.77 | 2900 | 0.1754 | 0.1402 |
| 0.2943 | 3.9 | 3000 | 0.1691 | 0.1379 |
| 0.2813 | 4.03 | 3100 | 0.1754 | 0.1376 |
| 0.2733 | 4.16 | 3200 | 0.1639 | 0.1363 |
| 0.2592 | 4.29 | 3300 | 0.1675 | 0.1349 |
| 0.2697 | 4.42 | 3400 | 0.1618 | 0.1360 |
| 0.2538 | 4.55 | 3500 | 0.1658 | 0.1348 |
| 0.2746 | 4.67 | 3600 | 0.1674 | 0.1325 |
| 0.2655 | 4.8 | 3700 | 0.1655 | 0.1319 |
| 0.2745 | 4.93 | 3800 | 0.1665 | 0.1316 |
| 0.2617 | 5.06 | 3900 | 0.1600 | 0.1311 |
| 0.2674 | 5.19 | 4000 | 0.1623 | 0.1311 |
| 0.237 | 5.32 | 4100 | 0.1591 | 0.1315 |
| 0.2669 | 5.45 | 4200 | 0.1584 | 0.1295 |
| 0.2476 | 5.58 | 4300 | 0.1572 | 0.1285 |
| 0.2445 | 5.71 | 4400 | 0.1580 | 0.1271 |
| 0.2207 | 5.84 | 4500 | 0.1567 | 0.1269 |
| 0.2289 | 5.97 | 4600 | 0.1536 | 0.1260 |
| 0.2438 | 6.1 | 4700 | 0.1530 | 0.1260 |
| 0.227 | 6.23 | 4800 | 0.1544 | 0.1249 |
| 0.2256 | 6.36 | 4900 | 0.1543 | 0.1254 |
| 0.2184 | 6.49 | 5000 | 0.1535 | 0.1254 |
### Framework versions
- Transformers 4.15.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.0
- Tokenizers 0.10.3
| {"language": ["pt"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "hf-asr-leaderboard", "mozilla-foundation/common_voice_7_0", "pt", "robust-speech-event"], "datasets": ["mozilla-foundation/common_voice_7_0"], "model-index": [{"name": "wav2vec2-xls-r-pt-cv7-from-bp400h", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 7", "type": "mozilla-foundation/common_voice_7_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 12.13, "name": "Test WER"}, {"type": "cer", "value": 3.68, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 28.23, "name": "Test WER"}, {"type": "cer", "value": 12.58, "name": "Test CER"}, {"type": "wer", "value": 26.58, "name": "Test WER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Test Data", "type": "speech-recognition-community-v2/eval_data", "args": "pt"}, "metrics": [{"type": "wer", "value": 26.86, "name": "Test WER"}]}]}]} | lgris/wav2vec2-xls-r-pt-cv7-from-bp400h | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"hf-asr-leaderboard",
"mozilla-foundation/common_voice_7_0",
"pt",
"robust-speech-event",
"dataset:mozilla-foundation/common_voice_7_0",
"license:apache-2.0",
"model-index",
"end... | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #mozilla-foundation/common_voice_7_0 #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2-xls-r-pt-cv7-from-bp400h
=================================
This model is a fine-tuned version of lgris/bp\_400h\_xlsr2\_300M on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.1535
* Wer: 0.1254
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 5000
### Training results
### Framework versions
* Transformers 4.15.0
* Pytorch 1.10.0+cu111
* Datasets 1.18.0
* Tokenizers 0.10.3
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #mozilla-foundation/common_voice_7_0 #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Trai... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2_base_10k_8khz_pt_cv7_2
This model is a fine-tuned version of [lgris/seasr_2022_base_10k_8khz_pt](https://huggingface.co/lgris/seasr_2022_base_10k_8khz_pt) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 76.3426
- Wer: 0.1979
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 10000
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 189.1362 | 0.65 | 500 | 80.6347 | 0.2139 |
| 174.2587 | 1.3 | 1000 | 80.2062 | 0.2116 |
| 164.676 | 1.95 | 1500 | 78.2161 | 0.2073 |
| 176.5856 | 2.6 | 2000 | 78.8920 | 0.2074 |
| 164.3583 | 3.25 | 2500 | 77.2865 | 0.2066 |
| 161.414 | 3.9 | 3000 | 77.8888 | 0.2048 |
| 158.283 | 4.55 | 3500 | 77.3472 | 0.2033 |
| 159.2265 | 5.19 | 4000 | 79.0953 | 0.2036 |
| 156.3967 | 5.84 | 4500 | 76.6855 | 0.2029 |
| 154.2743 | 6.49 | 5000 | 77.7785 | 0.2015 |
| 156.6497 | 7.14 | 5500 | 77.1220 | 0.2033 |
| 157.3038 | 7.79 | 6000 | 76.2926 | 0.2027 |
| 162.8151 | 8.44 | 6500 | 76.7602 | 0.2013 |
| 151.8613 | 9.09 | 7000 | 77.4777 | 0.2011 |
| 153.0225 | 9.74 | 7500 | 76.5206 | 0.2001 |
| 157.52 | 10.39 | 8000 | 76.1061 | 0.2006 |
| 145.0592 | 11.04 | 8500 | 76.7855 | 0.1992 |
| 150.0066 | 11.69 | 9000 | 76.0058 | 0.1988 |
| 146.8128 | 12.34 | 9500 | 76.2853 | 0.1987 |
| 146.9148 | 12.99 | 10000 | 76.3426 | 0.1979 |
### Framework versions
- Transformers 4.16.2
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.11.0
| {"language": ["pt"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "generated_from_trainer", "hf-asr-leaderboard", "mozilla-foundation/common_voice_7_0", "pt", "robust-speech-event"], "datasets": ["mozilla-foundation/common_voice_7_0"], "model-index": [{"name": "wav2vec2_base_10k_8khz_pt_cv7_2", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Common Voice 7", "type": "mozilla-foundation/common_voice_7_0", "args": "pt"}, "metrics": [{"type": "wer", "value": 36.9, "name": "Test WER"}, {"type": "cer", "value": 14.82, "name": "Test CER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Dev Data", "type": "speech-recognition-community-v2/dev_data", "args": "sv"}, "metrics": [{"type": "wer", "value": 40.53, "name": "Test WER"}, {"type": "cer", "value": 16.95, "name": "Test CER"}, {"type": "wer", "value": 37.15, "name": "Test WER"}]}, {"task": {"type": "automatic-speech-recognition", "name": "Automatic Speech Recognition"}, "dataset": {"name": "Robust Speech Event - Test Data", "type": "speech-recognition-community-v2/eval_data", "args": "pt"}, "metrics": [{"type": "wer", "value": 38.95, "name": "Test WER"}]}]}]} | lgris/wav2vec2_base_10k_8khz_pt_cv7_2 | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"hf-asr-leaderboard",
"mozilla-foundation/common_voice_7_0",
"pt",
"robust-speech-event",
"dataset:mozilla-foundation/common_voice_7_0",
"license:apache-2.0",
"model-index",
"end... | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #mozilla-foundation/common_voice_7_0 #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us
| wav2vec2\_base\_10k\_8khz\_pt\_cv7\_2
=====================================
This model is a fine-tuned version of lgris/seasr\_2022\_base\_10k\_8khz\_pt on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 76.3426
* Wer: 0.1979
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 1e-05
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 10000
### Training results
### Framework versions
* Transformers 4.16.2
* Pytorch 1.10.0+cu111
* Datasets 1.18.3
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 1e-05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #hf-asr-leaderboard #mozilla-foundation/common_voice_7_0 #pt #robust-speech-event #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #model-index #endpoints_compatible #region-us \n",
"### Trai... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wavlm-large-CORAA-pt-cv7
This model is a fine-tuned version of [lgris/WavLM-large-CORAA-pt](https://huggingface.co/lgris/WavLM-large-CORAA-pt) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2546
- Wer: 0.2261
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 5000
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 0.6029 | 0.13 | 100 | 0.3679 | 0.3347 |
| 0.5297 | 0.26 | 200 | 0.3516 | 0.3227 |
| 0.5134 | 0.39 | 300 | 0.3327 | 0.3167 |
| 0.4941 | 0.52 | 400 | 0.3281 | 0.3122 |
| 0.4816 | 0.65 | 500 | 0.3154 | 0.3102 |
| 0.4649 | 0.78 | 600 | 0.3199 | 0.3058 |
| 0.461 | 0.91 | 700 | 0.3047 | 0.2974 |
| 0.4613 | 1.04 | 800 | 0.3006 | 0.2900 |
| 0.4198 | 1.17 | 900 | 0.2951 | 0.2891 |
| 0.3864 | 1.3 | 1000 | 0.2989 | 0.2862 |
| 0.3963 | 1.43 | 1100 | 0.2932 | 0.2830 |
| 0.3953 | 1.56 | 1200 | 0.2936 | 0.2829 |
| 0.3962 | 1.69 | 1300 | 0.2952 | 0.2773 |
| 0.3811 | 1.82 | 1400 | 0.2915 | 0.2748 |
| 0.3736 | 1.95 | 1500 | 0.2839 | 0.2684 |
| 0.3507 | 2.08 | 1600 | 0.2914 | 0.2678 |
| 0.3277 | 2.21 | 1700 | 0.2895 | 0.2652 |
| 0.3344 | 2.34 | 1800 | 0.2843 | 0.2673 |
| 0.335 | 2.47 | 1900 | 0.2821 | 0.2635 |
| 0.3559 | 2.6 | 2000 | 0.2830 | 0.2599 |
| 0.3254 | 2.73 | 2100 | 0.2711 | 0.2577 |
| 0.3263 | 2.86 | 2200 | 0.2685 | 0.2546 |
| 0.3266 | 2.99 | 2300 | 0.2679 | 0.2521 |
| 0.3066 | 3.12 | 2400 | 0.2727 | 0.2526 |
| 0.2998 | 3.25 | 2500 | 0.2648 | 0.2537 |
| 0.2961 | 3.38 | 2600 | 0.2630 | 0.2519 |
| 0.3046 | 3.51 | 2700 | 0.2684 | 0.2506 |
| 0.3006 | 3.64 | 2800 | 0.2604 | 0.2492 |
| 0.2992 | 3.77 | 2900 | 0.2682 | 0.2508 |
| 0.2775 | 3.9 | 3000 | 0.2732 | 0.2440 |
| 0.2903 | 4.03 | 3100 | 0.2659 | 0.2427 |
| 0.2535 | 4.16 | 3200 | 0.2650 | 0.2433 |
| 0.2714 | 4.29 | 3300 | 0.2588 | 0.2394 |
| 0.2636 | 4.42 | 3400 | 0.2652 | 0.2434 |
| 0.2647 | 4.55 | 3500 | 0.2624 | 0.2371 |
| 0.2796 | 4.67 | 3600 | 0.2611 | 0.2373 |
| 0.2644 | 4.8 | 3700 | 0.2604 | 0.2341 |
| 0.2657 | 4.93 | 3800 | 0.2567 | 0.2331 |
| 0.2423 | 5.06 | 3900 | 0.2594 | 0.2322 |
| 0.2556 | 5.19 | 4000 | 0.2587 | 0.2323 |
| 0.2327 | 5.32 | 4100 | 0.2639 | 0.2299 |
| 0.2613 | 5.45 | 4200 | 0.2569 | 0.2310 |
| 0.2382 | 5.58 | 4300 | 0.2585 | 0.2298 |
| 0.2404 | 5.71 | 4400 | 0.2543 | 0.2287 |
| 0.2368 | 5.84 | 4500 | 0.2553 | 0.2286 |
| 0.2514 | 5.97 | 4600 | 0.2517 | 0.2279 |
| 0.2415 | 6.1 | 4700 | 0.2524 | 0.2270 |
| 0.2338 | 6.23 | 4800 | 0.2540 | 0.2265 |
| 0.219 | 6.36 | 4900 | 0.2549 | 0.2263 |
| 0.2428 | 6.49 | 5000 | 0.2546 | 0.2261 |
### Framework versions
- Transformers 4.16.1
- Pytorch 1.10.0+cu111
- Datasets 1.18.2
- Tokenizers 0.11.0
| {"language": ["pt"], "license": "apache-2.0", "tags": ["automatic-speech-recognition", "mozilla-foundation/common_voice_7_0", "generated_from_trainer", "pt"], "datasets": ["mozilla-foundation/common_voice_7_0"], "model-index": [{"name": "wavlm-large-CORAA-pt-cv7", "results": []}]} | lgris/wavlm-large-CORAA-pt-cv7 | null | [
"transformers",
"pytorch",
"tensorboard",
"wavlm",
"automatic-speech-recognition",
"mozilla-foundation/common_voice_7_0",
"generated_from_trainer",
"pt",
"dataset:mozilla-foundation/common_voice_7_0",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"pt"
] | TAGS
#transformers #pytorch #tensorboard #wavlm #automatic-speech-recognition #mozilla-foundation/common_voice_7_0 #generated_from_trainer #pt #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #endpoints_compatible #region-us
| wavlm-large-CORAA-pt-cv7
========================
This model is a fine-tuned version of lgris/WavLM-large-CORAA-pt on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 0.2546
* Wer: 0.2261
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0001
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 2
* total\_train\_batch\_size: 16
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 100
* training\_steps: 5000
### Training results
### Framework versions
* Transformers 4.16.1
* Pytorch 1.10.0+cu111
* Datasets 1.18.2
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 16\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=... | [
"TAGS\n#transformers #pytorch #tensorboard #wavlm #automatic-speech-recognition #mozilla-foundation/common_voice_7_0 #generated_from_trainer #pt #dataset-mozilla-foundation/common_voice_7_0 #license-apache-2.0 #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters w... |
token-classification | transformers |
ELECTRA-minuscule-discriminator
===============================
A ridiculously small ELECTRA discriminator model for testing purposes.
**THIS MODEL HAS NOT BEEN TRAINED, DO NOT EXPECT ANYThING OF IT.**
| {"language": "multilingual", "license": "cc0-1.0", "tags": ["electra", "testing", "minuscule"], "thumbnail": "url to a thumbnail used in social sharing"} | lgrobol/electra-minuscule-discriminator | null | [
"transformers",
"pytorch",
"electra",
"token-classification",
"testing",
"minuscule",
"multilingual",
"license:cc0-1.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"multilingual"
] | TAGS
#transformers #pytorch #electra #token-classification #testing #minuscule #multilingual #license-cc0-1.0 #autotrain_compatible #endpoints_compatible #region-us
|
ELECTRA-minuscule-discriminator
===============================
A ridiculously small ELECTRA discriminator model for testing purposes.
THIS MODEL HAS NOT BEEN TRAINED, DO NOT EXPECT ANYThING OF IT.
| [] | [
"TAGS\n#transformers #pytorch #electra #token-classification #testing #minuscule #multilingual #license-cc0-1.0 #autotrain_compatible #endpoints_compatible #region-us \n"
] |
fill-mask | transformers |
ELECTRA-minuscule-generator
===============================
A ridiculously small ELECTRA generator model for testing purposes.
**THIS MODEL HAS NOT BEEN TRAINED, DO NOT EXPECT ANYThING OF IT.**
| {"language": "multilingual", "license": "cc0-1.0", "tags": ["electra", "testing", "minuscule"]} | lgrobol/electra-minuscule-generator | null | [
"transformers",
"pytorch",
"safetensors",
"electra",
"fill-mask",
"testing",
"minuscule",
"multilingual",
"license:cc0-1.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"multilingual"
] | TAGS
#transformers #pytorch #safetensors #electra #fill-mask #testing #minuscule #multilingual #license-cc0-1.0 #autotrain_compatible #endpoints_compatible #region-us
|
ELECTRA-minuscule-generator
===============================
A ridiculously small ELECTRA generator model for testing purposes.
THIS MODEL HAS NOT BEEN TRAINED, DO NOT EXPECT ANYThING OF IT.
| [] | [
"TAGS\n#transformers #pytorch #safetensors #electra #fill-mask #testing #minuscule #multilingual #license-cc0-1.0 #autotrain_compatible #endpoints_compatible #region-us \n"
] |
fill-mask | transformers | FlauBERT-minuscule
==================
A ridiculously small model for testing purposes. | {} | lgrobol/flaubert-minuscule | null | [
"transformers",
"pytorch",
"safetensors",
"flaubert",
"fill-mask",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #safetensors #flaubert #fill-mask #autotrain_compatible #endpoints_compatible #region-us
| FlauBERT-minuscule
==================
A ridiculously small model for testing purposes. | [] | [
"TAGS\n#transformers #pytorch #safetensors #flaubert #fill-mask #autotrain_compatible #endpoints_compatible #region-us \n"
] |
fill-mask | transformers | RoBERTa-minuscule
==================
A ridiculously small model for testing purposes. | {} | lgrobol/roberta-minuscule | null | [
"transformers",
"pytorch",
"safetensors",
"roberta",
"fill-mask",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #safetensors #roberta #fill-mask #autotrain_compatible #endpoints_compatible #region-us
| RoBERTa-minuscule
==================
A ridiculously small model for testing purposes. | [] | [
"TAGS\n#transformers #pytorch #safetensors #roberta #fill-mask #autotrain_compatible #endpoints_compatible #region-us \n"
] |
text-generation | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# distilgpt2-finetuned-wikitext2
This model is a fine-tuned version of [distilgpt2](https://huggingface.co/distilgpt2) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 3.6423
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0
### Training results
| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 3.7602 | 1.0 | 2334 | 3.6669 |
| 3.633 | 2.0 | 4668 | 3.6455 |
| 3.6078 | 3.0 | 7002 | 3.6423 |
### Framework versions
- Transformers 4.16.2
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.11.0
| {"license": "apache-2.0", "tags": ["generated_from_trainer"], "model-index": [{"name": "distilgpt2-finetuned-wikitext2", "results": []}]} | lhbit20010120/distilgpt2-finetuned-wikitext2 | null | [
"transformers",
"pytorch",
"tensorboard",
"gpt2",
"text-generation",
"generated_from_trainer",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #tensorboard #gpt2 #text-generation #generated_from_trainer #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
| distilgpt2-finetuned-wikitext2
==============================
This model is a fine-tuned version of distilgpt2 on the None dataset.
It achieves the following results on the evaluation set:
* Loss: 3.6423
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 2e-05
* train\_batch\_size: 8
* eval\_batch\_size: 8
* seed: 42
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* num\_epochs: 3.0
### Training results
### Framework versions
* Transformers 4.16.2
* Pytorch 1.10.0+cu111
* Datasets 1.18.3
* Tokenizers 0.11.0
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 3.0",
"### Traini... | [
"TAGS\n#transformers #pytorch #tensorboard #gpt2 #text-generation #generated_from_trainer #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2... |
text-classification | transformers | Distilbert finetuned for Aspect-Based Sentiment Analysis (ABSA) with auxiliary sentence.
Fine-tuned using a dataset provided by NAVER for the CentraleSupélec NLP course.
```bibtex
@inproceedings{sun-etal-2019-utilizing,
title = "Utilizing {BERT} for Aspect-Based Sentiment Analysis via Constructing Auxiliary Sentence",
author = "Sun, Chi and
Huang, Luyao and
Qiu, Xipeng",
booktitle = "Proceedings of the 2019 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers)",
month = jun,
year = "2019",
address = "Minneapolis, Minnesota",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/N19-1035",
doi = "10.18653/v1/N19-1035",
pages = "380--385",
abstract = "Aspect-based sentiment analysis (ABSA), which aims to identify fine-grained opinion polarity towards a specific aspect, is a challenging subtask of sentiment analysis (SA). In this paper, we construct an auxiliary sentence from the aspect and convert ABSA to a sentence-pair classification task, such as question answering (QA) and natural language inference (NLI). We fine-tune the pre-trained model from BERT and achieve new state-of-the-art results on SentiHood and SemEval-2014 Task 4 datasets. The source codes are available at https://github.com/HSLCY/ABSA-BERT-pair.",
}
``` | {} | lhoestq/distilbert-base-uncased-finetuned-absa-as | null | [
"transformers",
"pytorch",
"distilbert",
"text-classification",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #distilbert #text-classification #autotrain_compatible #endpoints_compatible #region-us
| Distilbert finetuned for Aspect-Based Sentiment Analysis (ABSA) with auxiliary sentence.
Fine-tuned using a dataset provided by NAVER for the CentraleSupélec NLP course.
| [] | [
"TAGS\n#transformers #pytorch #distilbert #text-classification #autotrain_compatible #endpoints_compatible #region-us \n"
] |
fill-mask | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# char-bert-base-uncased
This model is a fine-tuned version of [char-bert-base-uncased/checkpoint-1840240](https://huggingface.co/char-bert-base-uncased/checkpoint-1840240) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1760
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 32
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 25
### Training results
| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:-------:|:---------------:|
| 0.8329 | 1.0 | 92012 | 0.4066 |
| 0.4066 | 2.0 | 184024 | 0.3223 |
| 0.3422 | 3.0 | 276036 | 0.2803 |
| 0.3044 | 4.0 | 368048 | 0.2560 |
| 0.2782 | 5.0 | 460060 | 0.2399 |
| 0.2593 | 6.0 | 552072 | 0.2265 |
| 0.2693 | 7.0 | 644084 | 0.2366 |
| 0.2559 | 8.0 | 736096 | 0.2228 |
| 0.2431 | 9.0 | 828108 | 0.2112 |
| 0.2334 | 10.0 | 920120 | 0.2103 |
| 0.2453 | 11.0 | 1012132 | 0.2164 |
| 0.2372 | 12.0 | 1104144 | 0.2113 |
| 0.2288 | 13.0 | 1196156 | 0.2004 |
| 0.2208 | 14.0 | 1288168 | 0.2002 |
| 0.2152 | 15.0 | 1380180 | 0.1941 |
| 0.2241 | 16.0 | 1472192 | 0.1940 |
| 0.2188 | 17.0 | 1564204 | 0.1954 |
| 0.2132 | 18.0 | 1656216 | 0.1968 |
| 0.2077 | 19.0 | 1748228 | 0.1887 |
| 0.2036 | 20.0 | 1840240 | 0.1863 |
| 0.2109 | 21.0 | 1932252 | 0.2009 |
| 0.2075 | 22.0 | 2024264 | 0.1840 |
| 0.2031 | 23.0 | 2116276 | 0.1884 |
| 0.1992 | 24.0 | 2208288 | 0.1902 |
| 0.196 | 25.0 | 2300300 | 0.1760 |
### Framework versions
- Transformers 4.20.1
- Pytorch 1.9.0+cu111
- Datasets 2.3.2
- Tokenizers 0.12.1
| {"tags": ["generated_from_trainer"], "model-index": [{"name": "char-bert-base-uncased", "results": []}]} | lhy/char-bert-base-uncased | null | [
"transformers",
"pytorch",
"bert",
"fill-mask",
"generated_from_trainer",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #bert #fill-mask #generated_from_trainer #autotrain_compatible #endpoints_compatible #region-us
| char-bert-base-uncased
======================
This model is a fine-tuned version of char-bert-base-uncased/checkpoint-1840240 on the None dataset.
It achieves the following results on the evaluation set:
* Loss: 0.1760
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 5e-05
* train\_batch\_size: 32
* eval\_batch\_size: 16
* seed: 42
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* num\_epochs: 25
### Training results
### Framework versions
* Transformers 4.20.1
* Pytorch 1.9.0+cu111
* Datasets 2.3.2
* Tokenizers 0.12.1
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 5e-05\n* train\\_batch\\_size: 32\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 25",
"### Train... | [
"TAGS\n#transformers #pytorch #bert #fill-mask #generated_from_trainer #autotrain_compatible #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 5e-05\n* train\\_batch\\_size: 32\n* eval\\_batch\\_size: 16\n* see... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-large-xls-r-300m-zh-CN-colab
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
### Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.13.3
- Tokenizers 0.10.3
| {"license": "apache-2.0", "tags": ["generated_from_trainer"], "datasets": ["common_voice"], "model-index": [{"name": "wav2vec2-large-xls-r-300m-zh-CN-colab", "results": []}]} | li666/wav2vec2-large-xls-r-300m-zh-CN-colab | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"dataset:common_voice",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-common_voice #license-apache-2.0 #endpoints_compatible #region-us
|
# wav2vec2-large-xls-r-300m-zh-CN-colab
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset.
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
### Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.13.3
- Tokenizers 0.10.3
| [
"# wav2vec2-large-xls-r-300m-zh-CN-colab\n\nThis model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset.",
"## Model description\n\nMore information needed",
"## Intended uses & limitations\n\nMore information needed",
"## Training and evaluation data\n\nMore information nee... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-common_voice #license-apache-2.0 #endpoints_compatible #region-us \n",
"# wav2vec2-large-xls-r-300m-zh-CN-colab\n\nThis model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_vo... |
feature-extraction | transformers |
# mBERT fine-tuned on English semantic role labeling
## Model description
This model is the [`bert-base-multilingual-cased`](https://huggingface.co/bert-base-multilingual-cased) fine-tuned on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-en_mbert-base")
model = AutoModel.from_pretrained("liaad/srl-en_mbert-base")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- The models were trained only for 5 epochs.
- The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
## Training procedure
The model was trained on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data. They were tested on the PropBank.Br data set as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt", "en"], "license": "apache-2.0", "tags": ["bert-base-multilingual-cased", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br", "CoNLL-2012"], "metrics": ["F1 Measure"]} | liaad/srl-en_mbert-base | null | [
"transformers",
"pytorch",
"tf",
"jax",
"bert",
"feature-extraction",
"bert-base-multilingual-cased",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"en",
"dataset:PropBank.Br",
"dataset:CoNLL-2012",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"r... | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt",
"en"
] | TAGS
#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-multilingual-cased #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| mBERT fine-tuned on English semantic role labeling
==================================================
Model description
-----------------
This model is the 'bert-base-multilingual-cased' fine-tuned on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* The models were trained only for 5 epochs.
* The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
Training procedure
------------------
The model was trained on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data. They were tested on the PropBank.Br data set as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* The models were trained only for 5 epochs.\n* The English data was preprocessed to match the Portuguese ... | [
"TAGS\n#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-multilingual-cased #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers... |
feature-extraction | transformers |
# XLM-R base fine-tuned on English semantic role labeling
## Model description
This model is the [`xlm-roberta-base`](https://huggingface.co/xlm-roberta-base) fine-tuned on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-en_xlmr-base")
model = AutoModel.from_pretrained("liaad/srl-en_xlmr-base")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
- The models were trained only for 5 epochs.
- The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
## Training procedure
The models were trained on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data. They were tested on the PropBank.Br data set as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt", "en"], "license": "apache-2.0", "tags": ["xlm-roberta-base", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br", "CoNLL-2012"], "metrics": ["F1 Measure"]} | liaad/srl-en_xlmr-base | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-base",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"en",
"dataset:PropBank.Br",
"dataset:CoNLL-2012",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt",
"en"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-base #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R base fine-tuned on English semantic role labeling
=======================================================
Model description
-----------------
This model is the 'xlm-roberta-base' fine-tuned on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
* The models were trained only for 5 epochs.
* The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
Training procedure
------------------
The models were trained on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data. They were tested on the PropBank.Br data set as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-base #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of th... |
feature-extraction | transformers |
# XLM-R large fine-tuned on English semantic role labeling
## Model description
This model is the [`xlm-roberta-large`](https://huggingface.co/xlm-roberta-large) fine-tuned on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-en_xlmr-large")
model = AutoModel.from_pretrained("liaad/srl-en_xlmr-large")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
- The models were trained only for 5 epochs.
- The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
## Training procedure
The models were trained on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data. They were tested on the PropBank.Br data set as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt", "en"], "license": "apache-2.0", "tags": ["xlm-roberta-large", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br", "CoNLL-2012"], "metrics": ["F1 Measure"]} | liaad/srl-en_xlmr-large | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-large",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"en",
"dataset:PropBank.Br",
"dataset:CoNLL-2012",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt",
"en"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R large fine-tuned on English semantic role labeling
========================================================
Model description
-----------------
This model is the 'xlm-roberta-large' fine-tuned on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
* The models were trained only for 5 epochs.
* The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
Training procedure
------------------
The models were trained on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data. They were tested on the PropBank.Br data set as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of t... |
feature-extraction | transformers |
# mBERT base fine-tune in English and Portuguese semantic role labeling
## Model description
This model is the [`bert-base-multilingual-cased`](https://huggingface.co/bert-base-multilingual-cased) fine-tuned first on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-enpt_mbert-base")
model = AutoModel.from_pretrained("liaad/srl-enpt_mbert-base")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
## Training procedure
The model was first fine-tuned on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data; then it was fine-tuned in the PropBank.Br dataset using 10-fold Cross-Validation. The resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt", "en"], "license": "apache-2.0", "tags": ["bert-base-multilingual-cased", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br", "CoNLL-2012"], "metrics": ["F1 Measure"]} | liaad/srl-enpt_mbert-base | null | [
"transformers",
"pytorch",
"tf",
"jax",
"bert",
"feature-extraction",
"bert-base-multilingual-cased",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"en",
"dataset:PropBank.Br",
"dataset:CoNLL-2012",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"r... | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt",
"en"
] | TAGS
#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-multilingual-cased #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| mBERT base fine-tune in English and Portuguese semantic role labeling
=====================================================================
Model description
-----------------
This model is the 'bert-base-multilingual-cased' fine-tuned first on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
Training procedure
------------------
The model was first fine-tuned on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data; then it was fine-tuned in the PropBank.Br dataset using 10-fold Cross-Validation. The resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* The English data was preprocessed to match the Portuguese data, so there are some differences in role at... | [
"TAGS\n#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-multilingual-cased #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers... |
feature-extraction | transformers |
# XLM-R base fine-tune in English and Portuguese semantic role labeling
## Model description
This model is the [`xlm-roberta-base`](https://huggingface.co/xlm-roberta-base) fine-tuned first on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-enpt_xlmr-base")
model = AutoModel.from_pretrained("liaad/srl-enpt_xlmr-base")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
- The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
## Training procedure
The model was first fine-tuned on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data; then it was fine-tuned in the PropBank.Br dataset using 10-fold Cross-Validation. The resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt", "en"], "license": "apache-2.0", "tags": ["xlm-roberta-base", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br", "CoNLL-2012"], "metrics": ["F1 Measure"]} | liaad/srl-enpt_xlmr-base | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-base",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"en",
"dataset:PropBank.Br",
"dataset:CoNLL-2012",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt",
"en"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-base #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R base fine-tune in English and Portuguese semantic role labeling
=====================================================================
Model description
-----------------
This model is the 'xlm-roberta-base' fine-tuned first on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
* The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
Training procedure
------------------
The model was first fine-tuned on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data; then it was fine-tuned in the PropBank.Br dataset using 10-fold Cross-Validation. The resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-base #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of th... |
feature-extraction | transformers |
# XLM-R large fine-tuned in English and Portuguese semantic role labeling
## Model description
This model is the [`xlm-roberta-large`](https://huggingface.co/xlm-roberta-large) fine-tuned first on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-enpt_xlmr-large")
model = AutoModel.from_pretrained("liaad/srl-enpt_xlmr-large")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
- The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
## Training procedure
The model was first fine-tuned on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data; then it was fine-tuned in the PropBank.Br dataset using 10-fold Cross-Validation. The resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt", "en"], "license": "apache-2.0", "tags": ["xlm-roberta-large", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br", "CoNLL-2012"], "metrics": ["F1 Measure"]} | liaad/srl-enpt_xlmr-large | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-large",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"en",
"dataset:PropBank.Br",
"dataset:CoNLL-2012",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt",
"en"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R large fine-tuned in English and Portuguese semantic role labeling
=======================================================================
Model description
-----------------
This model is the 'xlm-roberta-large' fine-tuned first on the English CoNLL formatted OntoNotes v5.0 semantic role labeling data and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
* The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
Training procedure
------------------
The model was first fine-tuned on the CoNLL-2012 dataset, preprocessed to match the Portuguese PropBank.Br data; then it was fine-tuned in the PropBank.Br dataset using 10-fold Cross-Validation. The resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #multilingual #pt #en #dataset-PropBank.Br #dataset-CoNLL-2012 #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of t... |
feature-extraction | transformers |
# BERTimbau base fine-tuned on Portuguese semantic role labeling
## Model description
This model is the [`neuralmind/bert-base-portuguese-cased`](https://huggingface.co/neuralmind/bert-base-portuguese-cased) fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-pt_bertimbau-base")
model = AutoModel.from_pretrained("liaad/srl-pt_bertimbau-base")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Training procedure
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt"], "license": "apache-2.0", "tags": ["bert-base-portuguese-cased", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br"], "metrics": ["F1 Measure"]} | liaad/srl-pt_bertimbau-base | null | [
"transformers",
"pytorch",
"tf",
"jax",
"bert",
"feature-extraction",
"bert-base-portuguese-cased",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"dataset:PropBank.Br",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt"
] | TAGS
#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-portuguese-cased #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| BERTimbau base fine-tuned on Portuguese semantic role labeling
==============================================================
Model description
-----------------
This model is the 'neuralmind/bert-base-portuguese-cased' fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
Training procedure
------------------
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.\n\n\nTraining procedure\n------------------\n\n\nThe model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 res... | [
"TAGS\n#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-portuguese-cased #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\n\... |
feature-extraction | transformers |
# BERTimbau large fine-tuned on Portuguese semantic role labeling
## Model description
This model is the [`neuralmind/bert-large-portuguese-cased`](https://huggingface.co/neuralmind/bert-large-portuguese-cased) fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-pt_bertimbau-large")
model = AutoModel.from_pretrained("liaad/srl-pt_bertimbau-large")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Training procedure
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt"], "license": "apache-2.0", "tags": ["bert-large-portuguese-cased", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br"], "metrics": ["F1 Measure"]} | liaad/srl-pt_bertimbau-large | null | [
"transformers",
"pytorch",
"tf",
"jax",
"bert",
"feature-extraction",
"bert-large-portuguese-cased",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"dataset:PropBank.Br",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt"
] | TAGS
#transformers #pytorch #tf #jax #bert #feature-extraction #bert-large-portuguese-cased #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| BERTimbau large fine-tuned on Portuguese semantic role labeling
===============================================================
Model description
-----------------
This model is the 'neuralmind/bert-large-portuguese-cased' fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
Training procedure
------------------
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.\n\n\nTraining procedure\n------------------\n\n\nThe model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 res... | [
"TAGS\n#transformers #pytorch #tf #jax #bert #feature-extraction #bert-large-portuguese-cased #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\n... |
feature-extraction | transformers |
# mBERT fine-tuned on Portuguese semantic role labeling
## Model description
This model is the [`bert-base-multilingual-cased`](https://huggingface.co/bert-base-multilingual-cased) fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-pt_mbert-base")
model = AutoModel.from_pretrained("liaad/srl-pt_mbert-base")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Training procedure
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt"], "license": "apache-2.0", "tags": ["bert-base-multilingual-cased", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br"], "metrics": ["F1 Measure"]} | liaad/srl-pt_mbert-base | null | [
"transformers",
"pytorch",
"tf",
"jax",
"bert",
"feature-extraction",
"bert-base-multilingual-cased",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"dataset:PropBank.Br",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt"
] | TAGS
#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-multilingual-cased #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| mBERT fine-tuned on Portuguese semantic role labeling
=====================================================
Model description
-----------------
This model is the 'bert-base-multilingual-cased' fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
Training procedure
------------------
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.\n\n\nTraining procedure\n------------------\n\n\nThe model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 res... | [
"TAGS\n#transformers #pytorch #tf #jax #bert #feature-extraction #bert-base-multilingual-cased #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\... |
feature-extraction | transformers |
# XLM-R base fine-tuned on Portuguese semantic role labeling
## Model description
This model is the [`xlm-roberta-base`](https://huggingface.co/xlm-roberta-base) fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-pt_xlmr-base")
model = AutoModel.from_pretrained("liaad/srl-pt_xlmr-base")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
## Training procedure
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt"], "license": "apache-2.0", "tags": ["xlm-roberta-base", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br"], "metrics": ["F1 Measure"]} | liaad/srl-pt_xlmr-base | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-base",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"dataset:PropBank.Br",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-base #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R base fine-tuned on Portuguese semantic role labeling
==========================================================
Model description
-----------------
This model is the 'xlm-roberta-base' fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
Training procedure
------------------
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-base #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use th... |
feature-extraction | transformers |
# XLM-R large fine-tuned on Portuguese semantic role labeling
## Model description
This model is the [`xlm-roberta-large`](https://huggingface.co/xlm-roberta-large) fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/srl-pt_xlmr-large")
model = AutoModel.from_pretrained("liaad/srl-pt_xlmr-large")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
## Training procedure
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt"], "license": "apache-2.0", "tags": ["xlm-roberta-large", "semantic role labeling", "finetuned"], "datasets": ["PropBank.Br"], "metrics": ["F1 Measure"]} | liaad/srl-pt_xlmr-large | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-large",
"semantic role labeling",
"finetuned",
"multilingual",
"pt",
"dataset:PropBank.Br",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R large fine-tuned on Portuguese semantic role labeling
===========================================================
Model description
-----------------
This model is the 'xlm-roberta-large' fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
Training procedure
------------------
The model was trained on the PropBank.Br datasets, using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #multilingual #pt #dataset-PropBank.Br #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use t... |
feature-extraction | transformers |
# XLM-R large fine-tuned in Portuguese Universal Dependencies and English and Portuguese semantic role labeling
## Model description
This model is the [`xlm-roberta-large`](https://huggingface.co/xlm-roberta-large) fine-tuned first on the Universal Dependencies Portuguese dataset, then fine-tuned on the CoNLL formatted OntoNotes v5.0 and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/ud_srl-enpt_xlmr-large")
model = AutoModel.from_pretrained("liaad/ud_srl-enpt_xlmr-large")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
- The model was trained only for 10 epochs in the Universal Dependencies dataset.
- The model was trained only for 5 epochs in the CoNLL formatted OntoNotes v5.0.
- The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
## Training procedure
The model was trained on the Universal Dependencies Portuguese dataset; then on the CoNLL formatted OntoNotes v5.0; then on Portuguese semantic role labeling data (PropBank.Br) using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt", "en"], "license": "apache-2.0", "tags": ["xlm-roberta-large", "semantic role labeling", "finetuned", "dependency parsing"], "datasets": ["PropBank.Br", "CoNLL-2012", "Universal Dependencies"], "metrics": "f1"} | liaad/ud_srl-enpt_xlmr-large | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-large",
"semantic role labeling",
"finetuned",
"dependency parsing",
"multilingual",
"pt",
"en",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt",
"en"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #dependency parsing #multilingual #pt #en #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R large fine-tuned in Portuguese Universal Dependencies and English and Portuguese semantic role labeling
=============================================================================================================
Model description
-----------------
This model is the 'xlm-roberta-large' fine-tuned first on the Universal Dependencies Portuguese dataset, then fine-tuned on the CoNLL formatted OntoNotes v5.0 and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
* The model was trained only for 10 epochs in the Universal Dependencies dataset.
* The model was trained only for 5 epochs in the CoNLL formatted OntoNotes v5.0.
* The English data was preprocessed to match the Portuguese data, so there are some differences in role attributions and some roles were removed from the data.
Training procedure
------------------
The model was trained on the Universal Dependencies Portuguese dataset; then on the CoNLL formatted OntoNotes v5.0; then on Portuguese semantic role labeling data (PropBank.Br) using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #dependency parsing #multilingual #pt #en #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo us... |
feature-extraction | transformers |
# BERTimbau large fine-tune in Portuguese Universal Dependencies and semantic role labeling
## Model description
This model is the [`neuralmind/bert-large-portuguese-cased`](https://huggingface.co/neuralmind/bert-large-portuguese-cased) fine-tuned first on the Universal Dependencies Portuguese dataset and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/ud_srl-pt_bertimbau-large")
model = AutoModel.from_pretrained("liaad/ud_srl-pt_bertimbau-large")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- The model was trained only for 10 epochs in the Universal Dependencies dataset.
## Training procedure
The model was trained on the Universal Dependencies Portuguese dataset; then on the CoNLL formatted OntoNotes v5.0; then on Portuguese semantic role labeling data (PropBank.Br) using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt"], "license": "apache-2.0", "tags": ["bert-large-portuguese-cased", "semantic role labeling", "finetuned", "dependency parsing"], "datasets": ["PropBank.Br", "CoNLL-2012", "Universal Dependencies"], "metrics": ["F1 Measure"]} | liaad/ud_srl-pt_bertimbau-large | null | [
"transformers",
"pytorch",
"tf",
"jax",
"bert",
"feature-extraction",
"bert-large-portuguese-cased",
"semantic role labeling",
"finetuned",
"dependency parsing",
"multilingual",
"pt",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt"
] | TAGS
#transformers #pytorch #tf #jax #bert #feature-extraction #bert-large-portuguese-cased #semantic role labeling #finetuned #dependency parsing #multilingual #pt #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| BERTimbau large fine-tune in Portuguese Universal Dependencies and semantic role labeling
=========================================================================================
Model description
-----------------
This model is the 'neuralmind/bert-large-portuguese-cased' fine-tuned first on the Universal Dependencies Portuguese dataset and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* The model was trained only for 10 epochs in the Universal Dependencies dataset.
Training procedure
------------------
The model was trained on the Universal Dependencies Portuguese dataset; then on the CoNLL formatted OntoNotes v5.0; then on Portuguese semantic role labeling data (PropBank.Br) using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* The model was trained only for 10 epochs in the Universal Dependencies dataset.\n\n\nTraining procedure\... | [
"TAGS\n#transformers #pytorch #tf #jax #bert #feature-extraction #bert-large-portuguese-cased #semantic role labeling #finetuned #dependency parsing #multilingual #pt #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\n\... |
feature-extraction | transformers |
# XLM-R large fine-tune in Portuguese Universal Dependencies and semantic role labeling
## Model description
This model is the [`xlm-roberta-large`](https://huggingface.co/xlm-roberta-large) fine-tuned first on the Universal Dependencies Portuguese dataset and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
* [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
* [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
* [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
* [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
* [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
* [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
* [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
* [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
* [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
* [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
* [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
* [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
* [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Intended uses & limitations
#### How to use
To use the transformers portion of this model:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("liaad/ud_srl-pt_xlmr-large")
model = AutoModel.from_pretrained("liaad/ud_srl-pt_xlmr-large")
```
To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
#### Limitations and bias
- This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
- The model was trained only for 10 epochs in the Universal Dependencies dataset.
## Training procedure
The model was trained on the Universal Dependencies Portuguese dataset; then on the CoNLL formatted OntoNotes v5.0; then on Portuguese semantic role labeling data (PropBank.Br) using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
## Eval results
| Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
| --------------- | ------ | ----- |
| `srl-pt_bertimbau-base` | 76.30 | 73.33 |
| `srl-pt_bertimbau-large` | 77.42 | 74.85 |
| `srl-pt_xlmr-base` | 75.22 | 72.82 |
| `srl-pt_xlmr-large` | 77.59 | 73.84 |
| `srl-pt_mbert-base` | 72.76 | 66.89 |
| `srl-en_xlmr-base` | 66.59 | 65.24 |
| `srl-en_xlmr-large` | 67.60 | 64.94 |
| `srl-en_mbert-base` | 63.07 | 58.56 |
| `srl-enpt_xlmr-base` | 76.50 | 73.74 |
| `srl-enpt_xlmr-large` | **78.22** | 74.55 |
| `srl-enpt_mbert-base` | 74.88 | 69.19 |
| `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
| `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
| `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
### BibTeX entry and citation info
```bibtex
@misc{oliveira2021transformers,
title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
year={2021},
eprint={2101.01213},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["multilingual", "pt"], "license": "apache-2.0", "tags": ["xlm-roberta-large", "semantic role labeling", "finetuned", "dependency parsing"], "datasets": ["PropBank.Br", "CoNLL-2012", "Universal Dependencies"], "metrics": ["F1 Measure"]} | liaad/ud_srl-pt_xlmr-large | null | [
"transformers",
"pytorch",
"xlm-roberta",
"feature-extraction",
"xlm-roberta-large",
"semantic role labeling",
"finetuned",
"dependency parsing",
"multilingual",
"pt",
"arxiv:2101.01213",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2101.01213"
] | [
"multilingual",
"pt"
] | TAGS
#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #dependency parsing #multilingual #pt #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us
| XLM-R large fine-tune in Portuguese Universal Dependencies and semantic role labeling
=====================================================================================
Model description
-----------------
This model is the 'xlm-roberta-large' fine-tuned first on the Universal Dependencies Portuguese dataset and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
* liaad/srl-pt\_bertimbau-base
* liaad/srl-pt\_bertimbau-large
* liaad/srl-pt\_xlmr-base
* liaad/srl-pt\_xlmr-large
* liaad/srl-pt\_mbert-base
* liaad/srl-en\_xlmr-base
* liaad/srl-en\_xlmr-large
* liaad/srl-en\_mbert-base
* liaad/srl-enpt\_xlmr-base
* liaad/srl-enpt\_xlmr-large
* liaad/srl-enpt\_mbert-base
* liaad/ud\_srl-pt\_bertimbau-large
* liaad/ud\_srl-pt\_xlmr-large
* liaad/ud\_srl-enpt\_xlmr-large
For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Intended uses & limitations
---------------------------
#### How to use
To use the transformers portion of this model:
To use the full SRL model (transformers portion + a decoding layer), refer to the project's github.
#### Limitations and bias
* This model does not include a Tensorflow version. This is because the "type\_vocab\_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
* The model was trained only for 10 epochs in the Universal Dependencies dataset.
Training procedure
------------------
The model was trained on the Universal Dependencies Portuguese dataset; then on the CoNLL formatted OntoNotes v5.0; then on Portuguese semantic role labeling data (PropBank.Br) using 10-fold Cross-Validation. The 10 resulting models were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the project's github.
Eval results
------------
Model Name: 'srl-pt\_bertimbau-base', F1 CV PropBank.Br (in domain): 76.30, F1 Buscapé (out of domain): 73.33
Model Name: 'srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.42, F1 Buscapé (out of domain): 74.85
Model Name: 'srl-pt\_xlmr-base', F1 CV PropBank.Br (in domain): 75.22, F1 Buscapé (out of domain): 72.82
Model Name: 'srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.59, F1 Buscapé (out of domain): 73.84
Model Name: 'srl-pt\_mbert-base', F1 CV PropBank.Br (in domain): 72.76, F1 Buscapé (out of domain): 66.89
Model Name: 'srl-en\_xlmr-base', F1 CV PropBank.Br (in domain): 66.59, F1 Buscapé (out of domain): 65.24
Model Name: 'srl-en\_xlmr-large', F1 CV PropBank.Br (in domain): 67.60, F1 Buscapé (out of domain): 64.94
Model Name: 'srl-en\_mbert-base', F1 CV PropBank.Br (in domain): 63.07, F1 Buscapé (out of domain): 58.56
Model Name: 'srl-enpt\_xlmr-base', F1 CV PropBank.Br (in domain): 76.50, F1 Buscapé (out of domain): 73.74
Model Name: 'srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 78.22, F1 Buscapé (out of domain): 74.55
Model Name: 'srl-enpt\_mbert-base', F1 CV PropBank.Br (in domain): 74.88, F1 Buscapé (out of domain): 69.19
Model Name: 'ud\_srl-pt\_bertimbau-large', F1 CV PropBank.Br (in domain): 77.53, F1 Buscapé (out of domain): 74.49
Model Name: 'ud\_srl-pt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.69, F1 Buscapé (out of domain): 74.91
Model Name: 'ud\_srl-enpt\_xlmr-large', F1 CV PropBank.Br (in domain): 77.97, F1 Buscapé (out of domain): 75.05
### BibTeX entry and citation info
| [
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use the full SRL model (transformers portion + a decoding layer), refer to the project's github.",
"#### Limitations and bias\n\n\n* This model does not include a Tensorflow version. This is because the \"type\\_vocab\\_size\" in this mo... | [
"TAGS\n#transformers #pytorch #xlm-roberta #feature-extraction #xlm-roberta-large #semantic role labeling #finetuned #dependency parsing #multilingual #pt #arxiv-2101.01213 #license-apache-2.0 #endpoints_compatible #region-us \n",
"#### How to use\n\n\nTo use the transformers portion of this model:\n\n\nTo use th... |
text-classification | transformers |
# liam168/c2-roberta-base-finetuned-dianping-chinese
## Model description
用中文对话情绪语料训练的模型，2分类：乐观和悲观。
## Overview
- **Language model**: BertForSequenceClassification
- **Model size**: 410M
- **Language**: Chinese
## Example
```python
>>> from transformers import AutoModelForSequenceClassification , AutoTokenizer, pipeline
>>> model_name = "liam168/c2-roberta-base-finetuned-dianping-chinese"
>>> class_num = 2
>>> ts_texts = ["我喜欢下雨。", "我讨厌他."]
>>> model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=class_num)
>>> tokenizer = AutoTokenizer.from_pretrained(model_name)
>>> classifier = pipeline('sentiment-analysis', model=model, tokenizer=tokenizer)
>>> classifier(ts_texts[0])
>>> classifier(ts_texts[1])
[{'label': 'positive', 'score': 0.9973447918891907}]
[{'label': 'negative', 'score': 0.9972558617591858}]
```
| {"language": "zh", "widget": [{"text": "\u6211\u559c\u6b22\u4e0b\u96e8\u3002"}, {"text": "\u6211\u8ba8\u538c\u4ed6\u3002"}]} | liam168/c2-roberta-base-finetuned-dianping-chinese | null | [
"transformers",
"pytorch",
"bert",
"text-classification",
"zh",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"zh"
] | TAGS
#transformers #pytorch #bert #text-classification #zh #autotrain_compatible #endpoints_compatible #has_space #region-us
|
# liam168/c2-roberta-base-finetuned-dianping-chinese
## Model description
用中文对话情绪语料训练的模型，2分类：乐观和悲观。
## Overview
- Language model: BertForSequenceClassification
- Model size: 410M
- Language: Chinese
## Example
| [
"# liam168/c2-roberta-base-finetuned-dianping-chinese",
"## Model description\n\n用中文对话情绪语料训练的模型，2分类：乐观和悲观。",
"## Overview\n\n- Language model: BertForSequenceClassification\n- Model size: 410M\n- Language: Chinese",
"## Example"
] | [
"TAGS\n#transformers #pytorch #bert #text-classification #zh #autotrain_compatible #endpoints_compatible #has_space #region-us \n",
"# liam168/c2-roberta-base-finetuned-dianping-chinese",
"## Model description\n\n用中文对话情绪语料训练的模型，2分类：乐观和悲观。",
"## Overview\n\n- Language model: BertForSequenceClassification\n- Mo... |
text-classification | transformers |
# liam168/c4-zh-distilbert-base-uncased
## Model description
用 ["女性","体育","文学","校园"]4类数据训练的分类模型。
## Overview
- **Language model**: DistilBERT
- **Model size**: 280M
- **Language**: Chinese
## Example
```python
>>> from transformers import DistilBertForSequenceClassification , AutoTokenizer, pipeline
>>> model_name = "liam168/c4-zh-distilbert-base-uncased"
>>> class_num = 4
>>> ts_texts = ["女人做得越纯粹，皮肤和身材就越好", "我喜欢篮球"]
>>> model = DistilBertForSequenceClassification.from_pretrained(model_name, num_labels=class_num)
>>> tokenizer = AutoTokenizer.from_pretrained(model_name)
>>> classifier = pipeline('sentiment-analysis', model=model, tokenizer=tokenizer)
>>> classifier(ts_texts[0])
>>> classifier(ts_texts[1])
[{'label': 'Female', 'score': 0.9137857556343079}]
[{'label': 'Sports', 'score': 0.8206522464752197}]
```
| {"language": "zh", "license": "apache-2.0", "tags": ["exbert"], "widget": [{"text": "\u5973\u4eba\u505a\u5f97\u8d8a\u7eaf\u7cb9\uff0c\u76ae\u80a4\u548c\u8eab\u6750\u5c31\u8d8a\u597d"}, {"text": "\u6211\u559c\u6b22\u7bee\u7403"}]} | liam168/c4-zh-distilbert-base-uncased | null | [
"transformers",
"pytorch",
"distilbert",
"text-classification",
"exbert",
"zh",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"zh"
] | TAGS
#transformers #pytorch #distilbert #text-classification #exbert #zh #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
|
# liam168/c4-zh-distilbert-base-uncased
## Model description
用 ["女性","体育","文学","校园"]4类数据训练的分类模型。
## Overview
- Language model: DistilBERT
- Model size: 280M
- Language: Chinese
## Example
| [
"# liam168/c4-zh-distilbert-base-uncased",
"## Model description\n\n用 [\"女性\",\"体育\",\"文学\",\"校园\"]4类数据训练的分类模型。",
"## Overview\n\n- Language model: DistilBERT\n- Model size: 280M \n- Language: Chinese",
"## Example"
] | [
"TAGS\n#transformers #pytorch #distilbert #text-classification #exbert #zh #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"# liam168/c4-zh-distilbert-base-uncased",
"## Model description\n\n用 [\"女性\",\"体育\",\"文学\",\"校园\"]4类数据训练的分类模型。",
"## Overview\n\n- Language model: DistilB... |
text-generation | transformers |
# liam168/chat-DialoGPT-small-en
## Model description
用英文聊天数据训练的模型；
### How to use
Now we are ready to try out how the model works as a chatting partner!
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
mode_name = 'liam168/chat-DialoGPT-small-en'
tokenizer = AutoTokenizer.from_pretrained(mode_name)
model = AutoModelForCausalLM.from_pretrained(mode_name)
# Let's chat for 5 lines
for step in range(5):
# encode the new user input, add the eos_token and return a tensor in Pytorch
new_user_input_ids = tokenizer.encode(input(">> User:") + tokenizer.eos_token, return_tensors='pt')
# append the new user input tokens to the chat history
bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
# generated a response while limiting the total chat history to 1000 tokens,
chat_history_ids = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)
# pretty print last ouput tokens from bot
print("Answer: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))
```
| {"language": "en", "license": "apache-2.0", "widget": [{"text": "I got a surprise for you, Morty."}]} | liam168/chat-DialoGPT-small-en | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"en",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en"
] | TAGS
#transformers #pytorch #gpt2 #text-generation #en #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
# liam168/chat-DialoGPT-small-en
## Model description
用英文聊天数据训练的模型；
### How to use
Now we are ready to try out how the model works as a chatting partner!
| [
"# liam168/chat-DialoGPT-small-en",
"## Model description\n\n用英文聊天数据训练的模型；",
"### How to use\n\nNow we are ready to try out how the model works as a chatting partner!"
] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #en #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"# liam168/chat-DialoGPT-small-en",
"## Model description\n\n用英文聊天数据训练的模型；",
"### How to use\n\nNow we are ready to try out how the model works as... |
text-generation | transformers |
# liam168/chat-DialoGPT-small-zh
## Model description
用中文聊天数据训练的模型；
### How to use
Now we are ready to try out how the model works as a chatting partner!
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
mode_name = 'liam168/chat-DialoGPT-small-zh'
tokenizer = AutoTokenizer.from_pretrained(mode_name)
model = AutoModelForCausalLM.from_pretrained(mode_name)
# Let's chat for 5 lines
for step in range(5):
# encode the new user input, add the eos_token and return a tensor in Pytorch
new_user_input_ids = tokenizer.encode(input(">> User:") + tokenizer.eos_token, return_tensors='pt')
# append the new user input tokens to the chat history
bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
# generated a response while limiting the total chat history to 1000 tokens,
chat_history_ids = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)
# pretty print last ouput tokens from bot
print("Answer: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))
```
| {"language": "zh", "license": "apache-2.0", "widget": [{"text": "\u4f60\u4eec\u5bbf\u820d\u90fd\u662f\u8fd9\u4e48\u5389\u5bb3\u7684\u4eba\u5417"}]} | liam168/chat-DialoGPT-small-zh | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"zh",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"zh"
] | TAGS
#transformers #pytorch #gpt2 #text-generation #zh #license-apache-2.0 #autotrain_compatible #endpoints_compatible #has_space #text-generation-inference #region-us
|
# liam168/chat-DialoGPT-small-zh
## Model description
用中文聊天数据训练的模型；
### How to use
Now we are ready to try out how the model works as a chatting partner!
| [
"# liam168/chat-DialoGPT-small-zh",
"## Model description\n\n用中文聊天数据训练的模型；",
"### How to use\n\nNow we are ready to try out how the model works as a chatting partner!"
] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #zh #license-apache-2.0 #autotrain_compatible #endpoints_compatible #has_space #text-generation-inference #region-us \n",
"# liam168/chat-DialoGPT-small-zh",
"## Model description\n\n用中文聊天数据训练的模型；",
"### How to use\n\nNow we are ready to try out how the mod... |
text-generation | transformers |
# gen-gpt2-medium-chinese
# Overview
- **Language model**: GPT2-Medium
- **Model size**: 68M
- **Language**: Chinese
# Example
```python
from transformers import TFGPT2LMHeadModel,AutoTokenizer
from transformers import TextGenerationPipeline
mode_name = 'liam168/gen-gpt2-medium-chinese'
tokenizer = AutoTokenizer.from_pretrained(mode_name)
model = TFGPT2LMHeadModel.from_pretrained(mode_name)
text_generator = TextGenerationPipeline(model, tokenizer)
print(text_generator("晓日千红", max_length=64, do_sample=True))
print(text_generator("加餐小语", max_length=50, do_sample=False))
```
输出
```text
[{'generated_text': '晓日千红 独 远 客 。 孤 夜 云 云 梦 到 冷 。 著 剩 笑 、 人 远 。 灯 啼 鸦 最 回 吟 。 望 ， 枕 付 孤 灯 、 客 。 对 梅 残 照 偏 相 思 ， 玉 弦 语 。 翠 台 新 妆 、 沉 、 登 临 水 。 空'}]
[{'generated_text': '加餐小语 有 有 骨 ， 有 人 诗 成 自 远 诗 。 死 了 自 喜 乐 ， 独 撑 天 下 诗 事 小 诗 柴 。 桃 花 谁 知 何 处 何 处 高 吟 诗 从 今 死 火 ， 此 事'}]
```
| {"language": "zh", "widget": [{"text": "\u6653\u65e5\u5343\u7ea2"}, {"text": "\u957f\u8857\u8e9e\u8e40"}]} | liam168/gen-gpt2-medium-chinese | null | [
"transformers",
"pytorch",
"tf",
"gpt2",
"text-generation",
"zh",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"zh"
] | TAGS
#transformers #pytorch #tf #gpt2 #text-generation #zh #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
# gen-gpt2-medium-chinese
# Overview
- Language model: GPT2-Medium
- Model size: 68M
- Language: Chinese
# Example
输出
| [
"# gen-gpt2-medium-chinese",
"# Overview\n\n- Language model: GPT2-Medium\n- Model size: 68M \n- Language: Chinese",
"# Example\n\n\n输出"
] | [
"TAGS\n#transformers #pytorch #tf #gpt2 #text-generation #zh #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"# gen-gpt2-medium-chinese",
"# Overview\n\n- Language model: GPT2-Medium\n- Model size: 68M \n- Language: Chinese",
"# Example\n\n\n输出"
] |
question-answering | transformers | # Chinese RoBERTa-Base Model for QA
## Model description
用中文预料微调的QA模型.
## Overview
- **Language model**: RoBERTa-Base
- **Model size**: 400M
- **Language**: Chinese
## How to use
You can use the model directly with a pipeline for extractive question answering:
```python
>>> from transformers import AutoModelForQuestionAnswering,AutoTokenizer,pipeline
>>> context = '卡利亚·基拔（，）生于英国汉默史密斯，是一名英格兰籍职业足球员，于2010年夏季约满离开母会阿仙奴。直到2005/06年，基拔通常在阿仙奴的青年后备队效力。他在首次在2005年11月29日的联赛杯赛事上场，并于12月7日，在一个欧洲联赛冠军杯比赛对阿积士，作为替代左后卫，入替受伤的劳伦。2006年7月21日阿仙奴宣布，将基拔出借卡迪夫城整个2006-07赛季，其后转借给修安联。2008年1月3日返回阿仙奴授予46号码。2008年2月11日，阿仙奴的英超联赛比赛中对布莱克本作为后备球员。但2008年7月10日，基拔被出借莱斯特城的一个赛季之久。2009年3月3日主场对-{zh-hans:斯托克港;zh-hk:史托港}-，开赛后仅两分钟，基拔的传中球「挞Q」却直入网角，是他个人首个入球。基拔在外借期间成为常规正选，整季上阵达39场及射入1球，协助莱斯特城赢取英甲联赛冠军及重返英冠。2009/10年上半季仅于两场英格兰联赛杯及一场无关痛痒的欧联分组赛上阵，将于季后约满的基拔获外借到英冠榜末球会彼德堡直到球季结束，期间上阵10场。2010年夏季基拔约满阿仙奴成为自由球员，仅为母会合共上阵10场，英超「升班马」黑池有意罗致，其后前往-{zh-hans:谢菲尔德联; zh-hk:锡菲联;}-参加试训，惟未有获得录用。'
>>> mode_name = 'liam168/qa-roberta-base-chinese-extractive'
>>> model = AutoModelForQuestionAnswering.from_pretrained(mode_name)
>>> tokenizer = AutoTokenizer.from_pretrained(mode_name)
>>> QA = pipeline('question-answering', model=model, tokenizer=tokenizer)
>>> QA_input = {'question': "卡利亚·基拔的职业是什么?",'context': context}
>>> QA(QA_input)
{'score': 0.9999, 'start': 20, 'end': 31, 'answer': '一名英格兰籍职业足球员'}
```
## Contact
liam168520@gmail.com
| {"language": "zh", "widget": [{"text": "\u8457\u540d\u8bd7\u6b4c\u300a\u5047\u5982\u751f\u6d3b\u6b3a\u9a97\u4e86\u4f60\u300b\u7684\u4f5c\u8005\u662f", "context": "\u666e\u5e0c\u91d1\u4ece\u90a3\u91cc\u5b66\u4e60\u4eba\u6c11\u7684\u8bed\u8a00\uff0c\u5438\u53d6\u4e86\u8bb8\u591a\u6709\u76ca\u7684\u517b\u6599\uff0c\u8fd9\u4e00\u5207\u5bf9\u666e\u5e0c\u91d1\u540e\u6765\u7684\u521b\u4f5c\u4ea7\u751f\u4e86\u5f88\u5927\u7684\u5f71\u54cd\u3002\u8fd9\u4e24\u5e74\u91cc\uff0c\u666e\u5e0c\u91d1\u521b\u4f5c\u4e86\u4e0d\u5c11\u4f18\u79c0\u7684\u4f5c\u54c1\uff0c\u5982\u300a\u56da\u5f92\u300b\u3001\u300a\u81f4\u5927\u6d77\u300b\u3001\u300a\u81f4\u51ef\u6069\u300b\u548c\u300a\u5047\u5982\u751f\u6d3b\u6b3a\u9a97\u4e86\u4f60\u300b\u7b49\u51e0\u5341\u9996\u6292\u60c5\u8bd7\uff0c\u53d9\u4e8b\u8bd7\u300a\u52aa\u6797\u4f2f\u7235\u300b\uff0c\u5386\u53f2\u5267\u300a\u9c8d\u91cc\u65af\u00b7\u6208\u90fd\u8bfa\u592b\u300b\uff0c\u4ee5\u53ca\u300a\u53f6\u752b\u76d6\u5c3c\u00b7\u5965\u6d85\u91d1\u300b\u524d\u516d\u7ae0\u3002"}]} | liam168/qa-roberta-base-chinese-extractive | null | [
"transformers",
"pytorch",
"bert",
"question-answering",
"zh",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"zh"
] | TAGS
#transformers #pytorch #bert #question-answering #zh #endpoints_compatible #region-us
| # Chinese RoBERTa-Base Model for QA
## Model description
用中文预料微调的QA模型.
## Overview
- Language model: RoBERTa-Base
- Model size: 400M
- Language: Chinese
## How to use
You can use the model directly with a pipeline for extractive question answering:
## Contact
liam168520@URL
| [
"# Chinese RoBERTa-Base Model for QA",
"## Model description\n\n用中文预料微调的QA模型.",
"## Overview\n\n- Language model: RoBERTa-Base\n- Model size: 400M\n- Language: Chinese",
"## How to use\n\nYou can use the model directly with a pipeline for extractive question answering:",
"## Contact\n\nliam168520@URL"
] | [
"TAGS\n#transformers #pytorch #bert #question-answering #zh #endpoints_compatible #region-us \n",
"# Chinese RoBERTa-Base Model for QA",
"## Model description\n\n用中文预料微调的QA模型.",
"## Overview\n\n- Language model: RoBERTa-Base\n- Model size: 400M\n- Language: Chinese",
"## How to use\n\nYou can use the model ... |
translation | transformers |
# liam168/trans-opus-mt-en-zh
## Model description
* source group: English
* target group: Chinese
* model: transformer
* source language(s): eng
* target language(s): cjy_Hans cjy_Hant cmn cmn_Hans cmn_Hant gan lzh lzh_Hans nan wuu yue yue_Hans yue_Hant
## How to use
```python
>>> from transformers import AutoModelWithLMHead,AutoTokenizer,pipeline
>>> mode_name = 'liam168/trans-opus-mt-en-zh'
>>> model = AutoModelWithLMHead.from_pretrained(mode_name)
>>> tokenizer = AutoTokenizer.from_pretrained(mode_name)
>>> translation = pipeline("translation_en_to_zh", model=model, tokenizer=tokenizer)
>>> translation('I like to study Data Science and Machine Learning.', max_length=400)
[{'translation_text': '我喜欢学习数据科学和机器学习'}]
```
## Contact
liam168520@gmail.com
| {"language": ["en", "zh"], "tags": ["translation"], "widget": [{"text": "I like to study Data Science and Machine Learning."}]} | liam168/trans-opus-mt-en-zh | null | [
"transformers",
"pytorch",
"marian",
"text2text-generation",
"translation",
"en",
"zh",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en",
"zh"
] | TAGS
#transformers #pytorch #marian #text2text-generation #translation #en #zh #autotrain_compatible #endpoints_compatible #has_space #region-us
|
# liam168/trans-opus-mt-en-zh
## Model description
* source group: English
* target group: Chinese
* model: transformer
* source language(s): eng
* target language(s): cjy_Hans cjy_Hant cmn cmn_Hans cmn_Hant gan lzh lzh_Hans nan wuu yue yue_Hans yue_Hant
## How to use
## Contact
liam168520@URL
| [
"# liam168/trans-opus-mt-en-zh",
"## Model description\n\n* source group: English\n* target group: Chinese\n* model: transformer\n* source language(s): eng\n* target language(s): cjy_Hans cjy_Hant cmn cmn_Hans cmn_Hant gan lzh lzh_Hans nan wuu yue yue_Hans yue_Hant",
"## How to use",
"## Contact\n\nliam168520... | [
"TAGS\n#transformers #pytorch #marian #text2text-generation #translation #en #zh #autotrain_compatible #endpoints_compatible #has_space #region-us \n",
"# liam168/trans-opus-mt-en-zh",
"## Model description\n\n* source group: English\n* target group: Chinese\n* model: transformer\n* source language(s): eng\n* t... |
translation | transformers |
# liam168/trans-opus-mt-zh-en
## Model description
* source group: English
* target group: Chinese
* model: transformer
* source language(s): eng
## How to use
```python
>>> from transformers import AutoModelWithLMHead,AutoTokenizer,pipeline
>>> mode_name = 'liam168/trans-opus-mt-zh-en'
>>> model = AutoModelWithLMHead.from_pretrained(mode_name)
>>> tokenizer = AutoTokenizer.from_pretrained(mode_name)
>>> translation = pipeline("translation_zh_to_en", model=model, tokenizer=tokenizer)
>>> translation('我喜欢学习数据科学和机器学习。', max_length=400)
[{'translation_text': 'I like to study data science and machine learning.'}]
```
## Contact
liam168520@gmail.com
| {"language": ["en", "zh"], "tags": ["translation"], "widget": [{"text": "\u6211\u559c\u6b22\u5b66\u4e60\u6570\u636e\u79d1\u5b66\u548c\u673a\u5668\u5b66\u4e60\u3002"}]} | liam168/trans-opus-mt-zh-en | null | [
"transformers",
"pytorch",
"marian",
"text2text-generation",
"translation",
"en",
"zh",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en",
"zh"
] | TAGS
#transformers #pytorch #marian #text2text-generation #translation #en #zh #autotrain_compatible #endpoints_compatible #has_space #region-us
|
# liam168/trans-opus-mt-zh-en
## Model description
* source group: English
* target group: Chinese
* model: transformer
* source language(s): eng
## How to use
## Contact
liam168520@URL
| [
"# liam168/trans-opus-mt-zh-en",
"## Model description\n\n* source group: English \n* target group: Chinese \n* model: transformer\n* source language(s): eng",
"## How to use",
"## Contact\n\nliam168520@URL"
] | [
"TAGS\n#transformers #pytorch #marian #text2text-generation #translation #en #zh #autotrain_compatible #endpoints_compatible #has_space #region-us \n",
"# liam168/trans-opus-mt-zh-en",
"## Model description\n\n* source group: English \n* target group: Chinese \n* model: transformer\n* source language(s): eng",
... |
null | null | ## Title Generator
References this [notebook](https://shivanandroy.com/transformers-generating-arxiv-papers-title-from-abstracts/)
Using `t5-small`, trained on a batch size of 16 for 4 epochs, utilising the ArXiV dataset through the `SimpleTransformers` library. Around 15k data was used for training and 3.7k data for evaluation.
This is a `.pkl` file.
### Prerequisites
Install `simpletransformers` library.
```bsh
pip install simpletransformers
```
### Example Usage
```py
import pickle
model = pickle.load(open("title-generator-t5-arxiv-16-4.pkl", "rb"))
# Prefix your text with 'summarize: '
text = ["summarize: " + """Venetian commodes imitated the curving lines and carved ornament of the French rocaille, but with a particular Venetian variation; the pieces were painted, often with landscapes or flowers or scenes from Guardi or other painters, or Chinoiserie, against a blue or green background, matching the colours of the Venetian school of painters whose work decorated the salons. 24] Ceiling of church of Santi Giovanni e Paolo in Venice, by Piazzetta (1727) Juno and Luna by Giovanni Battista Tiepolo (1735–45) Murano glass chandelier at the Ca Rezzonico (1758) Ballroom ceiling of the Ca Rezzonico with ceiling by Giovanni Battista Crosato (1753) In church construction, especially in the southern German-Austrian region, gigantic spatial creations are sometimes created for practical reasons alone, which, however, do not appear monumental, but are characterized by a unique fusion of architecture, painting, stucco, etc. ,."""]
print("Generated title: " + model.predict(text))
``` | {} | lianaling/title-generator-t5 | null | [
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#region-us
| ## Title Generator
References this notebook
Using 't5-small', trained on a batch size of 16 for 4 epochs, utilising the ArXiV dataset through the 'SimpleTransformers' library. Around 15k data was used for training and 3.7k data for evaluation.
This is a '.pkl' file.
### Prerequisites
Install 'simpletransformers' library.
### Example Usage
| [
"## Title Generator\nReferences this notebook\n\nUsing 't5-small', trained on a batch size of 16 for 4 epochs, utilising the ArXiV dataset through the 'SimpleTransformers' library. Around 15k data was used for training and 3.7k data for evaluation.\n\nThis is a '.pkl' file.",
"### Prerequisites\nInstall 'simpletr... | [
"TAGS\n#region-us \n",
"## Title Generator\nReferences this notebook\n\nUsing 't5-small', trained on a batch size of 16 for 4 epochs, utilising the ArXiV dataset through the 'SimpleTransformers' library. Around 15k data was used for training and 3.7k data for evaluation.\n\nThis is a '.pkl' file.",
"### Prerequ... |
null | null | https://arthritisrelieftx.com/123movies-watch-space-jam-a-new-legacy-2021-full-online-free-hd/
https://www.mycentraloregon.com/2021/07/16/how-to-watch-space-jam-a-new-legacy-free-streaming-space-jam-2-on-hbo-max-available-online/ | {} | liano/aura | null | [
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#region-us
| URL
URL | [] | [
"TAGS\n#region-us \n"
] |
null | null | https://ulmerderm.com/blog/space-jam/official-watch-space-jam-2-2021-online-for-free-123movies/
https://ulmerderm.com/blog/space-jam/official-watch-space-jam-2-2021-online-for-free-123movies/
https://ulmerderm.com/blog/space-jam/official-watch-space-jam-2-2021-online-for-free-123movies/
https://ulmerderm.com/blog/space-jam/official-watch-space-jam-2-2021-online-for-free-123movies/
https://ulmerderm.com/blog/space-jam/official-watch-space-jam-2-2021-online-for-free-123movies/
https://arthritisrelieftx.com/123moves-watch-here-space-jam-a-new-legacy-2021-hd-movie-ful-online-for-free
https://superbrandpublishing.com/123moves-watch-here-space-jam-a-new-legacy-2021-hd-movie-ful-online-for-free
https://www.mycentraloregon.com/2021/07/15/exclusive-watch-space-jam-a-new-legacy-free-streaming-space-jam-2-on-hbo-max-available-online/
https://www.mycentraloregon.com/2021/07/15/exclusive-watch-fast-and-furious-9-free-streaming-f9-on-hbo-max-available-online-14/
https://www.mycentraloregon.com/2021/07/15/exclusive-watch-space-jam-a-new-legacy-free-streaming-space-jam-2-on-hbo-max-available-online-2/
https://arthritisrelieftx.com/123movies-watch-space-jam-a-new-legacy-2021-full-online-free-hd/
https://www.mycentraloregon.com/2021/07/16/how-to-watch-space-jam-a-new-legacy-free-streaming-space-jam-2-on-hbo-max-available-online/
https://ulmerderm.com/blog/space-jam/official-watch-space-jam-2-2021-online-for-free-123movies/ | {} | liano/vioan | null | [
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#region-us
| URL
URL
URL
URL
URL
URL
URL
URL
URL
URL
URL
URL
URL | [] | [
"TAGS\n#region-us \n"
] |
null | null | this is a model for ecom representation | {} | liatwilight/sbert-ecom | null | [
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#region-us
| this is a model for ecom representation | [] | [
"TAGS\n#region-us \n"
] |
audio-to-audio | espnet |
## ESPnet2 ENH model
### `lichenda/wsj0_2mix_skim_noncausal`
This model was trained by LiChenda using wsj0_2mix recipe in [espnet](https://github.com/espnet/espnet/).
### Demo: How to use in ESPnet2
```bash
cd espnet
git checkout ac3c10cfe4faf82c0bb30f8b32d9e8692363e0a9
pip install -e .
cd egs2/wsj0_2mix/enh1
./run.sh --skip_data_prep false --skip_train true --download_model lichenda/wsj0_2mix_skim_noncausal
```
<!-- Generated by ./scripts/utils/show_enh_score.sh -->
# RESULTS
## Environments
- date: `Wed Feb 23 16:42:06 CST 2022`
- python version: `3.7.11 (default, Jul 27 2021, 14:32:16) [GCC 7.5.0]`
- espnet version: `espnet 0.10.7a1`
- pytorch version: `pytorch 1.8.1`
- Git hash: `ac3c10cfe4faf82c0bb30f8b32d9e8692363e0a9`
- Commit date: `Fri Feb 11 16:22:52 2022 +0800`
## ..
config: conf/tuning/train_enh_skim_tasnet_noncausal.yaml
|dataset|STOI|SAR|SDR|SIR|
|---|---|---|---|---|
|enhanced_cv_min_8k|0.96|19.17|18.70|29.56|
|enhanced_tt_min_8k|0.97|18.96|18.45|29.31|
## ENH config
<details><summary>expand</summary>
```
config: conf/tuning/train_enh_skim_tasnet_noncausal.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: chunk
output_dir: exp/enh_train_enh_skim_tasnet_noncausal_raw
ngpu: 1
seed: 0
num_workers: 4
num_att_plot: 3
dist_backend: nccl
dist_init_method: env://
dist_world_size: null
dist_rank: null
local_rank: 0
dist_master_addr: null
dist_master_port: null
dist_launcher: null
multiprocessing_distributed: false
unused_parameters: false
sharded_ddp: false
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 150
patience: 20
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
- - valid
- si_snr
- max
- - valid
- loss
- min
keep_nbest_models: 1
nbest_averaging_interval: 0
grad_clip: 5.0
grad_clip_type: 2.0
grad_noise: false
accum_grad: 1
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: null
use_matplotlib: true
use_tensorboard: true
use_wandb: false
wandb_project: null
wandb_id: null
wandb_entity: null
wandb_name: null
wandb_model_log_interval: -1
detect_anomaly: false
pretrain_path: null
init_param: []
ignore_init_mismatch: false
freeze_param: []
num_iters_per_epoch: null
batch_size: 8
valid_batch_size: null
batch_bins: 1000000
valid_batch_bins: null
train_shape_file:
- exp/enh_stats_8k/train/speech_mix_shape
- exp/enh_stats_8k/train/speech_ref1_shape
- exp/enh_stats_8k/train/speech_ref2_shape
valid_shape_file:
- exp/enh_stats_8k/valid/speech_mix_shape
- exp/enh_stats_8k/valid/speech_ref1_shape
- exp/enh_stats_8k/valid/speech_ref2_shape
batch_type: folded
valid_batch_type: null
fold_length:
- 80000
- 80000
- 80000
sort_in_batch: descending
sort_batch: descending
multiple_iterator: false
chunk_length: 16000
chunk_shift_ratio: 0.5
num_cache_chunks: 1024
train_data_path_and_name_and_type:
- - dump/raw/tr_min_8k/wav.scp
- speech_mix
- sound
- - dump/raw/tr_min_8k/spk1.scp
- speech_ref1
- sound
- - dump/raw/tr_min_8k/spk2.scp
- speech_ref2
- sound
valid_data_path_and_name_and_type:
- - dump/raw/cv_min_8k/wav.scp
- speech_mix
- sound
- - dump/raw/cv_min_8k/spk1.scp
- speech_ref1
- sound
- - dump/raw/cv_min_8k/spk2.scp
- speech_ref2
- sound
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
valid_max_cache_size: null
optim: adam
optim_conf:
lr: 0.001
eps: 1.0e-08
weight_decay: 0
scheduler: reducelronplateau
scheduler_conf:
mode: min
factor: 0.7
patience: 1
init: xavier_uniform
model_conf:
stft_consistency: false
loss_type: mask_mse
mask_type: null
criterions:
- name: si_snr
conf:
eps: 1.0e-07
wrapper: pit
wrapper_conf:
weight: 1.0
independent_perm: true
use_preprocessor: false
encoder: conv
encoder_conf:
channel: 64
kernel_size: 2
stride: 1
separator: skim
separator_conf:
causal: false
num_spk: 2
layer: 6
nonlinear: relu
unit: 128
segment_size: 250
dropout: 0.1
mem_type: hc
seg_overlap: true
decoder: conv
decoder_conf:
channel: 64
kernel_size: 2
stride: 1
required:
- output_dir
version: 0.10.7a1
distributed: false
```
</details>
### Citing ESPnet
```BibTex
@inproceedings{watanabe2018espnet,
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
title={{ESPnet}: End-to-End Speech Processing Toolkit},
year={2018},
booktitle={Proceedings of Interspeech},
pages={2207--2211},
doi={10.21437/Interspeech.2018-1456},
url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
}
@inproceedings{ESPnet-SE,
author = {Chenda Li and Jing Shi and Wangyou Zhang and Aswin Shanmugam Subramanian and Xuankai Chang and
Naoyuki Kamo and Moto Hira and Tomoki Hayashi and Christoph B{"{o}}ddeker and Zhuo Chen and Shinji Watanabe},
title = {ESPnet-SE: End-To-End Speech Enhancement and Separation Toolkit Designed for {ASR} Integration},
booktitle = {{IEEE} Spoken Language Technology Workshop, {SLT} 2021, Shenzhen, China, January 19-22, 2021},
pages = {785--792},
publisher = {{IEEE}},
year = {2021},
url = {https://doi.org/10.1109/SLT48900.2021.9383615},
doi = {10.1109/SLT48900.2021.9383615},
timestamp = {Mon, 12 Apr 2021 17:08:59 +0200},
biburl = {https://dblp.org/rec/conf/slt/Li0ZSCKHHBC021.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
```
or arXiv:
```bibtex
@misc{watanabe2018espnet,
title={ESPnet: End-to-End Speech Processing Toolkit},
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
year={2018},
eprint={1804.00015},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
Citing SkiM:
```bibtex
@article{li2022skim,
title={SkiM: Skipping Memory LSTM for Low-Latency Real-Time Continuous Speech Separation},
author={Li, Chenda and Yang, Lei and Wang, Weiqin and Qian, Yanmin},
journal={arXiv preprint arXiv:2201.10800},
year={2022}
}
```
| {"language": "en", "license": "cc-by-4.0", "tags": ["espnet", "audio", "audio-to-audio"], "datasets": ["wsj0_2mix"]} | lichenda/wsj0_2mix_skim_noncausal | null | [
"espnet",
"audio",
"audio-to-audio",
"en",
"dataset:wsj0_2mix",
"arxiv:1804.00015",
"license:cc-by-4.0",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"1804.00015"
] | [
"en"
] | TAGS
#espnet #audio #audio-to-audio #en #dataset-wsj0_2mix #arxiv-1804.00015 #license-cc-by-4.0 #region-us
| ESPnet2 ENH model
-----------------
### 'lichenda/wsj0\_2mix\_skim\_noncausal'
This model was trained by LiChenda using wsj0\_2mix recipe in espnet.
### Demo: How to use in ESPnet2
RESULTS
=======
Environments
------------
* date: 'Wed Feb 23 16:42:06 CST 2022'
* python version: '3.7.11 (default, Jul 27 2021, 14:32:16) [GCC 7.5.0]'
* espnet version: 'espnet 0.10.7a1'
* pytorch version: 'pytorch 1.8.1'
* Git hash: 'ac3c10cfe4faf82c0bb30f8b32d9e8692363e0a9'
+ Commit date: 'Fri Feb 11 16:22:52 2022 +0800'
..
--
config: conf/tuning/train\_enh\_skim\_tasnet\_noncausal.yaml
ENH config
----------
expand
### Citing ESPnet
or arXiv:
Citing SkiM:
| [
"### 'lichenda/wsj0\\_2mix\\_skim\\_noncausal'\n\n\nThis model was trained by LiChenda using wsj0\\_2mix recipe in espnet.",
"### Demo: How to use in ESPnet2\n\n\nRESULTS\n=======\n\n\nEnvironments\n------------\n\n\n* date: 'Wed Feb 23 16:42:06 CST 2022'\n* python version: '3.7.11 (default, Jul 27 2021, 14:32:16... | [
"TAGS\n#espnet #audio #audio-to-audio #en #dataset-wsj0_2mix #arxiv-1804.00015 #license-cc-by-4.0 #region-us \n",
"### 'lichenda/wsj0\\_2mix\\_skim\\_noncausal'\n\n\nThis model was trained by LiChenda using wsj0\\_2mix recipe in espnet.",
"### Demo: How to use in ESPnet2\n\n\nRESULTS\n=======\n\n\nEnvironments\... |
text-classification | transformers |
# Model Trained Using AutoNLP
- Problem type: Binary Classification
- Model ID: 32957902
- CO2 Emissions (in grams): 0.9756221672668951
## Validation Metrics
- Loss: 0.2765039801597595
- Accuracy: 0.8939828080229226
- Precision: 0.7757009345794392
- Recall: 0.8645833333333334
- AUC: 0.9552659749670619
- F1: 0.8177339901477833
## Usage
You can use cURL to access this model:
```
$ curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "I love AutoNLP"}' https://api-inference.huggingface.co/models/lidiia/autonlp-trans_class_arg-32957902
```
Or Python API:
```
from transformers import AutoModelForSequenceClassification, AutoTokenizer
model = AutoModelForSequenceClassification.from_pretrained("lidiia/autonlp-trans_class_arg-32957902", use_auth_token=True)
tokenizer = AutoTokenizer.from_pretrained("lidiia/autonlp-trans_class_arg-32957902", use_auth_token=True)
inputs = tokenizer("I love AutoNLP", return_tensors="pt")
outputs = model(**inputs)
``` | {"language": "unk", "tags": "autonlp", "datasets": ["lidiia/autonlp-data-trans_class_arg"], "widget": [{"text": "I love AutoNLP \ud83e\udd17"}], "co2_eq_emissions": 0.9756221672668951} | lidiia/autonlp-trans_class_arg-32957902 | null | [
"transformers",
"pytorch",
"bert",
"text-classification",
"autonlp",
"unk",
"dataset:lidiia/autonlp-data-trans_class_arg",
"co2_eq_emissions",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"unk"
] | TAGS
#transformers #pytorch #bert #text-classification #autonlp #unk #dataset-lidiia/autonlp-data-trans_class_arg #co2_eq_emissions #autotrain_compatible #endpoints_compatible #region-us
|
# Model Trained Using AutoNLP
- Problem type: Binary Classification
- Model ID: 32957902
- CO2 Emissions (in grams): 0.9756221672668951
## Validation Metrics
- Loss: 0.2765039801597595
- Accuracy: 0.8939828080229226
- Precision: 0.7757009345794392
- Recall: 0.8645833333333334
- AUC: 0.9552659749670619
- F1: 0.8177339901477833
## Usage
You can use cURL to access this model:
Or Python API:
| [
"# Model Trained Using AutoNLP\n\n- Problem type: Binary Classification\n- Model ID: 32957902\n- CO2 Emissions (in grams): 0.9756221672668951",
"## Validation Metrics\n\n- Loss: 0.2765039801597595\n- Accuracy: 0.8939828080229226\n- Precision: 0.7757009345794392\n- Recall: 0.8645833333333334\n- AUC: 0.955265974967... | [
"TAGS\n#transformers #pytorch #bert #text-classification #autonlp #unk #dataset-lidiia/autonlp-data-trans_class_arg #co2_eq_emissions #autotrain_compatible #endpoints_compatible #region-us \n",
"# Model Trained Using AutoNLP\n\n- Problem type: Binary Classification\n- Model ID: 32957902\n- CO2 Emissions (in grams... |
summarization | transformers | ## `bart-base-samsum`
This model was obtained by fine-tuning `facebook/bart-base` on Samsum dataset.
## Usage
```python
from transformers import pipeline
summarizer = pipeline("summarization", model="lidiya/bart-base-samsum")
conversation = '''Jeff: Can I train a 🤗 Transformers model on Amazon SageMaker?
Philipp: Sure you can use the new Hugging Face Deep Learning Container.
Jeff: ok.
Jeff: and how can I get started?
Jeff: where can I find documentation?
Philipp: ok, ok you can find everything here. https://huggingface.co/blog/the-partnership-amazon-sagemaker-and-hugging-face
'''
summarizer(conversation)
```
## Training procedure
- Colab notebook: https://colab.research.google.com/drive/1RInRjLLso9E2HG_xjA6j8JO3zXzSCBRF?usp=sharing
## Results
| key | value |
| --- | ----- |
| eval_rouge1 | 46.6619 |
| eval_rouge2 | 23.3285 |
| eval_rougeL | 39.4811 |
| eval_rougeLsum | 43.0482 |
| test_rouge1 | 44.9932 |
| test_rouge2 | 21.7286 |
| test_rougeL | 38.1921 |
| test_rougeLsum | 41.2672 |
| {"language": "en", "license": "apache-2.0", "tags": ["bart", "seq2seq", "summarization"], "datasets": ["samsum"], "widget": [{"text": "Jeff: Can I train a \ud83e\udd17 Transformers model on Amazon SageMaker? \nPhilipp: Sure you can use the new Hugging Face Deep Learning Container. \nJeff: ok.\nJeff: and how can I get started? \nJeff: where can I find documentation? \nPhilipp: ok, ok you can find everything here. https://huggingface.co/blog/the-partnership-amazon-sagemaker-and-hugging-face\n"}], "model-index": [{"name": "bart-base-samsum", "results": [{"task": {"type": "abstractive-text-summarization", "name": "Abstractive Text Summarization"}, "dataset": {"name": "SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization", "type": "samsum"}, "metrics": [{"type": "rouge-1", "value": 46.6619, "name": "Validation ROUGE-1"}, {"type": "rouge-2", "value": 23.3285, "name": "Validation ROUGE-2"}, {"type": "rouge-l", "value": 39.4811, "name": "Validation ROUGE-L"}, {"type": "rouge-1", "value": 44.9932, "name": "Test ROUGE-1"}, {"type": "rouge-2", "value": 21.7286, "name": "Test ROUGE-2"}, {"type": "rouge-l", "value": 38.1921, "name": "Test ROUGE-L"}]}, {"task": {"type": "summarization", "name": "Summarization"}, "dataset": {"name": "samsum", "type": "samsum", "config": "samsum", "split": "test"}, "metrics": [{"type": "rouge", "value": 45.0148, "name": "ROUGE-1", "verified": true, "verifyToken": "eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiZWNlYWIyNzI4MDg5YTcxNzE2NDg3MTBkZGMzMGFmNjVhNDhiMjdiM2YxODdiMDRhZWYyYTdlY2ZkOTZlMThkNyIsInZlcnNpb24iOjF9.hUpQMm2qHUkBPstp7nldJFNy-9B75Z6zunEQCstfGSxIUYXdIlI9u-o0Y9DHIBr4ZLx_CvBtvR2e0shcFFbUBg"}, {"type": "rouge", "value": 21.6861, "name": "ROUGE-2", "verified": true, "verifyToken": "eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiOTAwNjdmM2MwMTcxYjNjMTA4ODk4ZDRhODQ1M2UwN2U2ZjM0MDAyZTJhMTRmMTg0ZThiYThiYTJiN2FiYTk1ZiIsInZlcnNpb24iOjF9._QzKtHvIc_oi1VO-Maxofu-LKINnu9NuAwHmLKka_KwEwrTUZkL74zLa-r4ojKNWpRLRicu02L8W_AQafYoZCw"}, {"type": "rouge", "value": 38.1728, "name": "ROUGE-L", "verified": true, "verifyToken": "eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiNGU0OTEzZTFhMGExOTkzYTI3NzljYjg2YzAxNDM4YzBhM2NjNjI4NWMxYjUwYmFjYzc5YTcxMGVmMTI3YThmMiIsInZlcnNpb24iOjF9.2JgzUAzdOOxUlt8HOWYa8mQuqyRBdyn-LqPiZI-h72zT8mrEO3sIEmmBOvmW40Gf5rvlErYtq87BgxzNwwYUAA"}, {"type": "rouge", "value": 41.2794, "name": "ROUGE-LSUM", "verified": true, "verifyToken": "eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiNjI3ODg4YWQ5MjgwZmZkYTMzMGRjMGI2OWU2MDQ0ZDI3MThkZmYzN2U0OGMwMWJlMjhlMTc5YzgwMDBiM2JiZSIsInZlcnNpb24iOjF9.EnYKG7MuM-lNLkKOrlsb6mB94HqOg9sDBG1mCOni8hi7kM0rveSgSDVLk5Z6Adp-cfdRlho8zK-15TJTHJRxAw"}, {"type": "loss", "value": 1.597476601600647, "name": "loss", "verified": true, "verifyToken": "eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiZTBmYjJmZDhiYmJiMTcxODM5M2ZmMTBkZTcwYzM2NDFiMDJjNjJhOGMyNGQ3MGI1Y2UxZTBhNTBiMjFjZGZiNyIsInZlcnNpb24iOjF9.UdOhxHcBJGRM-kz46st_vVQR_-KWr9EtsaQnLvj7YjCzE6JqHA2LPXnDogpUQX96PISJj32XoK7jlj-2z-CGBQ"}, {"type": "gen_len", "value": 17.6606, "name": "gen_len", "verified": true, "verifyToken": "eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9.eyJoYXNoIjoiMWNlM2IyY2EzZGNiOWE0ZGMxZmJmZjhmMDI2YzE1YTQ3NmM3OGQ1NjY2ODllYjI5MDllODNhMjNmMWMyMDAyMiIsInZlcnNpb24iOjF9.sewPQx2WKY8IOBgr0XZkmzOzgwsvJko2iK0noBHpgbyWp41akxWHiaxmvipTOLcx7rbIroXQEr_UgE_LMv46Dw"}]}]}]} | lidiya/bart-base-samsum | null | [
"transformers",
"pytorch",
"safetensors",
"bart",
"text2text-generation",
"seq2seq",
"summarization",
"en",
"dataset:samsum",
"license:apache-2.0",
"model-index",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en"
] | TAGS
#transformers #pytorch #safetensors #bart #text2text-generation #seq2seq #summarization #en #dataset-samsum #license-apache-2.0 #model-index #autotrain_compatible #endpoints_compatible #has_space #region-us
| 'bart-base-samsum'
------------------
This model was obtained by fine-tuning 'facebook/bart-base' on Samsum dataset.
Usage
-----
Training procedure
------------------
* Colab notebook: URL
Results
-------
| [] | [
"TAGS\n#transformers #pytorch #safetensors #bart #text2text-generation #seq2seq #summarization #en #dataset-samsum #license-apache-2.0 #model-index #autotrain_compatible #endpoints_compatible #has_space #region-us \n"
] |
summarization | transformers | ## `bart-large-xsum-samsum`
This model was obtained by fine-tuning `facebook/bart-large-xsum` on [Samsum](https://huggingface.co/datasets/samsum) dataset.
## Usage
```python
from transformers import pipeline
summarizer = pipeline("summarization", model="lidiya/bart-large-xsum-samsum")
conversation = '''Hannah: Hey, do you have Betty's number?
Amanda: Lemme check
Amanda: Sorry, can't find it.
Amanda: Ask Larry
Amanda: He called her last time we were at the park together
Hannah: I don't know him well
Amanda: Don't be shy, he's very nice
Hannah: If you say so..
Hannah: I'd rather you texted him
Amanda: Just text him 🙂
Hannah: Urgh.. Alright
Hannah: Bye
Amanda: Bye bye
'''
summarizer(conversation)
```
## Training procedure
- Colab notebook: https://colab.research.google.com/drive/1dul0Sg-TTMy9xZCJzmDRajXbyzDwtYx6?usp=sharing
## Results
| key | value |
| --- | ----- |
| eval_rouge1 | 54.3921 |
| eval_rouge2 | 29.8078 |
| eval_rougeL | 45.1543 |
| eval_rougeLsum | 49.942 |
| test_rouge1 | 53.3059 |
| test_rouge2 | 28.355 |
| test_rougeL | 44.0953 |
| test_rougeLsum | 48.9246 | | {"language": "en", "license": "apache-2.0", "tags": ["bart", "seq2seq", "summarization"], "datasets": ["samsum"], "widget": [{"text": "Hannah: Hey, do you have Betty's number?\nAmanda: Lemme check\nAmanda: Sorry, can't find it.\nAmanda: Ask Larry\nAmanda: He called her last time we were at the park together\nHannah: I don't know him well\nAmanda: Don't be shy, he's very nice\nHannah: If you say so..\nHannah: I'd rather you texted him\nAmanda: Just text him \ud83d\ude42\nHannah: Urgh.. Alright\nHannah: Bye\nAmanda: Bye bye\n"}], "model-index": [{"name": "bart-large-xsum-samsum", "results": [{"task": {"type": "abstractive-text-summarization", "name": "Abstractive Text Summarization"}, "dataset": {"name": "SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization", "type": "samsum"}, "metrics": [{"type": "rouge-1", "value": 54.3921, "name": "Validation ROUGE-1"}, {"type": "rouge-2", "value": 29.8078, "name": "Validation ROUGE-2"}, {"type": "rouge-l", "value": 45.1543, "name": "Validation ROUGE-L"}, {"type": "rouge-1", "value": 53.3059, "name": "Test ROUGE-1"}, {"type": "rouge-2", "value": 28.355, "name": "Test ROUGE-2"}, {"type": "rouge-l", "value": 44.0953, "name": "Test ROUGE-L"}]}]}]} | lidiya/bart-large-xsum-samsum | null | [
"transformers",
"pytorch",
"safetensors",
"bart",
"text2text-generation",
"seq2seq",
"summarization",
"en",
"dataset:samsum",
"license:apache-2.0",
"model-index",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en"
] | TAGS
#transformers #pytorch #safetensors #bart #text2text-generation #seq2seq #summarization #en #dataset-samsum #license-apache-2.0 #model-index #autotrain_compatible #endpoints_compatible #has_space #region-us
| 'bart-large-xsum-samsum'
------------------------
This model was obtained by fine-tuning 'facebook/bart-large-xsum' on Samsum dataset.
Usage
-----
Training procedure
------------------
* Colab notebook: URL
Results
-------
| [] | [
"TAGS\n#transformers #pytorch #safetensors #bart #text2text-generation #seq2seq #summarization #en #dataset-samsum #license-apache-2.0 #model-index #autotrain_compatible #endpoints_compatible #has_space #region-us \n"
] |
text-generation | transformers |
#Teia Moranta | {"tags": ["conversational"]} | life4free96/DialogGPT-med-TeiaMoranta | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"conversational",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
#Teia Moranta | [] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n"
] |
text-generation | transformers |
#rick sanchez | {"tags": ["conversational"]} | light/small-rickk | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"conversational",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
#rick sanchez | [] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n"
] |
text-classification | transformers |
# BiomedNLP-PubMedBERT finetuned on textual entailment (NLI)
The [microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext](https://huggingface.co/microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext?text=%5BMASK%5D+is+a+tumor+suppressor+gene) finetuned on the MNLI dataset. It should be useful in textual entailment tasks involving biomedical corpora.
## Usage
Given two sentences (a premise and a hypothesis), the model outputs the logits of entailment, neutral or contradiction.
You can test the model using the HuggingFace model widget on the side:
- Input two sentences (premise and hypothesis) one after the other.
- The model returns the probabilities of 3 labels: entailment(LABEL:0), neutral(LABEL:1) and contradiction(LABEL:2) respectively.
To use the model locally on your machine:
```python
# import torch
# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("lighteternal/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext-finetuned-mnli")
model = AutoModelForSequenceClassification.from_pretrained("lighteternal/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext-finetuned-mnli")
premise = 'EpCAM is overexpressed in breast cancer'
hypothesis = 'EpCAM is downregulated in breast cancer.'
# run through model pre-trained on MNLI
x = tokenizer.encode(premise, hypothesis, return_tensors='pt',
truncation_strategy='only_first')
logits = model(x)[0]
probs = logits.softmax(dim=1)
print('Probabilities for entailment, neutral, contradiction \n', np.around(probs.cpu().
detach().numpy(),3))
# Probabilities for entailment, neutral, contradiction
# 0.001 0.001 0.998
```
## Metrics
Evaluation on classification accuracy (entailment, contradiction, neutral) on MNLI test set:
| Metric | Value |
| --- | --- |
| Accuracy | 0.8338|
See Training Metrics tab for detailed info. | {"language": "en", "license": "mit", "tags": ["textual-entailment", "nli", "pytorch"], "datasets": ["mnli"], "widget": [{"text": "EpCAM is overexpressed in breast cancer. </s></s> EpCAM is downregulated in breast cancer."}]} | lighteternal/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext-finetuned-mnli | null | [
"transformers",
"pytorch",
"tensorboard",
"bert",
"text-classification",
"textual-entailment",
"nli",
"en",
"dataset:mnli",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en"
] | TAGS
#transformers #pytorch #tensorboard #bert #text-classification #textual-entailment #nli #en #dataset-mnli #license-mit #autotrain_compatible #endpoints_compatible #region-us
| BiomedNLP-PubMedBERT finetuned on textual entailment (NLI)
==========================================================
The microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext finetuned on the MNLI dataset. It should be useful in textual entailment tasks involving biomedical corpora.
Usage
-----
Given two sentences (a premise and a hypothesis), the model outputs the logits of entailment, neutral or contradiction.
You can test the model using the HuggingFace model widget on the side:
* Input two sentences (premise and hypothesis) one after the other.
* The model returns the probabilities of 3 labels: entailment(LABEL:0), neutral(LABEL:1) and contradiction(LABEL:2) respectively.
To use the model locally on your machine:
Metrics
-------
Evaluation on classification accuracy (entailment, contradiction, neutral) on MNLI test set:
See Training Metrics tab for detailed info.
| [] | [
"TAGS\n#transformers #pytorch #tensorboard #bert #text-classification #textual-entailment #nli #en #dataset-mnli #license-mit #autotrain_compatible #endpoints_compatible #region-us \n"
] |
translation | transformers |
## Greek to English NMT
## By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* source languages: el
* target languages: en
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + BPE segmentation
* metrics: bleu, chrf
### Model description
Trained using the Fairseq framework, transformer_iwslt_de_en architecture.\\
BPE segmentation (20k codes).\\
Mixed-case model.
### How to use
```
from transformers import FSMTTokenizer, FSMTForConditionalGeneration
mname = "lighteternal/SSE-TUC-mt-el-en-cased"
tokenizer = FSMTTokenizer.from_pretrained(mname)
model = FSMTForConditionalGeneration.from_pretrained(mname)
text = "Ο όρος τεχνητή νοημοσύνη αναφέρεται στον κλάδο της πληροφορικής ο οποίος ασχολείται με τη σχεδίαση και την υλοποίηση υπολογιστικών συστημάτων που μιμούνται στοιχεία της ανθρώπινης συμπεριφοράς ."
encoded = tokenizer.encode(text, return_tensors='pt')
outputs = model.generate(encoded, num_beams=5, num_return_sequences=5, early_stopping=True)
for i, output in enumerate(outputs):
i += 1
print(f"{i}: {output.tolist()}")
decoded = tokenizer.decode(output, skip_special_tokens=True)
print(f"{i}: {decoded}")
```
## Training data
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
## Eval results
Results on Tatoeba testset (EL-EN):
| BLEU | chrF |
| ------ | ------ |
| 79.3 | 0.795 |
Results on XNLI parallel (EL-EN):
| BLEU | chrF |
| ------ | ------ |
| 66.2 | 0.623 |
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| {"language": ["en", "el"], "license": "apache-2.0", "tags": ["translation"], "metrics": ["bleu"], "widget": [{"text": "\u039f \u03cc\u03c1\u03bf\u03c2 \u03c4\u03b5\u03c7\u03bd\u03b7\u03c4\u03ae \u03bd\u03bf\u03b7\u03bc\u03bf\u03c3\u03cd\u03bd\u03b7 \u03b1\u03bd\u03b1\u03c6\u03ad\u03c1\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c4\u03bf\u03bd \u03ba\u03bb\u03ac\u03b4\u03bf \u03c4\u03b7\u03c2 \u03c0\u03bb\u03b7\u03c1\u03bf\u03c6\u03bf\u03c1\u03b9\u03ba\u03ae\u03c2 \u03bf \u03bf\u03c0\u03bf\u03af\u03bf\u03c2 \u03b1\u03c3\u03c7\u03bf\u03bb\u03b5\u03af\u03c4\u03b1\u03b9 \u03bc\u03b5 \u03c4\u03b7 \u03c3\u03c7\u03b5\u03b4\u03af\u03b1\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd \u03c5\u03bb\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03c4\u03b9\u03ba\u03ce\u03bd \u03c3\u03c5\u03c3\u03c4\u03b7\u03bc\u03ac\u03c4\u03c9\u03bd \u03c0\u03bf\u03c5 \u03bc\u03b9\u03bc\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03c3\u03c4\u03bf\u03b9\u03c7\u03b5\u03af\u03b1 \u03c4\u03b7\u03c2 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7\u03c2 \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03b9\u03c6\u03bf\u03c1\u03ac\u03c2. "}]} | lighteternal/SSE-TUC-mt-el-en-cased | null | [
"transformers",
"pytorch",
"fsmt",
"text2text-generation",
"translation",
"en",
"el",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en",
"el"
] | TAGS
#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
| Greek to English NMT
--------------------
By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
------------------------------------------------------------------------------
* source languages: el
* target languages: en
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + BPE segmentation
* metrics: bleu, chrf
### Model description
Trained using the Fairseq framework, transformer\_iwslt\_de\_en architecture.\
BPE segmentation (20k codes).\
Mixed-case model.
### How to use
Training data
-------------
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
Eval results
------------
Results on Tatoeba testset (EL-EN):
Results on XNLI parallel (EL-EN):
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| [
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (20k codes).\\\nMixed-case model.",
"### How to use\n\n\nTraining data\n-------------\n\n\nConsolidated corpus from Opus and CC-Matrix (~6.6GB in total)\n\n\nEval results\n------------\... | [
"TAGS\n#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (20k codes).\\\nMixed-case... |
translation | transformers |
## Greek to English NMT (lower-case output)
## By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* source languages: el
* target languages: en
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + BPE segmentation
* metrics: bleu, chrf
* output: lowercase only, for mixed-cased model use this: https://huggingface.co/lighteternal/SSE-TUC-mt-el-en-cased
### Model description
Trained using the Fairseq framework, transformer_iwslt_de_en architecture.\\
BPE segmentation (10k codes).\\
Lower-case model.
### How to use
```
from transformers import FSMTTokenizer, FSMTForConditionalGeneration
mname = " <your_downloaded_model_folderpath_here> "
tokenizer = FSMTTokenizer.from_pretrained(mname)
model = FSMTForConditionalGeneration.from_pretrained(mname)
text = "Η τύχη βοηθάει τους τολμηρούς."
encoded = tokenizer.encode(text, return_tensors='pt')
outputs = model.generate(encoded, num_beams=5, num_return_sequences=5, early_stopping=True)
for i, output in enumerate(outputs):
i += 1
print(f"{i}: {output.tolist()}")
decoded = tokenizer.decode(output, skip_special_tokens=True)
print(f"{i}: {decoded}")
```
## Training data
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
## Eval results
Results on Tatoeba testset (EL-EN):
| BLEU | chrF |
| ------ | ------ |
| 79.3 | 0.795 |
Results on XNLI parallel (EL-EN):
| BLEU | chrF |
| ------ | ------ |
| 66.2 | 0.623 |
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| {"language": ["en", "el"], "license": "apache-2.0", "tags": ["translation"], "metrics": ["bleu"], "widget": [{"text": "\u0397 \u03c4\u03cd\u03c7\u03b7 \u03b2\u03bf\u03b7\u03b8\u03ac\u03b5\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03c4\u03bf\u03bb\u03bc\u03b7\u03c1\u03bf\u03cd\u03c2."}]} | lighteternal/SSE-TUC-mt-el-en-lowercase | null | [
"transformers",
"pytorch",
"fsmt",
"text2text-generation",
"translation",
"en",
"el",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en",
"el"
] | TAGS
#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
| Greek to English NMT (lower-case output)
----------------------------------------
By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
------------------------------------------------------------------------------
* source languages: el
* target languages: en
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + BPE segmentation
* metrics: bleu, chrf
* output: lowercase only, for mixed-cased model use this: URL
### Model description
Trained using the Fairseq framework, transformer\_iwslt\_de\_en architecture.\
BPE segmentation (10k codes).\
Lower-case model.
### How to use
Training data
-------------
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
Eval results
------------
Results on Tatoeba testset (EL-EN):
Results on XNLI parallel (EL-EN):
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| [
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (10k codes).\\\nLower-case model.",
"### How to use\n\n\nTraining data\n-------------\n\n\nConsolidated corpus from Opus and CC-Matrix (~6.6GB in total)\n\n\nEval results\n------------\... | [
"TAGS\n#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (10k codes).\\\nLower-case... |
translation | transformers |
## English to Greek NMT
## By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* source languages: en
* target languages: el
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + BPE segmentation
* metrics: bleu, chrf
### Model description
Trained using the Fairseq framework, transformer_iwslt_de_en architecture.\\
BPE segmentation (20k codes).\\
Mixed-case model.
### How to use
```
from transformers import FSMTTokenizer, FSMTForConditionalGeneration
mname = "lighteternal/SSE-TUC-mt-en-el-cased"
tokenizer = FSMTTokenizer.from_pretrained(mname)
model = FSMTForConditionalGeneration.from_pretrained(mname)
text = " 'Katerina', is the best name for a girl."
encoded = tokenizer.encode(text, return_tensors='pt')
outputs = model.generate(encoded, num_beams=5, num_return_sequences=5, early_stopping=True)
for i, output in enumerate(outputs):
i += 1
print(f"{i}: {output.tolist()}")
decoded = tokenizer.decode(output, skip_special_tokens=True)
print(f"{i}: {decoded}")
```
## Training data
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
## Eval results
Results on Tatoeba testset (EN-EL):
| BLEU | chrF |
| ------ | ------ |
| 76.9 | 0.733 |
Results on XNLI parallel (EN-EL):
| BLEU | chrF |
| ------ | ------ |
| 65.4 | 0.624 |
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| {"language": ["en", "el"], "license": "apache-2.0", "tags": ["translation"], "metrics": ["bleu"], "widget": [{"text": "'Katerina', is the best name for a girl."}]} | lighteternal/SSE-TUC-mt-en-el-cased | null | [
"transformers",
"pytorch",
"fsmt",
"text2text-generation",
"translation",
"en",
"el",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en",
"el"
] | TAGS
#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
| English to Greek NMT
--------------------
By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
------------------------------------------------------------------------------
* source languages: en
* target languages: el
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + BPE segmentation
* metrics: bleu, chrf
### Model description
Trained using the Fairseq framework, transformer\_iwslt\_de\_en architecture.\
BPE segmentation (20k codes).\
Mixed-case model.
### How to use
Training data
-------------
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
Eval results
------------
Results on Tatoeba testset (EN-EL):
Results on XNLI parallel (EN-EL):
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| [
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (20k codes).\\\nMixed-case model.",
"### How to use\n\n\nTraining data\n-------------\n\n\nConsolidated corpus from Opus and CC-Matrix (~6.6GB in total)\n\n\nEval results\n------------\... | [
"TAGS\n#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (20k codes).\\\nMixed-case... |
translation | transformers |
## English to Greek NMT (lower-case output)
## By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* source languages: en
* target languages: el
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + lower-casing + BPE segmentation
* metrics: bleu, chrf
* output: lowercase only, for mixed-cased model use this: https://huggingface.co/lighteternal/SSE-TUC-mt-en-el-cased
### Model description
Trained using the Fairseq framework, transformer_iwslt_de_en architecture.\\
BPE segmentation (10k codes).\\
Lower-case model.
### How to use
```
from transformers import FSMTTokenizer, FSMTForConditionalGeneration
mname = " <your_downloaded_model_folderpath_here> "
tokenizer = FSMTTokenizer.from_pretrained(mname)
model = FSMTForConditionalGeneration.from_pretrained(mname)
text = "Not all those who wander are lost."
encoded = tokenizer.encode(text, return_tensors='pt')
outputs = model.generate(encoded, num_beams=5, num_return_sequences=5, early_stopping=True)
for i, output in enumerate(outputs):
i += 1
print(f"{i}: {output.tolist()}")
decoded = tokenizer.decode(output, skip_special_tokens=True)
print(f"{i}: {decoded}")
```
## Training data
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
## Eval results
Results on Tatoeba testset (EN-EL):
| BLEU | chrF |
| ------ | ------ |
| 77.3 | 0.739 |
Results on XNLI parallel (EN-EL):
| BLEU | chrF |
| ------ | ------ |
| 66.1 | 0.606 |
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| {"language": ["en", "el"], "license": "apache-2.0", "tags": ["translation"], "metrics": ["bleu"], "widget": [{"text": "Not all those who wander are lost."}]} | lighteternal/SSE-TUC-mt-en-el-lowercase | null | [
"transformers",
"pytorch",
"fsmt",
"text2text-generation",
"translation",
"en",
"el",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en",
"el"
] | TAGS
#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
| English to Greek NMT (lower-case output)
----------------------------------------
By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
------------------------------------------------------------------------------
* source languages: en
* target languages: el
* licence: apache-2.0
* dataset: Opus, CCmatrix
* model: transformer(fairseq)
* pre-processing: tokenization + lower-casing + BPE segmentation
* metrics: bleu, chrf
* output: lowercase only, for mixed-cased model use this: URL
### Model description
Trained using the Fairseq framework, transformer\_iwslt\_de\_en architecture.\
BPE segmentation (10k codes).\
Lower-case model.
### How to use
Training data
-------------
Consolidated corpus from Opus and CC-Matrix (~6.6GB in total)
Eval results
------------
Results on Tatoeba testset (EN-EL):
Results on XNLI parallel (EN-EL):
### BibTeX entry and citation info
Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| [
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (10k codes).\\\nLower-case model.",
"### How to use\n\n\nTraining data\n-------------\n\n\nConsolidated corpus from Opus and CC-Matrix (~6.6GB in total)\n\n\nEval results\n------------\... | [
"TAGS\n#transformers #pytorch #fsmt #text2text-generation #translation #en #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"### Model description\n\n\nTrained using the Fairseq framework, transformer\\_iwslt\\_de\\_en architecture.\\\nBPE segmentation (10k codes).\\\nLower-case... |
text-classification | transformers |
# Fact vs. opinion binary classifier, trained on a mixed EN-EL annotated corpus.
### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
This is an XLM-Roberta-base model with a binary classification head. Given a sentence, it can classify it either as a fact or an opinion based on its content.
You can use this model in any of the XLM-R supported languages for the same task, taking advantage of its 0-shot learning capabilities. However, the model was trained only using English and Greek sentences.
Legend of HuggingFace API labels:
* Label 0: Opinion/Subjective sentence
* Label 1: Fact/Objective sentence
## Dataset training info
The original dataset (available here: https://github.com/1024er/cbert_aug/tree/crayon/datasets/subj) contained aprox. 9000 annotated sentences (classified as subjective or objective). It was translated to Greek using Google Translate. The Greek version was then concatenated with the original English one to create the mixed EN-EL dataset.
The model was trained for 5 epochs, using batch size = 8. Detailed metrics and hyperparameters available on the "Metrics" tab.
## Evaluation Results on test set
| accuracy | precision | recall | f1 |
| ----------- | ----------- | ----------- | ----------- |
|0.952 | 0.945 | 0.960 | 0.952 |
## Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| {"language": ["en", "el", "multilingual"], "license": "apache-2.0", "tags": ["text-classification", "fact-or-opinion", "transformers"], "widget": [{"text": "\u039e\u03b5\u03c7\u03c9\u03c1\u03af\u03b6\u03b5\u03b9 \u03b7 \u03ba\u03b1\u03b8\u03b7\u03bb\u03c9\u03c4\u03b9\u03ba\u03ae \u03b5\u03c1\u03bc\u03b7\u03bd\u03b5\u03af\u03b1 \u03c4\u03bf\u03c5 \u03c0\u03c1\u03c9\u03c4\u03b1\u03b3\u03c9\u03bd\u03b9\u03c3\u03c4\u03ae."}, {"text": "\u0397 \u0395\u03bb\u03bb\u03ac\u03b4\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c7\u03ce\u03c1\u03b1 \u03c4\u03b7\u03c2 \u0395\u03c5\u03c1\u03ce\u03c0\u03b7\u03c2."}, {"text": "Tolkien was an English writer"}, {"text": "Tolkien is my favorite writer."}], "pipeline_tag": "text-classification"} | lighteternal/fact-or-opinion-xlmr-el | null | [
"transformers",
"pytorch",
"tensorboard",
"xlm-roberta",
"text-classification",
"fact-or-opinion",
"en",
"el",
"multilingual",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en",
"el",
"multilingual"
] | TAGS
#transformers #pytorch #tensorboard #xlm-roberta #text-classification #fact-or-opinion #en #el #multilingual #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
| Fact vs. opinion binary classifier, trained on a mixed EN-EL annotated corpus.
==============================================================================
### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
This is an XLM-Roberta-base model with a binary classification head. Given a sentence, it can classify it either as a fact or an opinion based on its content.
You can use this model in any of the XLM-R supported languages for the same task, taking advantage of its 0-shot learning capabilities. However, the model was trained only using English and Greek sentences.
Legend of HuggingFace API labels:
* Label 0: Opinion/Subjective sentence
* Label 1: Fact/Objective sentence
Dataset training info
---------------------
The original dataset (available here: URL contained aprox. 9000 annotated sentences (classified as subjective or objective). It was translated to Greek using Google Translate. The Greek version was then concatenated with the original English one to create the mixed EN-EL dataset.
The model was trained for 5 epochs, using batch size = 8. Detailed metrics and hyperparameters available on the "Metrics" tab.
Evaluation Results on test set
------------------------------
Acknowledgement
---------------
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
| [
"### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)\n\n\nThis is an XLM-Roberta-base model with a binary classification head. Given a sentence, it can classify it either as a fact or an opinion based on its content.\n\n\nYou can use this model in any of the XLM-R supported languages ... | [
"TAGS\n#transformers #pytorch #tensorboard #xlm-roberta #text-classification #fact-or-opinion #en #el #multilingual #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)\n\n\nThis is an XLM-Roberta-base m... |
text-generation | transformers | # Greek (el) GPT2 model - small
<img src="https://huggingface.co/lighteternal/gpt2-finetuned-greek-small/raw/main/GPT2el.png" width="600"/>
#### A new version (recommended) trained on 5x more data is available at: https://huggingface.co/lighteternal/gpt2-finetuned-greek
### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* language: el
* licence: apache-2.0
* dataset: ~5GB of Greek corpora
* model: GPT2 (12-layer, 768-hidden, 12-heads, 117M parameters. OpenAI GPT-2 English model, finetuned for the Greek language)
* pre-processing: tokenization + BPE segmentation
### Model description
A text generation (autoregressive) model, using Huggingface transformers and fastai based on the English GPT-2(small). 

Finetuned with gradual layer unfreezing. This is a more efficient and sustainable alternative compared to training from scratch, especially for low-resource languages. 

Based on the work of Thomas Dehaene (ML6) for the creation of a Dutch GPT2: https://colab.research.google.com/drive/1Y31tjMkB8TqKKFlZ5OJ9fcMp3p8suvs4?usp=sharing
### How to use
```
from transformers import pipeline
model = "lighteternal/gpt2-finetuned-greek-small"
generator = pipeline(
'text-generation',
device=0,
model=f'{model}',
tokenizer=f'{model}')
text = "Μια φορά κι έναν καιρό"
print("\\\\
".join([x.get("generated_text") for x in generator(
text,
max_length=len(text.split(" "))+15,
do_sample=True,
top_k=50,
repetition_penalty = 1.2,
add_special_tokens=False,
num_return_sequences=5,
temperature=0.95,
top_p=0.95)]))
```
## Training data
We used a small (~5GB) sample from a consolidated Greek corpus based on CC100, Wikimatrix, Tatoeba, Books, SETIMES and GlobalVoices. A bigger corpus is expected to provide better results (T0D0).
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
Based on the work of Thomas Dehaene (ML6): https://blog.ml6.eu/dutch-gpt2-autoregressive-language-modelling-on-a-budget-cff3942dd020
| {"language": ["el"], "license": "apache-2.0", "tags": ["pytorch", "causal-lm"], "widget": [{"text": "\u03a4\u03bf \u03b1\u03b3\u03b1\u03c0\u03b7\u03bc\u03ad\u03bd\u03bf \u03bc\u03bf\u03c5 \u03bc\u03ad\u03c1\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9"}]} | lighteternal/gpt2-finetuned-greek-small | null | [
"transformers",
"pytorch",
"jax",
"gpt2",
"text-generation",
"causal-lm",
"el",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"el"
] | TAGS
#transformers #pytorch #jax #gpt2 #text-generation #causal-lm #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #has_space #text-generation-inference #region-us
| # Greek (el) GPT2 model - small
<img src="URL width="600"/>
#### A new version (recommended) trained on 5x more data is available at: URL
### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* language: el
* licence: apache-2.0
* dataset: ~5GB of Greek corpora
* model: GPT2 (12-layer, 768-hidden, 12-heads, 117M parameters. OpenAI GPT-2 English model, finetuned for the Greek language)
* pre-processing: tokenization + BPE segmentation
### Model description
A text generation (autoregressive) model, using Huggingface transformers and fastai based on the English GPT-2(small). 

Finetuned with gradual layer unfreezing. This is a more efficient and sustainable alternative compared to training from scratch, especially for low-resource languages. 

Based on the work of Thomas Dehaene (ML6) for the creation of a Dutch GPT2: URL
### How to use
## Training data
We used a small (~5GB) sample from a consolidated Greek corpus based on CC100, Wikimatrix, Tatoeba, Books, SETIMES and GlobalVoices. A bigger corpus is expected to provide better results (T0D0).
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
Based on the work of Thomas Dehaene (ML6): URL
| [
"# Greek (el) GPT2 model - small\n\n\n<img src=\"URL width=\"600\"/>",
"#### A new version (recommended) trained on 5x more data is available at: URL",
"### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)\n\n* language: el\n* licence: apache-2.0\n* dataset: ~5GB of Greek corpora \... | [
"TAGS\n#transformers #pytorch #jax #gpt2 #text-generation #causal-lm #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #has_space #text-generation-inference #region-us \n",
"# Greek (el) GPT2 model - small\n\n\n<img src=\"URL width=\"600\"/>",
"#### A new version (recommended) trained on 5x mo... |
text-generation | transformers |
# Greek (el) GPT2 model
<img src="https://huggingface.co/lighteternal/gpt2-finetuned-greek-small/raw/main/GPT2el.png" width="600"/>
### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* language: el
* licence: apache-2.0
* dataset: ~23.4 GB of Greek corpora
* model: GPT2 (12-layer, 768-hidden, 12-heads, 117M parameters. OpenAI GPT-2 English model, finetuned for the Greek language)
* pre-processing: tokenization + BPE segmentation
* metrics: perplexity
### Model description
A text generation (autoregressive) model, using Huggingface transformers and fastai based on the English GPT-2.
Finetuned with gradual layer unfreezing. This is a more efficient and sustainable alternative compared to training from scratch, especially for low-resource languages.
Based on the work of Thomas Dehaene (ML6) for the creation of a Dutch GPT2: https://colab.research.google.com/drive/1Y31tjMkB8TqKKFlZ5OJ9fcMp3p8suvs4?usp=sharing
### How to use
```
from transformers import pipeline
model = "lighteternal/gpt2-finetuned-greek"
generator = pipeline(
'text-generation',
device=0,
model=f'{model}',
tokenizer=f'{model}')
text = "Μια φορά κι έναν καιρό"
print("\
".join([x.get("generated_text") for x in generator(
text,
max_length=len(text.split(" "))+15,
do_sample=True,
top_k=50,
repetition_penalty = 1.2,
add_special_tokens=False,
num_return_sequences=5,
temperature=0.95,
top_p=0.95)]))
```
## Training data
We used a 23.4GB sample from a consolidated Greek corpus from CC100, Wikimatrix, Tatoeba, Books, SETIMES and GlobalVoices containing long senquences.
This is a better version of our GPT-2 small model (https://huggingface.co/lighteternal/gpt2-finetuned-greek-small)
## Metrics
| Metric | Value |
| ----------- | ----------- |
| Train Loss | 3.67 |
| Validation Loss | 3.83 |
| Perplexity | 39.12 |
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
Based on the work of Thomas Dehaene (ML6): https://blog.ml6.eu/dutch-gpt2-autoregressive-language-modelling-on-a-budget-cff3942dd020
| {"language": ["el"], "license": "apache-2.0", "tags": ["pytorch", "causal-lm"], "widget": [{"text": "\u03a4\u03bf \u03b1\u03b3\u03b1\u03c0\u03b7\u03bc\u03ad\u03bd\u03bf \u03bc\u03bf\u03c5 \u03bc\u03ad\u03c1\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9"}]} | lighteternal/gpt2-finetuned-greek | null | [
"transformers",
"pytorch",
"jax",
"gpt2",
"text-generation",
"causal-lm",
"el",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"el"
] | TAGS
#transformers #pytorch #jax #gpt2 #text-generation #causal-lm #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
| Greek (el) GPT2 model
=====================
<img src="URL width="600"/>
### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
* language: el
* licence: apache-2.0
* dataset: ~23.4 GB of Greek corpora
* model: GPT2 (12-layer, 768-hidden, 12-heads, 117M parameters. OpenAI GPT-2 English model, finetuned for the Greek language)
* pre-processing: tokenization + BPE segmentation
* metrics: perplexity
### Model description
A text generation (autoregressive) model, using Huggingface transformers and fastai based on the English GPT-2.
Finetuned with gradual layer unfreezing. This is a more efficient and sustainable alternative compared to training from scratch, especially for low-resource languages.
Based on the work of Thomas Dehaene (ML6) for the creation of a Dutch GPT2: URL
### How to use
Training data
-------------
We used a 23.4GB sample from a consolidated Greek corpus from CC100, Wikimatrix, Tatoeba, Books, SETIMES and GlobalVoices containing long senquences.
This is a better version of our GPT-2 small model (URL
Metrics
-------
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
Based on the work of Thomas Dehaene (ML6): URL
| [
"### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)\n\n\n* language: el\n* licence: apache-2.0\n* dataset: ~23.4 GB of Greek corpora\n* model: GPT2 (12-layer, 768-hidden, 12-heads, 117M parameters. OpenAI GPT-2 English model, finetuned for the Greek language)\n* pre-processing: token... | [
"TAGS\n#transformers #pytorch #jax #gpt2 #text-generation #causal-lm #el #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"### By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)\n\n\n* language: el\n* licence: apache-2.0\n* data... |
zero-shot-classification | transformers |
# Cross-Encoder for Greek Natural Language Inference (Textual Entailment) & Zero-Shot Classification
## By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
This model was trained using [SentenceTransformers](https://sbert.net) [Cross-Encoder](https://www.sbert.net/examples/applications/cross-encoder/README.html) class.
## Training Data
The model was trained on the the combined Greek+English version of the AllNLI dataset(sum of [SNLI](https://nlp.stanford.edu/projects/snli/) and [MultiNLI](https://cims.nyu.edu/~sbowman/multinli/)). The Greek part was created using the EN2EL NMT model available [here](https://huggingface.co/lighteternal/SSE-TUC-mt-en-el-cased).
The model can be used in two ways:
* NLI/Textual Entailment: For a given sentence pair, it will output three scores corresponding to the labels: contradiction, entailment, neutral.
* Zero-shot classification through the Huggingface pipeline: Given a sentence and a set of labels/topics, it will output the likelihood of the sentence belonging to each of the topic. Under the hood, the logit for entailment between the sentence and each label is taken as the logit for the candidate label being valid.
## Performance
Evaluation on classification accuracy (entailment, contradiction, neutral) on mixed (Greek+English) AllNLI-dev set:
| Metric | Value |
| --- | --- |
| Accuracy | 0.8409 |
## To use the model for NLI/Textual Entailment
#### Usage with sentence_transformers
Pre-trained models can be used like this:
```python
from sentence_transformers import CrossEncoder
model = CrossEncoder('lighteternal/nli-xlm-r-greek')
scores = model.predict([('Δύο άνθρωποι συναντιούνται στο δρόμο', 'Ο δρόμος έχει κόσμο'),
('Ένα μαύρο αυτοκίνητο ξεκινάει στη μέση του πλήθους.', 'Ένας άντρας οδηγάει σε ένα μοναχικό δρόμο'),
('Δυο γυναίκες μιλάνε στο κινητό', 'Το τραπέζι ήταν πράσινο')])
#Convert scores to labels
label_mapping = ['contradiction', 'entailment', 'neutral']
labels = [label_mapping[score_max] for score_max in scores.argmax(axis=1)]
print(scores, labels)
# Οutputs
#[[-3.1526504 2.9981945 -0.3108107]
# [ 5.0549307 -2.757949 -1.6220676]
# [-0.5124733 -2.2671669 3.1630592]] ['entailment', 'contradiction', 'neutral']
```
#### Usage with Transformers AutoModel
You can use the model also directly with Transformers library (without SentenceTransformers library):
```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
model = AutoModelForSequenceClassification.from_pretrained('lighteternal/nli-xlm-r-greek')
tokenizer = AutoTokenizer.from_pretrained('lighteternal/nli-xlm-r-greek')
features = tokenizer(['Δύο άνθρωποι συναντιούνται στο δρόμο', 'Ο δρόμος έχει κόσμο'],
['Ένα μαύρο αυτοκίνητο ξεκινάει στη μέση του πλήθους.', 'Ένας άντρας οδηγάει σε ένα μοναχικό δρόμο.'],
padding=True, truncation=True, return_tensors="pt")
model.eval()
with torch.no_grad():
scores = model(**features).logits
label_mapping = ['contradiction', 'entailment', 'neutral']
labels = [label_mapping[score_max] for score_max in scores.argmax(dim=1)]
print(labels)
```
## To use the model for Zero-Shot Classification
This model can also be used for zero-shot-classification:
```python
from transformers import pipeline
classifier = pipeline("zero-shot-classification", model='lighteternal/nli-xlm-r-greek')
sent = "Το Facebook κυκλοφόρησε τα πρώτα «έξυπνα» γυαλιά επαυξημένης πραγματικότητας"
candidate_labels = ["πολιτική", "τεχνολογία", "αθλητισμός"]
res = classifier(sent, candidate_labels)
print(res)
#outputs:
#{'sequence': 'Το Facebook κυκλοφόρησε τα πρώτα «έξυπνα» γυαλιά επαυξημένης πραγματικότητας', 'labels': ['τεχνολογία', 'αθλητισμός', 'πολιτική'], 'scores': [0.8380699157714844, 0.09086982160806656, 0.07106029987335205]}
```
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
### Citation info
Citation for the Greek model TBA.
Based on the work [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://arxiv.org/abs/1908.10084)
Kudos to @nreimers (Nils Reimers) for his support on Github .
| {"language": ["el", "en"], "license": "apache-2.0", "tags": ["xlm-roberta-base"], "datasets": ["multi_nli", "snli", "allnli_greek"], "metrics": ["accuracy"], "pipeline_tag": "zero-shot-classification", "widget": [{"text": "\u0397 Facebook \u03ba\u03c5\u03ba\u03bb\u03bf\u03c6\u03cc\u03c1\u03b7\u03c3\u03b5 \u03c4\u03b1 \u03c0\u03c1\u03ce\u03c4\u03b1 \u00ab\u03ad\u03be\u03c5\u03c0\u03bd\u03b1\u00bb \u03b3\u03c5\u03b1\u03bb\u03b9\u03ac \u03b5\u03c0\u03b1\u03c5\u03be\u03b7\u03bc\u03ad\u03bd\u03b7\u03c2 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2.", "candidate_labels": "\u03c4\u03b5\u03c7\u03bd\u03bf\u03bb\u03bf\u03b3\u03af\u03b1, \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae, \u03b1\u03b8\u03bb\u03b7\u03c4\u03b9\u03c3\u03bc\u03cc\u03c2", "multi_class": false}]} | lighteternal/nli-xlm-r-greek | null | [
"transformers",
"pytorch",
"xlm-roberta",
"text-classification",
"xlm-roberta-base",
"zero-shot-classification",
"el",
"en",
"dataset:multi_nli",
"dataset:snli",
"dataset:allnli_greek",
"arxiv:1908.10084",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us... | null | 2022-03-02T23:29:05+00:00 | [
"1908.10084"
] | [
"el",
"en"
] | TAGS
#transformers #pytorch #xlm-roberta #text-classification #xlm-roberta-base #zero-shot-classification #el #en #dataset-multi_nli #dataset-snli #dataset-allnli_greek #arxiv-1908.10084 #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
| Cross-Encoder for Greek Natural Language Inference (Textual Entailment) & Zero-Shot Classification
==================================================================================================
By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
------------------------------------------------------------------------------
This model was trained using SentenceTransformers Cross-Encoder class.
Training Data
-------------
The model was trained on the the combined Greek+English version of the AllNLI dataset(sum of SNLI and MultiNLI). The Greek part was created using the EN2EL NMT model available here.
The model can be used in two ways:
* NLI/Textual Entailment: For a given sentence pair, it will output three scores corresponding to the labels: contradiction, entailment, neutral.
* Zero-shot classification through the Huggingface pipeline: Given a sentence and a set of labels/topics, it will output the likelihood of the sentence belonging to each of the topic. Under the hood, the logit for entailment between the sentence and each label is taken as the logit for the candidate label being valid.
Performance
-----------
Evaluation on classification accuracy (entailment, contradiction, neutral) on mixed (Greek+English) AllNLI-dev set:
To use the model for NLI/Textual Entailment
-------------------------------------------
#### Usage with sentence\_transformers
Pre-trained models can be used like this:
#### Usage with Transformers AutoModel
You can use the model also directly with Transformers library (without SentenceTransformers library):
To use the model for Zero-Shot Classification
---------------------------------------------
This model can also be used for zero-shot-classification:
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
info
Citation for the Greek model TBA.
Based on the work Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
Kudos to @nreimers (Nils Reimers) for his support on Github .
| [
"#### Usage with sentence\\_transformers\n\n\nPre-trained models can be used like this:",
"#### Usage with Transformers AutoModel\n\n\nYou can use the model also directly with Transformers library (without SentenceTransformers library):\n\n\nTo use the model for Zero-Shot Classification\n-------------------------... | [
"TAGS\n#transformers #pytorch #xlm-roberta #text-classification #xlm-roberta-base #zero-shot-classification #el #en #dataset-multi_nli #dataset-snli #dataset-allnli_greek #arxiv-1908.10084 #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"#### Usage with sentence\\_transformers\n\n\... |
automatic-speech-recognition | transformers |
# Greek (el) version of the XLSR-Wav2Vec2 automatic speech recognition (ASR) model
### By the Hellenic Army Academy and the Technical University of Crete
* language: el
* licence: apache-2.0
* dataset: CommonVoice (EL), 364MB: https://commonvoice.mozilla.org/el/datasets + CSS10 (EL), 1.22GB: https://github.com/Kyubyong/css10
* model: XLSR-Wav2Vec2, trained for 50 epochs
* metrics: Word Error Rate (WER)
## Model description
UPDATE: We repeated the fine-tuning process using an additional 1.22GB dataset from CSS10.
Wav2Vec2 is a pretrained model for Automatic Speech Recognition (ASR) and was released in September 2020 by Alexei Baevski, Michael Auli, and Alex Conneau. Soon after the superior performance of Wav2Vec2 was demonstrated on the English ASR dataset LibriSpeech, Facebook AI presented XLSR-Wav2Vec2. XLSR stands for cross-lingual speech representations and refers to XLSR-Wav2Vec2`s ability to learn speech representations that are useful across multiple languages.
Similar to Wav2Vec2, XLSR-Wav2Vec2 learns powerful speech representations from hundreds of thousands of hours of speech in more than 50 languages of unlabeled speech. Similar, to BERT's masked language modeling, the model learns contextualized speech representations by randomly masking feature vectors before passing them to a transformer network.
This model was trained for 50 epochs on a single NVIDIA RTX 3080, for aprox. 8hrs.
## How to use for inference:
For live demo, make sure that speech files are sampled at 16kHz.
Instructions to test on CommonVoice extracts are provided in the ASR_Inference.ipynb. Snippet also available below:
```python
#!/usr/bin/env python
# coding: utf-8
# Loading dependencies and defining preprocessing functions
from transformers import Wav2Vec2ForCTC
from transformers import Wav2Vec2Processor
from datasets import load_dataset, load_metric
import re
import torchaudio
import librosa
import numpy as np
from datasets import load_dataset, load_metric
import torch
chars_to_ignore_regex = '[\\\\\\\\,\\\\\\\\?\\\\\\\\.\\\\\\\\!\\\\\\\\-\\\\\\\\;\\\\\\\\:\\\\\\\\"\\\\\\\\“\\\\\\\\%\\\\\\\\‘\\\\\\\\”\\\\\\\\�]'
def remove_special_characters(batch):
batch["text"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower() + " "
return batch
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = speech_array[0].numpy()
batch["sampling_rate"] = sampling_rate
batch["target_text"] = batch["text"]
return batch
def resample(batch):
batch["speech"] = librosa.resample(np.asarray(batch["speech"]), 48_000, 16_000)
batch["sampling_rate"] = 16_000
return batch
def prepare_dataset(batch):
# check that all files have the correct sampling rate
assert (
len(set(batch["sampling_rate"])) == 1
), f"Make sure all inputs have the same sampling rate of {processor.feature_extractor.sampling_rate}."
batch["input_values"] = processor(batch["speech"], sampling_rate=batch["sampling_rate"][0]).input_values
with processor.as_target_processor():
batch["labels"] = processor(batch["target_text"]).input_ids
return batch
# Loading model and dataset processor
model = Wav2Vec2ForCTC.from_pretrained("lighteternal/wav2vec2-large-xlsr-53-greek").to("cuda")
processor = Wav2Vec2Processor.from_pretrained("lighteternal/wav2vec2-large-xlsr-53-greek")
# Preparing speech dataset to be suitable for inference
common_voice_test = load_dataset("common_voice", "el", split="test")
common_voice_test = common_voice_test.remove_columns(["accent", "age", "client_id", "down_votes", "gender", "locale", "segment", "up_votes"])
common_voice_test = common_voice_test.map(remove_special_characters, remove_columns=["sentence"])
common_voice_test = common_voice_test.map(speech_file_to_array_fn, remove_columns=common_voice_test.column_names)
common_voice_test = common_voice_test.map(resample, num_proc=8)
common_voice_test = common_voice_test.map(prepare_dataset, remove_columns=common_voice_test.column_names, batch_size=8, num_proc=8, batched=True)
# Loading test dataset
common_voice_test_transcription = load_dataset("common_voice", "el", split="test")
#Performing inference on a random sample. Change the "example" value to try inference on different CommonVoice extracts
example = 123
input_dict = processor(common_voice_test["input_values"][example], return_tensors="pt", sampling_rate=16_000, padding=True)
logits = model(input_dict.input_values.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
print("Prediction:")
print(processor.decode(pred_ids[0]))
# πού θέλεις να πάμε ρώτησε φοβισμένα ο βασιλιάς
print("\\\\
Reference:")
print(common_voice_test_transcription["sentence"][example].lower())
# πού θέλεις να πάμε; ρώτησε φοβισμένα ο βασιλιάς.
```
## Evaluation
The model can be evaluated as follows on the Greek test data of Common Voice.
```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
test_dataset = load_dataset("common_voice", "el", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("lighteternal/wav2vec2-large-xlsr-53-greek")
model = Wav2Vec2ForCTC.from_pretrained("lighteternal/wav2vec2-large-xlsr-53-greek")
model.to("cuda")
chars_to_ignore_regex = '[\\\\\\\\,\\\\\\\\?\\\\\\\\.\\\\\\\\!\\\\\\\\-\\\\\\\\;\\\\\\\\:\\\\\\\\"\\\\\\\\“\\\\\\\\%\\\\\\\\‘\\\\\\\\”\\\\\\\\�]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
**Test Result**: 10.497628 %
### How to use for training:
Instructions and code to replicate the process are provided in the Fine_Tune_XLSR_Wav2Vec2_on_Greek_ASR_with_🤗_Transformers.ipynb notebook.
## Metrics
| Metric | Value |
| ----------- | ----------- |
| Training Loss | 0.0545 |
| Validation Loss | 0.1661 |
| CER on CommonVoice Test (%) *| 2.8753 |
| WER on CommonVoice Test (%) *| 10.4976 |
* Reference transcripts were lower-cased and striped of punctuation and special characters.
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
Based on the tutorial of Patrick von Platen: https://huggingface.co/blog/fine-tune-xlsr-wav2vec2
Original colab notebook here: https://colab.research.google.com/github/patrickvonplaten/notebooks/blob/master/Fine_Tune_XLSR_Wav2Vec2_on_Turkish_ASR_with_%F0%9F%A4%97_Transformers.ipynb#scrollTo=V7YOT2mnUiea
| {"language": "el", "license": "apache-2.0", "tags": ["audio", "hf-asr-leaderboard", "automatic-speech-recognition", "speech", "xlsr-fine-tuning-week"], "datasets": ["common_voice"], "model-index": [{"name": "XLSR Wav2Vec2 Greek by Lighteternal", "results": [{"task": {"type": "automatic-speech-recognition", "name": "Speech Recognition"}, "dataset": {"name": "CommonVoice (EL), CSS10 (EL)", "type": "CCS10 + mozilla-foundation/common_voice_7_0", "args": "el"}, "metrics": [{"type": "wer", "value": 10.497628, "name": "Test WER"}, {"type": "cer", "value": 2.87526, "name": "Test CER"}]}]}]} | lighteternal/wav2vec2-large-xlsr-53-greek | null | [
"transformers",
"pytorch",
"jax",
"wav2vec2",
"automatic-speech-recognition",
"audio",
"hf-asr-leaderboard",
"speech",
"xlsr-fine-tuning-week",
"el",
"dataset:common_voice",
"license:apache-2.0",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"el"
] | TAGS
#transformers #pytorch #jax #wav2vec2 #automatic-speech-recognition #audio #hf-asr-leaderboard #speech #xlsr-fine-tuning-week #el #dataset-common_voice #license-apache-2.0 #endpoints_compatible #has_space #region-us
| Greek (el) version of the XLSR-Wav2Vec2 automatic speech recognition (ASR) model
================================================================================
### By the Hellenic Army Academy and the Technical University of Crete
* language: el
* licence: apache-2.0
* dataset: CommonVoice (EL), 364MB: URL + CSS10 (EL), 1.22GB: URL
* model: XLSR-Wav2Vec2, trained for 50 epochs
* metrics: Word Error Rate (WER)
Model description
-----------------
UPDATE: We repeated the fine-tuning process using an additional 1.22GB dataset from CSS10.
Wav2Vec2 is a pretrained model for Automatic Speech Recognition (ASR) and was released in September 2020 by Alexei Baevski, Michael Auli, and Alex Conneau. Soon after the superior performance of Wav2Vec2 was demonstrated on the English ASR dataset LibriSpeech, Facebook AI presented XLSR-Wav2Vec2. XLSR stands for cross-lingual speech representations and refers to XLSR-Wav2Vec2's ability to learn speech representations that are useful across multiple languages.
Similar to Wav2Vec2, XLSR-Wav2Vec2 learns powerful speech representations from hundreds of thousands of hours of speech in more than 50 languages of unlabeled speech. Similar, to BERT's masked language modeling, the model learns contextualized speech representations by randomly masking feature vectors before passing them to a transformer network.
This model was trained for 50 epochs on a single NVIDIA RTX 3080, for aprox. 8hrs.
How to use for inference:
-------------------------
For live demo, make sure that speech files are sampled at 16kHz.
Instructions to test on CommonVoice extracts are provided in the ASR\_Inference.ipynb. Snippet also available below:
Evaluation
----------
The model can be evaluated as follows on the Greek test data of Common Voice.
Test Result: 10.497628 %
### How to use for training:
Instructions and code to replicate the process are provided in the Fine\_Tune\_XLSR\_Wav2Vec2\_on\_Greek\_ASR\_with\_\_Transformers.ipynb notebook.
Metrics
-------
### Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
Based on the tutorial of Patrick von Platen: URL
Original colab notebook here: URL
| [
"### By the Hellenic Army Academy and the Technical University of Crete\n\n\n* language: el\n* licence: apache-2.0\n* dataset: CommonVoice (EL), 364MB: URL + CSS10 (EL), 1.22GB: URL\n* model: XLSR-Wav2Vec2, trained for 50 epochs\n* metrics: Word Error Rate (WER)\n\n\nModel description\n-----------------\n\n\nUPDATE... | [
"TAGS\n#transformers #pytorch #jax #wav2vec2 #automatic-speech-recognition #audio #hf-asr-leaderboard #speech #xlsr-fine-tuning-week #el #dataset-common_voice #license-apache-2.0 #endpoints_compatible #has_space #region-us \n",
"### By the Hellenic Army Academy and the Technical University of Crete\n\n\n* languag... |
feature-extraction | sentence-transformers | ## Testing Sentence Transformer
This Roberta model is trained from scratch using Masked Language Modelling task on a collection of medical reports | {"tags": ["sentence-transformers"], "pipeline_tag": "feature-extraction"} | ligolab/DxRoberta | null | [
"sentence-transformers",
"pytorch",
"roberta",
"feature-extraction",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#sentence-transformers #pytorch #roberta #feature-extraction #endpoints_compatible #has_space #region-us
| ## Testing Sentence Transformer
This Roberta model is trained from scratch using Masked Language Modelling task on a collection of medical reports | [
"## Testing Sentence Transformer\nThis Roberta model is trained from scratch using Masked Language Modelling task on a collection of medical reports"
] | [
"TAGS\n#sentence-transformers #pytorch #roberta #feature-extraction #endpoints_compatible #has_space #region-us \n",
"## Testing Sentence Transformer\nThis Roberta model is trained from scratch using Masked Language Modelling task on a collection of medical reports"
] |
fill-mask | transformers |
<!-- This model card has been generated automatically according to the information Keras had access to. You should
probably proofread and complete it, then remove this comment. -->
# dummy-model
This model is a fine-tuned version of [camembert-base](https://huggingface.co/camembert-base) on an unknown dataset.
It achieves the following results on the evaluation set:
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- optimizer: None
- training_precision: float32
### Training results
### Framework versions
- Transformers 4.15.0
- TensorFlow 2.7.0
- Datasets 1.17.0
- Tokenizers 0.10.3
| {"license": "mit", "tags": ["generated_from_keras_callback"], "model-index": [{"name": "dummy-model", "results": []}]} | lijingxin/dummy-model | null | [
"transformers",
"tf",
"camembert",
"fill-mask",
"generated_from_keras_callback",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #tf #camembert #fill-mask #generated_from_keras_callback #license-mit #autotrain_compatible #endpoints_compatible #region-us
|
# dummy-model
This model is a fine-tuned version of camembert-base on an unknown dataset.
It achieves the following results on the evaluation set:
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- optimizer: None
- training_precision: float32
### Training results
### Framework versions
- Transformers 4.15.0
- TensorFlow 2.7.0
- Datasets 1.17.0
- Tokenizers 0.10.3
| [
"# dummy-model\n\nThis model is a fine-tuned version of camembert-base on an unknown dataset.\nIt achieves the following results on the evaluation set:",
"## Model description\n\nMore information needed",
"## Intended uses & limitations\n\nMore information needed",
"## Training and evaluation data\n\nMore inf... | [
"TAGS\n#transformers #tf #camembert #fill-mask #generated_from_keras_callback #license-mit #autotrain_compatible #endpoints_compatible #region-us \n",
"# dummy-model\n\nThis model is a fine-tuned version of camembert-base on an unknown dataset.\nIt achieves the following results on the evaluation set:",
"## Mod... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-large-xls-r-300m-turkish-colab
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 1.7126
- Wer: 0.8198
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 1
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 120
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:------:|:-----:|:---------------:|:------:|
| 6.7419 | 2.38 | 200 | 3.1913 | 1.0 |
| 3.0446 | 4.76 | 400 | 2.3247 | 1.0 |
| 1.3163 | 7.14 | 600 | 1.2629 | 0.9656 |
| 0.6058 | 9.52 | 800 | 1.2203 | 0.9343 |
| 0.3687 | 11.9 | 1000 | 1.2157 | 0.8849 |
| 0.2644 | 14.29 | 1200 | 1.3693 | 0.8992 |
| 0.2147 | 16.67 | 1400 | 1.3321 | 0.8623 |
| 0.1962 | 19.05 | 1600 | 1.3476 | 0.8886 |
| 0.1631 | 21.43 | 1800 | 1.3984 | 0.8755 |
| 0.15 | 23.81 | 2000 | 1.4602 | 0.8798 |
| 0.1311 | 26.19 | 2200 | 1.4727 | 0.8836 |
| 0.1174 | 28.57 | 2400 | 1.5257 | 0.8805 |
| 0.1155 | 30.95 | 2600 | 1.4697 | 0.9337 |
| 0.1046 | 33.33 | 2800 | 1.6076 | 0.8667 |
| 0.1063 | 35.71 | 3000 | 1.5012 | 0.8861 |
| 0.0996 | 38.1 | 3200 | 1.6204 | 0.8605 |
| 0.088 | 40.48 | 3400 | 1.4788 | 0.8586 |
| 0.089 | 42.86 | 3600 | 1.5983 | 0.8648 |
| 0.0805 | 45.24 | 3800 | 1.5045 | 0.8298 |
| 0.0718 | 47.62 | 4000 | 1.6361 | 0.8611 |
| 0.0718 | 50.0 | 4200 | 1.5088 | 0.8548 |
| 0.0649 | 52.38 | 4400 | 1.5491 | 0.8554 |
| 0.0685 | 54.76 | 4600 | 1.5939 | 0.8442 |
| 0.0588 | 57.14 | 4800 | 1.6321 | 0.8536 |
| 0.0591 | 59.52 | 5000 | 1.6468 | 0.8442 |
| 0.0529 | 61.9 | 5200 | 1.6086 | 0.8661 |
| 0.0482 | 64.29 | 5400 | 1.6622 | 0.8517 |
| 0.0396 | 66.67 | 5600 | 1.6191 | 0.8436 |
| 0.0463 | 69.05 | 5800 | 1.6231 | 0.8661 |
| 0.0415 | 71.43 | 6000 | 1.6874 | 0.8511 |
| 0.0383 | 73.81 | 6200 | 1.7054 | 0.8411 |
| 0.0411 | 76.19 | 6400 | 1.7073 | 0.8486 |
| 0.0346 | 78.57 | 6600 | 1.7137 | 0.8342 |
| 0.0318 | 80.95 | 6800 | 1.6523 | 0.8329 |
| 0.0299 | 83.33 | 7000 | 1.6893 | 0.8579 |
| 0.029 | 85.71 | 7200 | 1.7162 | 0.8429 |
| 0.025 | 88.1 | 7400 | 1.7589 | 0.8529 |
| 0.025 | 90.48 | 7600 | 1.7581 | 0.8398 |
| 0.0232 | 92.86 | 7800 | 1.8459 | 0.8442 |
| 0.0215 | 95.24 | 8000 | 1.7942 | 0.8448 |
| 0.0222 | 97.62 | 8200 | 1.6848 | 0.8442 |
| 0.0179 | 100.0 | 8400 | 1.7223 | 0.8298 |
| 0.0176 | 102.38 | 8600 | 1.7426 | 0.8404 |
| 0.016 | 104.76 | 8800 | 1.7501 | 0.8411 |
| 0.0153 | 107.14 | 9000 | 1.7185 | 0.8235 |
| 0.0136 | 109.52 | 9200 | 1.7250 | 0.8292 |
| 0.0117 | 111.9 | 9400 | 1.7159 | 0.8185 |
| 0.0123 | 114.29 | 9600 | 1.7135 | 0.8248 |
| 0.0121 | 116.67 | 9800 | 1.7189 | 0.8210 |
| 0.0116 | 119.05 | 10000 | 1.7126 | 0.8198 |
### Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu113
- Datasets 1.18.3
- Tokenizers 0.10.3
| {"license": "apache-2.0", "tags": ["generated_from_trainer"], "datasets": ["common_voice"], "model-index": [{"name": "wav2vec2-large-xls-r-300m-turkish-colab", "results": []}]} | lilitket/wav2vec2-large-xls-r-300m-turkish-colab | null | [
"transformers",
"pytorch",
"tensorboard",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"dataset:common_voice",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-common_voice #license-apache-2.0 #endpoints_compatible #region-us
| wav2vec2-large-xls-r-300m-turkish-colab
=======================================
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common\_voice dataset.
It achieves the following results on the evaluation set:
* Loss: 1.7126
* Wer: 0.8198
Model description
-----------------
More information needed
Intended uses & limitations
---------------------------
More information needed
Training and evaluation data
----------------------------
More information needed
Training procedure
------------------
### Training hyperparameters
The following hyperparameters were used during training:
* learning\_rate: 0.0003
* train\_batch\_size: 1
* eval\_batch\_size: 8
* seed: 42
* gradient\_accumulation\_steps: 4
* total\_train\_batch\_size: 4
* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
* lr\_scheduler\_type: linear
* lr\_scheduler\_warmup\_steps: 500
* num\_epochs: 120
* mixed\_precision\_training: Native AMP
### Training results
### Framework versions
* Transformers 4.11.3
* Pytorch 1.10.0+cu113
* Datasets 1.18.3
* Tokenizers 0.10.3
| [
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0003\n* train\\_batch\\_size: 1\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 4\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1... | [
"TAGS\n#transformers #pytorch #tensorboard #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-common_voice #license-apache-2.0 #endpoints_compatible #region-us \n",
"### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0003\n* t... |
automatic-speech-recognition | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# wav2vec2-large-xls-r-armenian-colab
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
### Training results
### Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu113
- Datasets 1.18.3
- Tokenizers 0.10.3
| {"license": "apache-2.0", "tags": ["generated_from_trainer"], "datasets": ["common_voice"], "model-index": [{"name": "wav2vec2-large-xls-r-armenian-colab", "results": []}]} | lilitket/wav2vec2-large-xls-r-armenian-colab | null | [
"transformers",
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"dataset:common_voice",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-common_voice #license-apache-2.0 #endpoints_compatible #region-us
|
# wav2vec2-large-xls-r-armenian-colab
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset.
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
### Training results
### Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu113
- Datasets 1.18.3
- Tokenizers 0.10.3
| [
"# wav2vec2-large-xls-r-armenian-colab\n\nThis model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset.",
"## Model description\n\nMore information needed",
"## Intended uses & limitations\n\nMore information needed",
"## Training and evaluation data\n\nMore information neede... | [
"TAGS\n#transformers #pytorch #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-common_voice #license-apache-2.0 #endpoints_compatible #region-us \n",
"# wav2vec2-large-xls-r-armenian-colab\n\nThis model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset.",
... |
text-generation | transformers |
#C3PO DialoGPT Model | {"tags": ["conversational"]} | limivan/DialoGPT-small-c3po | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"conversational",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
#C3PO DialoGPT Model | [] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n"
] |
fill-mask | transformers |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# bert-hateful-memes-expanded
This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on texts from the following datasets:
- [Hateful Memes](https://hatefulmemeschallenge.com/), `train`, `dev_seen` and `dev_unseen`
- [HarMeme](https://github.com/di-dimitrov/harmeme), `train`, `val` and `test`
- [MultiOFF](https://github.com/bharathichezhiyan/Multimodal-Meme-Classification-Identifying-Offensive-Content-in-Image-and-Text), `Training`, `Validation` and `Testing`
It achieves the following results on the evaluation set:
- Loss: 3.7600
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0
### Training results
### Framework versions
- Transformers 4.11.0
- Pytorch 1.8.1+cu102
- Datasets 1.8.0
- Tokenizers 0.10.2
| {"license": "apache-2.0", "tags": ["generated_from_trainer"], "model-index": [{"name": "bert-hateful-memes-expanded", "results": []}]} | limjiayi/bert-hateful-memes-expanded | null | [
"transformers",
"pytorch",
"bert",
"fill-mask",
"generated_from_trainer",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #bert #fill-mask #generated_from_trainer #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us
|
# bert-hateful-memes-expanded
This model is a fine-tuned version of bert-base-uncased on texts from the following datasets:
- Hateful Memes, 'train', 'dev_seen' and 'dev_unseen'
- HarMeme, 'train', 'val' and 'test'
- MultiOFF, 'Training', 'Validation' and 'Testing'
It achieves the following results on the evaluation set:
- Loss: 3.7600
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0
### Training results
### Framework versions
- Transformers 4.11.0
- Pytorch 1.8.1+cu102
- Datasets 1.8.0
- Tokenizers 0.10.2
| [
"# bert-hateful-memes-expanded\n\nThis model is a fine-tuned version of bert-base-uncased on texts from the following datasets:\n- Hateful Memes, 'train', 'dev_seen' and 'dev_unseen'\n- HarMeme, 'train', 'val' and 'test'\n- MultiOFF, 'Training', 'Validation' and 'Testing'\n\nIt achieves the following results on the... | [
"TAGS\n#transformers #pytorch #bert #fill-mask #generated_from_trainer #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n",
"# bert-hateful-memes-expanded\n\nThis model is a fine-tuned version of bert-base-uncased on texts from the following datasets:\n- Hateful Memes, 'train', 'dev_see... |
sentence-similarity | sentence-transformers |
## Modèle de représentation d'un message Twitch à l'aide de ConvBERT
Modèle [sentence-transformers](https://www.SBERT.net): cela permet de mapper une séquence de texte en un vecteur numérique de dimension 256 et peut être utilisé pour des tâches de clustering ou de recherche sémantique.
L'expérimentation menée au sein de Lincoln avait pour principal objectif de mettre en œuvre des techniques NLP from scratch sur un corpus de messages issus d’un chat Twitch. Ces derniers sont exprimés en français, mais sur une plateforme internet avec le vocabulaire internet que cela implique (fautes, vocabulaire communautaires, abréviations, anglicisme, emotes, ...).
Après avoir entrainé un modèle `ConvBert` puis `MLM` (cf section smodèles), nous avons entrainé un modèle _sentence-transformers_ à l'aide du framework d'apprentissage [SimCSE](https://www.sbert.net/examples/unsupervised_learning/SimCSE/README.html) en non supervisée.
L'objectif est de spécialiser la moyenne des tokens _CLS_ de chaque token de la séquence pour représenter un vecteur numérique cohérent avec l'ensemble du corpus. _SimCSE_ crée fictivement des exemples positifs et négatifs supervisées à l'aide du dropout pour revenir à une tâche classique.
_Nous garantissons pas la stabilité du modèle sur le long terme. Modèle réalisé dans le cadre d'un POC._
## Usage (Sentence-Transformers)
Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
```
pip install -U sentence-transformers
```
Then you can use the model like this:
```python
from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]
model = SentenceTransformer('2021twitchfr-conv-bert-small-mlm-simcse')
embeddings = model.encode(sentences)
print(embeddings)
```
## Semantic Textual Similarity
```python
from sentence_transformers import SentenceTransformer, models, util
# Two lists of sentences
sentences1 = ['zackFCZack',
'Team bons petits plats',
'sa commence a quelle heure de base popcorn ?',
'BibleThump']
sentences2 = ['zack titulaire',
'salade de pates c une dinguerie',
'ça commence à être long la',
'NotLikeThis']
# Compute embedding for both lists
embeddings1 = model.encode(sentences1, convert_to_tensor=True)
embeddings2 = model.encode(sentences2, convert_to_tensor=True)
# Compute cosine-similarits
cosine_scores = util.cos_sim(embeddings1, embeddings2)
# Output the pairs with their score
for i in range(len(sentences1)):
print("Score: {:.4f} | \"{}\" -vs- \"{}\" ".format(cosine_scores[i][i], sentences1[i], sentences2[i]))
# Score: 0.5783 | "zackFCZack" -vs- "zack titulaire"
# Score: 0.2881 | "Team bons petits plats" -vs- "salade de pates c une dinguerie"
# Score: 0.4529 | "sa commence a quelle heure de base popcorn ?" -vs- "ça commence à être long la"
# Score: 0.5805 | "BibleThump" -vs- "NotLikeThis"
```
## Entrainement
* 500 000 messages twitchs échantillonnés (cf description données des modèles de bases)
* Batch size: 24
* Epochs: 24
* Loss: MultipleNegativesRankingLoss
_A noter:_
* _ConvBert a été entrainé avec un longueur de 128 tokens max, mais est utilisé pour 512 dans ce modèle. Pas de problème._
* _La loss d'apprentissage n'est pas encore disponible: peu de visibilité sur les performances._
L'ensemble du code d'entrainement sur le github public [lincoln/twitchatds](https://github.com/Lincoln-France/twitchatds).
## Application:
Nous avons utilisé une approche détournée de [BERTopic](https://maartengr.github.io/BERTopic/) pour réaliser un clustering d'un stream en prenant en compte la dimension temporelle: i.e. le nombre de seconde écoulée depuis le début du stream.
Globalement, l'approche donnes des résultats satisfaisant pour identifier des messages dit "similaires" récurrents. L'approche en revanche est fortement influencée par la ponctuation et la structure d'un message. Cela est largement explicable par le manque d'entrainement de l'ensemble des modèles et une volumétrie faible.
### Clustering émission "Backseat":
Entre 19h30 et 20h00:
🎞️ en vidéo: [youtu.be/EcjvlE9aTls](https://youtu.be/EcjvlE9aTls)
### Exemple regroupement émission "PopCorn":
```txt
-------------------- LABEL 106 --------------------
circus (0.88)/sulli (0.23)/connu (0.19)/jure (0.12)/aime (0.11)
silouhette moyenne: 0.04
-------------------- LABEL 106 --------------------
2021-03-30 20:10:22 0.01: les gosse c est des animaux
2021-03-30 20:12:11 -0.03: oue c connu
2021-03-30 20:14:15 0.03: oh le circus !! <3
2021-03-30 20:14:19 0.12: le circus l'anciennnee
2021-03-30 20:14:22 0.06: jure le circus !
2021-03-30 20:14:27 -0.03: le sulli
2021-03-30 20:14:31 0.09: le circus??? j'aime po
2021-03-30 20:14:34 0.11: le Circus, hors de prix !
2021-03-30 20:14:35 -0.09: le Paddock a Rignac en Aveyron
2021-03-30 20:14:39 0.11: le circus ><
2021-03-30 20:14:39 0.04: le Titty Twister de Besançon
-------------------- LABEL 17 --------------------
pates (0.12)/riz (0.09)/pâtes (0.09)/salade (0.07)/emission (0.07)
silouhette moyenne: -0.05
-------------------- LABEL 17 --------------------
2021-03-30 20:11:18 -0.03: Des nanimaux trop beaux !
2021-03-30 20:13:11 -0.01: episode des simpsons ça...
2021-03-30 20:13:41 -0.01: des le debut d'emission ca tue mdrrrrr
2021-03-30 20:13:50 0.03: des "lasagnes"
2021-03-30 20:14:37 -0.18: poubelle la vie
2021-03-30 20:15:13 0.03: Une omelette
2021-03-30 20:15:35 -0.19: salade de bite
2021-03-30 20:15:36 -0.00: hahaha ce gastronome
2021-03-30 20:15:43 -0.08: salade de pates c une dinguerie
2021-03-30 20:17:00 -0.11: Une bonne femme !
2021-03-30 20:17:06 -0.05: bouffe des graines
2021-03-30 20:17:08 -0.06: des pokeball ?
2021-03-30 20:17:11 -0.12: le choux fleur cru
2021-03-30 20:17:15 0.05: des pockeball ?
2021-03-30 20:17:27 -0.00: du chou fleur crue
2021-03-30 20:17:36 -0.09: un râgout de Meynia !!!!
2021-03-30 20:17:43 -0.07: une line up Sa rd o ch Zack Ponce my dream
2021-03-30 20:17:59 -0.10: Pâtes/10
2021-03-30 20:18:09 -0.05: Team bons petits plats
2021-03-30 20:18:13 -0.10: pate level
2021-03-30 20:18:19 -0.03: que des trucs très basiques
2021-03-30 20:18:24 0.03: des pates et du jambon c'est de la cuisine?
2021-03-30 20:18:30 0.05: Des pates et du riz ouai
2021-03-30 20:18:37 -0.02: des gnocchis à la poele c'est cuisiner ?
2021-03-30 20:18:50 -0.03: Pâtes à pizzas, pulled pork, carbonade flamande, etc..
2021-03-30 20:19:01 -0.11: Des pâtes ou du riz ça compte ?
2021-03-30 20:19:22 -0.21: le noob
2021-03-30 20:19:47 -0.02: Une bonne escalope de milanaise les gars
2021-03-30 20:20:05 -0.04: faites des gratins et des quiches
-------------------- LABEL 67 --------------------
1 1 (0.25)/1 (0.19)/ (0.0)/ (0.0)/ (0.0)
silouhette moyenne: 0.96
-------------------- LABEL 67 --------------------
2021-03-30 20:24:17 0.94: +1
2021-03-30 20:24:37 0.97: +1
2021-03-30 20:24:37 0.97: +1
2021-03-30 20:24:38 0.97: +1
2021-03-30 20:24:39 0.97: +1
2021-03-30 20:24:43 0.97: +1
2021-03-30 20:24:44 0.97: +1
2021-03-30 20:24:47 0.97: +1
2021-03-30 20:24:49 0.97: +1
2021-03-30 20:25:00 0.97: +1
2021-03-30 20:25:21 0.95: +1
2021-03-30 20:25:25 0.95: +1
2021-03-30 20:25:28 0.94: +1
2021-03-30 20:25:30 0.94: +1
```
## Full Model Architecture
```
SentenceTransformer(
(0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: ConvBertModel
(1): Pooling({'word_embedding_dimension': 256, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)
```
## Modèles:
* [2021twitchfr-conv-bert-small](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small)
* [2021twitchfr-conv-bert-small-mlm](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small-mlm)
* [2021twitchfr-conv-bert-small-mlm-simcse](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small-mlm-simcse) | {"language": ["fr"], "license": "mit", "tags": ["sentence-transformers", "feature-extraction", "sentence-similarity", "transformers", "twitch", "convbert"], "pipeline_tag": "sentence-similarity", "widget": [{"source_sentence": "Bonsoir", "sentences": ["Salut !", "Hello", "Bonsoir!", "Bonsouar!", "Bonsouar !", "De rien", "LUL LUL"], "example_title": "Coucou"}, {"source_sentence": "elle s'en sort bien", "sentences": ["elle a raison", "elle a tellement raison", "Elle a pas tort", "C'est bien ce qu'elle dit l\u00e0", "Hello"], "example_title": "Raison or not"}, {"source_sentence": "et la question \u00e9nerg\u00e9tique n'est pas politique ?", "sentences": ["C'est le nucl\u00e9aire militaire qui a entach\u00e9 le nucl\u00e9aire pour l'\u00e9nergie.", "La fusion nucl\u00e9aire c'est pas pour maintenant malheureusement", "le pro nucl\u00e9aire redevient acceptable \u00e0 gauche j'ai l'impression", "La mer \u00e0 Nantes?", "c'est bien un olivier pour l'upr", "Moi je vois juste sa lavalli\u00e8re"], "example_title": "Nucl\u00e9aire"}]} | lincoln/2021twitchfr-conv-bert-small-mlm-simcse | null | [
"sentence-transformers",
"pytorch",
"convbert",
"feature-extraction",
"sentence-similarity",
"transformers",
"twitch",
"fr",
"license:mit",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"fr"
] | TAGS
#sentence-transformers #pytorch #convbert #feature-extraction #sentence-similarity #transformers #twitch #fr #license-mit #endpoints_compatible #region-us
|
## Modèle de représentation d'un message Twitch à l'aide de ConvBERT
Modèle sentence-transformers: cela permet de mapper une séquence de texte en un vecteur numérique de dimension 256 et peut être utilisé pour des tâches de clustering ou de recherche sémantique.
L'expérimentation menée au sein de Lincoln avait pour principal objectif de mettre en œuvre des techniques NLP from scratch sur un corpus de messages issus d’un chat Twitch. Ces derniers sont exprimés en français, mais sur une plateforme internet avec le vocabulaire internet que cela implique (fautes, vocabulaire communautaires, abréviations, anglicisme, emotes, ...).
Après avoir entrainé un modèle 'ConvBert' puis 'MLM' (cf section smodèles), nous avons entrainé un modèle _sentence-transformers_ à l'aide du framework d'apprentissage SimCSE en non supervisée.
L'objectif est de spécialiser la moyenne des tokens _CLS_ de chaque token de la séquence pour représenter un vecteur numérique cohérent avec l'ensemble du corpus. _SimCSE_ crée fictivement des exemples positifs et négatifs supervisées à l'aide du dropout pour revenir à une tâche classique.
_Nous garantissons pas la stabilité du modèle sur le long terme. Modèle réalisé dans le cadre d'un POC._
## Usage (Sentence-Transformers)
Using this model becomes easy when you have sentence-transformers installed:
Then you can use the model like this:
## Semantic Textual Similarity
## Entrainement
* 500 000 messages twitchs échantillonnés (cf description données des modèles de bases)
* Batch size: 24
* Epochs: 24
* Loss: MultipleNegativesRankingLoss
_A noter:_
* _ConvBert a été entrainé avec un longueur de 128 tokens max, mais est utilisé pour 512 dans ce modèle. Pas de problème._
* _La loss d'apprentissage n'est pas encore disponible: peu de visibilité sur les performances._
L'ensemble du code d'entrainement sur le github public lincoln/twitchatds.
## Application:
Nous avons utilisé une approche détournée de BERTopic pour réaliser un clustering d'un stream en prenant en compte la dimension temporelle: i.e. le nombre de seconde écoulée depuis le début du stream.
!approche_bertopic_lincoln
Globalement, l'approche donnes des résultats satisfaisant pour identifier des messages dit "similaires" récurrents. L'approche en revanche est fortement influencée par la ponctuation et la structure d'un message. Cela est largement explicable par le manque d'entrainement de l'ensemble des modèles et une volumétrie faible.
### Clustering émission "Backseat":
Entre 19h30 et 20h00:
!1930_2000
️ en vidéo: URL
### Exemple regroupement émission "PopCorn":
## Full Model Architecture
## Modèles:
* 2021twitchfr-conv-bert-small
* 2021twitchfr-conv-bert-small-mlm
* 2021twitchfr-conv-bert-small-mlm-simcse | [
"## Modèle de représentation d'un message Twitch à l'aide de ConvBERT\n\nModèle sentence-transformers: cela permet de mapper une séquence de texte en un vecteur numérique de dimension 256 et peut être utilisé pour des tâches de clustering ou de recherche sémantique.\n\nL'expérimentation menée au sein de Lincoln ava... | [
"TAGS\n#sentence-transformers #pytorch #convbert #feature-extraction #sentence-similarity #transformers #twitch #fr #license-mit #endpoints_compatible #region-us \n",
"## Modèle de représentation d'un message Twitch à l'aide de ConvBERT\n\nModèle sentence-transformers: cela permet de mapper une séquence de texte ... |
fill-mask | transformers |
## Modèle de Masking sur les données Twitch FR
L'expérimentation menée au sein de Lincoln avait pour principal objectif de mettre en œuvre des techniques NLP from scratch sur un corpus de messages issus d’un chat Twitch. Ces derniers sont exprimés en français, mais sur une plateforme internet avec le vocabulaire internet que cela implique (fautes, vocabulaire communautaires, abréviations, anglicisme, emotes, ...).
Nos contraintes sont celles d’une entreprise n’ayant pas une volumétrie excessive de données et une puissance infinie de calcul.
Il a été nécessaire de construire un nouveau tokenizer afin de mieux correspondre à notre corpus plutôt qu’un tokenizer français existant.
Note corpus étant faible en volumétrie par rapport aux données habituelles pour entrainer un modèle BERT, nous avons opté pour l’entrainement d’un modèle dit « small ». Et il a été montré dans la littérature qu’un corpus de quelques giga octets peut donner de bons résultats, c’est pourquoi nous avons continué avec notre corpus.
La limite de la puissance de calcul a été contourné à l’aide d’une nouvelle architecture d’apprentissage basée sur un double modèle générateur / discriminateur.
Ceci nous a permis d’entrainer un modèle de langue ConvBERT sur nos données, ainsi qu’un modèle de masking en quelques heures sur une carte GPU V100.
_Nous garantissons pas la stabilité du modèle sur le long terme. Modèle réalisé dans le cadre d'un POC._
## Données
| Streamer | Nbr de messages | Categories notables en 2021 |
| --------------------------------------------- | --------------- | ---------------------------------- |
| Ponce | 2 604 935 | Chatting/Mario Kart/FIFA |
| Domingo | 1 209 703 | Chatting/talk-shows/FM2O21 |
| Mistermv | 1 205 882 | Isaac/Special events/TFT |
| Zerator | 900 894 | New World/WOW/Valorant |
| Blitzstream | 821 585 | Chess |
| Squeezie | 602 148 | Chatting / Minecraft |
| Antoinedaniellive | 548 497 | Geoguessr |
| Jeanmassietaccropolis/jeanmassiet | 301 387 | Talk-shows/chatting/special events |
| Samueletienne | 215 956 | chatting |
Sur la période du 12/03/2021 au 22/07/2021. La totalité des messages comptent 9 410 987 messages sur ces neufs streamers. Ces messages sont issus du canal IRC, donc n’ont pas subi de modération
Les données d'entrainement du modèle de masking contient 899 652 instances de train et 99 962 instances de test. Les données ont été formaté en concaténant les messages sur une fenêtre de 10s. Cette fenêtre correspond à une fenêtre courte qui regroupe des messages très « proches » temporellement.
* 512 tokens max
* Probabilité du « mask » : 15%
## Application
Voir github public [lincoln/twitchatds](https://github.com/Lincoln-France/twitchatds) pour les détails d'implémentation et les résultats.
## Remarques
* Expérimentation ponctuelle
* Les métriques d'entrainement sont disponibles dans l'onglet _Training metrics_
* Pour une meilleure stabilité, les données doivent être plus hétérogènes et volumineuse. Le modèle doit être entrainé + de 24h.
* Le token `<mask>` fonctionne probablement mieux sans laisser d'espace à gauche. Cela est dû au fait que `lstrip=False` pour ce token spécial.
## Usage
```python
from transformers import AutoTokenizer, ConvBertForMaskedLM
from transformers import pipeline
model_name = 'lincoln/2021twitchfr-conv-bert-small-mlm'
tokenizer_name = 'lincoln/2021twitchfr-conv-bert-small'
loaded_tokenizer = AutoTokenizer.from_pretrained(tokenizer_name)
loaded_model = ConvBertForMaskedLM.from_pretrained(model_name)
nlp = pipeline('fill-mask', model=loaded_model, tokenizer=loaded_tokenizer)
nlp('<mask> les gens !')
```
## Modèles:
* [2021twitchfr-conv-bert-small](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small)
* [2021twitchfr-conv-bert-small-mlm](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small-mlm)
* [2021twitchfr-conv-bert-small-mlm-simcse](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small-mlm-simcse)
| {"language": ["fr"], "license": "mit", "tags": ["fill-mask", "convbert", "twitch"], "pipeline_tag": "fill-mask", "widget": [{"text": "<mask> tt le monde !"}, {"text": "cc<mask> va?"}, {"text": "<mask> la Fronce !"}]} | lincoln/2021twitchfr-conv-bert-small-mlm | null | [
"transformers",
"pytorch",
"tensorboard",
"convbert",
"fill-mask",
"twitch",
"fr",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"fr"
] | TAGS
#transformers #pytorch #tensorboard #convbert #fill-mask #twitch #fr #license-mit #autotrain_compatible #endpoints_compatible #region-us
| Modèle de Masking sur les données Twitch FR
-------------------------------------------
L'expérimentation menée au sein de Lincoln avait pour principal objectif de mettre en œuvre des techniques NLP from scratch sur un corpus de messages issus d’un chat Twitch. Ces derniers sont exprimés en français, mais sur une plateforme internet avec le vocabulaire internet que cela implique (fautes, vocabulaire communautaires, abréviations, anglicisme, emotes, ...).
Nos contraintes sont celles d’une entreprise n’ayant pas une volumétrie excessive de données et une puissance infinie de calcul.
Il a été nécessaire de construire un nouveau tokenizer afin de mieux correspondre à notre corpus plutôt qu’un tokenizer français existant.
Note corpus étant faible en volumétrie par rapport aux données habituelles pour entrainer un modèle BERT, nous avons opté pour l’entrainement d’un modèle dit « small ». Et il a été montré dans la littérature qu’un corpus de quelques giga octets peut donner de bons résultats, c’est pourquoi nous avons continué avec notre corpus.
La limite de la puissance de calcul a été contourné à l’aide d’une nouvelle architecture d’apprentissage basée sur un double modèle générateur / discriminateur.
Ceci nous a permis d’entrainer un modèle de langue ConvBERT sur nos données, ainsi qu’un modèle de masking en quelques heures sur une carte GPU V100.
*Nous garantissons pas la stabilité du modèle sur le long terme. Modèle réalisé dans le cadre d'un POC.*
Données
-------
Streamer: Ponce, Nbr de messages: 2 604 935, Categories notables en 2021: Chatting/Mario Kart/FIFA
Streamer: Domingo, Nbr de messages: 1 209 703, Categories notables en 2021: Chatting/talk-shows/FM2O21
Streamer: Mistermv, Nbr de messages: 1 205 882, Categories notables en 2021: Isaac/Special events/TFT
Streamer: Zerator, Nbr de messages: 900 894, Categories notables en 2021: New World/WOW/Valorant
Streamer: Blitzstream, Nbr de messages: 821 585, Categories notables en 2021: Chess
Streamer: Squeezie, Nbr de messages: 602 148, Categories notables en 2021: Chatting / Minecraft
Streamer: Antoinedaniellive, Nbr de messages: 548 497, Categories notables en 2021: Geoguessr
Streamer: Jeanmassietaccropolis/jeanmassiet, Nbr de messages: 301 387, Categories notables en 2021: Talk-shows/chatting/special events
Streamer: Samueletienne, Nbr de messages: 215 956, Categories notables en 2021: chatting
Sur la période du 12/03/2021 au 22/07/2021. La totalité des messages comptent 9 410 987 messages sur ces neufs streamers. Ces messages sont issus du canal IRC, donc n’ont pas subi de modération
Les données d'entrainement du modèle de masking contient 899 652 instances de train et 99 962 instances de test. Les données ont été formaté en concaténant les messages sur une fenêtre de 10s. Cette fenêtre correspond à une fenêtre courte qui regroupe des messages très « proches » temporellement.
* 512 tokens max
* Probabilité du « mask » : 15%
Application
-----------
Voir github public lincoln/twitchatds pour les détails d'implémentation et les résultats.
Remarques
---------
* Expérimentation ponctuelle
* Les métriques d'entrainement sont disponibles dans l'onglet *Training metrics*
* Pour une meilleure stabilité, les données doivent être plus hétérogènes et volumineuse. Le modèle doit être entrainé + de 24h.
* Le token '' fonctionne probablement mieux sans laisser d'espace à gauche. Cela est dû au fait que 'lstrip=False' pour ce token spécial.
Usage
-----
Modèles:
--------
* 2021twitchfr-conv-bert-small
* 2021twitchfr-conv-bert-small-mlm
* 2021twitchfr-conv-bert-small-mlm-simcse
| [] | [
"TAGS\n#transformers #pytorch #tensorboard #convbert #fill-mask #twitch #fr #license-mit #autotrain_compatible #endpoints_compatible #region-us \n"
] |
feature-extraction | transformers |
## Modèle de langue sur les données Twitch FR
L'expérimentation menée au sein de Lincoln avait pour principal objectif de mettre en œuvre des techniques NLP from scratch sur un corpus de messages issus d’un chat Twitch. Ces derniers sont exprimés en français, mais sur une plateforme internet avec le vocabulaire internet que cela implique (fautes, vocabulaire communautaires, abréviations, anglicisme, emotes, ...).
Nos contraintes sont celles d’une entreprise n’ayant pas une volumétrie excessive de données et une puissance infinie de calcul.
Il a été nécessaire de construire un nouveau tokenizer afin de mieux correspondre à notre corpus plutôt qu’un tokenizer français existant.
Note corpus étant faible en volumétrie par rapport aux données habituelles pour entrainer un modèle BERT, nous avons opté pour l’entrainement d’un modèle dit « small ». Et il a été montré dans la littérature qu’un corpus de quelques giga octets peut donner de bons résultats, c’est pourquoi nous avons continué avec notre corpus.
La limite de la puissance de calcul a été contourné à l’aide d’une nouvelle architecture d’apprentissage basée sur un double modèle générateur / discriminateur.
Ceci nous a permis d’entrainer un modèle de langue ConvBERT sur nos données, ainsi qu’un modèle de masking en quelques heures sur une carte GPU V100.
_Nous garantissons pas la stabilité du modèle sur le long terme. Modèle réalisé dans le cadre d'un POC._
## Données
| Streamer | Nbr de messages | Categories notables en 2021 |
| --------------------------------------------- | --------------- | ---------------------------------- |
| Ponce | 2 604 935 | Chatting/Mario Kart/FIFA |
| Domingo | 1 209 703 | Chatting/talk-shows/FM2O21 |
| Mistermv | 1 205 882 | Isaac/Special events/TFT |
| Zerator | 900 894 | New World/WOW/Valorant |
| Blitzstream | 821 585 | Chess |
| Squeezie | 602 148 | Chatting / Minecraft |
| Antoinedaniellive | 548 497 | Geoguessr |
| Jeanmassietaccropolis/jeanmassiet | 301 387 | Talk-shows/chatting/special events |
| Samueletienne | 215 956 | chatting |
Sur la période du 12/03/2021 au 22/07/2021. La totalité des messages comptent 9 410 987 messages sur ces neufs streamers. Ces messages sont issus du canal IRC, donc n’ont pas subi de modération
Les données d'entrainement sont basé sur le format d'entrainement du modèle ELECTRA. Cela nécessite de formater les données en paragraphe, séparés par phrase. Nous avons choisi de regrouper les messages dans une fenêtre de 60 secondes, faisant office de paragraphe, avec les conditions suivantes :
* Longueur supérieure à 170 (ce qui représente en moyenne 50 tokens) afin de ne pas créer des instances ayant pas d’information car majoritairement vide : un padding sera nécessaire et pénalise la vitesse d’apprentissage.
* 128 tokens maximums (défaut)
Si la longueur maximale est atteinte, une deuxième instance est créée. Au final, la volumétrie d'instance d'entrainement est de 554 974.
## Application
Voir github public [lincoln/twitchatds](https://github.com/Lincoln-France/twitchatds) pour les détails d'implémentation et les résultats.
## Remarques
* Expérimentation ponctuelle
* Les métriques d'entrainement sont disponibles dans l'onglet _Training metrics_
* Pour une meilleure stabilité, les données doivent être plus hétérogènes et volumineuse. Le modèle doit être entrainé + de 24h.
## Usage
```python
from transformers import AutoTokenizer, ConvBertModel
from transformers import FeatureExtractionPipeline
model_name = 'lincoln/2021twitchfr-conv-bert-small'
loaded_tokenizer = AutoTokenizer.from_pretrained(model_name)
loaded_model = ConvBertModel.from_pretrained(model_name)
nlp = FeatureExtractionPipeline(model=loaded_model, tokenizer=loaded_tokenizer)
nlp("<3 <3 les modos")
```
## Modèles:
* [2021twitchfr-conv-bert-small](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small)
* [2021twitchfr-conv-bert-small-mlm](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small-mlm)
* [2021twitchfr-conv-bert-small-mlm-simcse](https://huggingface.co/lincoln/2021twitchfr-conv-bert-small-mlm-simcse)
| {"language": ["fr"], "license": "mit", "tags": ["feature-extraction", "convbert", "twitch"], "pipeline_tag": "feature-extraction", "widget": [{"text": "LUL +1 xD La Fronce !"}]} | lincoln/2021twitchfr-conv-bert-small | null | [
"transformers",
"pytorch",
"tf",
"tensorboard",
"convbert",
"feature-extraction",
"twitch",
"fr",
"license:mit",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"fr"
] | TAGS
#transformers #pytorch #tf #tensorboard #convbert #feature-extraction #twitch #fr #license-mit #endpoints_compatible #region-us
| Modèle de langue sur les données Twitch FR
------------------------------------------
L'expérimentation menée au sein de Lincoln avait pour principal objectif de mettre en œuvre des techniques NLP from scratch sur un corpus de messages issus d’un chat Twitch. Ces derniers sont exprimés en français, mais sur une plateforme internet avec le vocabulaire internet que cela implique (fautes, vocabulaire communautaires, abréviations, anglicisme, emotes, ...).
Nos contraintes sont celles d’une entreprise n’ayant pas une volumétrie excessive de données et une puissance infinie de calcul.
Il a été nécessaire de construire un nouveau tokenizer afin de mieux correspondre à notre corpus plutôt qu’un tokenizer français existant.
Note corpus étant faible en volumétrie par rapport aux données habituelles pour entrainer un modèle BERT, nous avons opté pour l’entrainement d’un modèle dit « small ». Et il a été montré dans la littérature qu’un corpus de quelques giga octets peut donner de bons résultats, c’est pourquoi nous avons continué avec notre corpus.
La limite de la puissance de calcul a été contourné à l’aide d’une nouvelle architecture d’apprentissage basée sur un double modèle générateur / discriminateur.
Ceci nous a permis d’entrainer un modèle de langue ConvBERT sur nos données, ainsi qu’un modèle de masking en quelques heures sur une carte GPU V100.
*Nous garantissons pas la stabilité du modèle sur le long terme. Modèle réalisé dans le cadre d'un POC.*
Données
-------
Streamer: Ponce, Nbr de messages: 2 604 935, Categories notables en 2021: Chatting/Mario Kart/FIFA
Streamer: Domingo, Nbr de messages: 1 209 703, Categories notables en 2021: Chatting/talk-shows/FM2O21
Streamer: Mistermv, Nbr de messages: 1 205 882, Categories notables en 2021: Isaac/Special events/TFT
Streamer: Zerator, Nbr de messages: 900 894, Categories notables en 2021: New World/WOW/Valorant
Streamer: Blitzstream, Nbr de messages: 821 585, Categories notables en 2021: Chess
Streamer: Squeezie, Nbr de messages: 602 148, Categories notables en 2021: Chatting / Minecraft
Streamer: Antoinedaniellive, Nbr de messages: 548 497, Categories notables en 2021: Geoguessr
Streamer: Jeanmassietaccropolis/jeanmassiet, Nbr de messages: 301 387, Categories notables en 2021: Talk-shows/chatting/special events
Streamer: Samueletienne, Nbr de messages: 215 956, Categories notables en 2021: chatting
Sur la période du 12/03/2021 au 22/07/2021. La totalité des messages comptent 9 410 987 messages sur ces neufs streamers. Ces messages sont issus du canal IRC, donc n’ont pas subi de modération
Les données d'entrainement sont basé sur le format d'entrainement du modèle ELECTRA. Cela nécessite de formater les données en paragraphe, séparés par phrase. Nous avons choisi de regrouper les messages dans une fenêtre de 60 secondes, faisant office de paragraphe, avec les conditions suivantes :
* Longueur supérieure à 170 (ce qui représente en moyenne 50 tokens) afin de ne pas créer des instances ayant pas d’information car majoritairement vide : un padding sera nécessaire et pénalise la vitesse d’apprentissage.
* 128 tokens maximums (défaut)
Si la longueur maximale est atteinte, une deuxième instance est créée. Au final, la volumétrie d'instance d'entrainement est de 554 974.
Application
-----------
Voir github public lincoln/twitchatds pour les détails d'implémentation et les résultats.
Remarques
---------
* Expérimentation ponctuelle
* Les métriques d'entrainement sont disponibles dans l'onglet *Training metrics*
* Pour une meilleure stabilité, les données doivent être plus hétérogènes et volumineuse. Le modèle doit être entrainé + de 24h.
Usage
-----
Modèles:
--------
* 2021twitchfr-conv-bert-small
* 2021twitchfr-conv-bert-small-mlm
* 2021twitchfr-conv-bert-small-mlm-simcse
| [] | [
"TAGS\n#transformers #pytorch #tf #tensorboard #convbert #feature-extraction #twitch #fr #license-mit #endpoints_compatible #region-us \n"
] |
text2text-generation | transformers |
# Génération de question à partir d'un contexte
Le modèle est _fine tuné_ à partir du modèle [moussaKam/barthez](https://huggingface.co/moussaKam/barthez) afin de générer des questions à partir d'un paragraphe et d'une suite de token. La suite de token représente la réponse sur laquelle la question est basée.
Input: _Les projecteurs peuvent être utilisées pour \<hl\>illuminer\<hl\> des terrains de jeu extérieurs_
Output: _À quoi servent les projecteurs sur les terrains de jeu extérieurs?_
## Données d'apprentissage
La base d'entrainement est la concatenation des bases SquadFR, [fquad](https://huggingface.co/datasets/fquad), [piaf](https://huggingface.co/datasets/piaf). L'input est le context et nous avons entouré à l'aide du token spécial **\<hl\>** les réponses.
Volumétrie (nombre de triplet contexte/réponse/question):
* train: 98 211
* test: 12 277
* valid: 12 776
## Entrainement
L'apprentissage s'est effectué sur une carte Tesla V100.
* Batch size: 20
* Weight decay: 0.01
* Learning rate: 3x10-5 (décroit linéairement)
* < 24h d'entrainement
* Paramètres par défaut de la classe [TrainingArguments](https://huggingface.co/transformers/main_classes/trainer.html#trainingarguments)
* Total steps: 56 000
<img src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAj0AAAGOCAYAAAB8J7JHAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsMAAA7DAcdvqGQAAEKXSURBVHhe7d1/sB11fcf/zHSmihVDKagFpqQDrVUcQptaf9Q2acf6ox1M1NKpVkpGWrRaJ3FqW6f/JPX7HR0pNSkWKUMRqEz6bRSj8kMRh8QREWyQYtFCgw0Uwy8hKYIKWrvf+9y77+Rz9+45d2/u+bHnfJ6Pmc/cuz/Ont09e3Zf57Of3V1WSJIkZcDQI0mSsmDokSRJWTD0SJKkLBh6JElSFgw9kiQpC4YeSZKUBUOPJEnKgqFHkiRlwdAjSZKyYOiRJElZMPRIkqQsGHokSVIWDD2SJCkLhh5JkpQFQ48kScqCoUeSJGXB0CNJkrJg6JGm0I9+9KPioosuKu64446qz9Lt3Lmz2Lx5c/l31B5//PHi0ksvLd+fMizDnn4bXZiHQZmmZdF0MPRo4nEQXrZs2VgOxl31gx/8oFwnf//3f1/1aS/CTd041/OqVauKFStWFGvWrCnLsAx7+qlNmzYVe/furboO4f2nJSiwvRh61CWGHk08Q898Swk9HIx5bV2EoVGv50996lPl/Hz5y1+u+gzPKENPr212mkIPy2HoUZcYejTxFht6brvttuLAgQNVVzPG2bVrV9U1H7/QGd70S30hgwwNzGeTYYSehfSalzrWWdtxwUFzoflpOz3G6zduPfTwWX3rW9+quuZjO2I7WGh7Ckwvtpm2oYdxHn744aqrGdM8nG2xrcVMP11GqWsMPZp47GTbhJ44mEc57bTT5u2cN2zYUBx11FFzxjvrrLOqobMHTU6zpMPpXggHxXXr1s15Hd1xsGQ+6Ee7lbqY7xiXv+vXr58zLQ6U6YG3KfSsXr26LHX04z1QX0dR0Gs9s87ScZvWK/0ZL10HrOcdO3ZUYzRjuWL8KPH+vAfvlQ6r1yrEPKefW9M6CBF6tm3bNudzfte73lWNMatpO2Be6oEq1nm6Xtme0tdFieWK0EM5+uijDw4///zzy+Eptpd0e+X/9PPp9XlS4jPvh2nVp1//zJgOy5jOS0yb/+ufiTROhh5NvDiwpTv7utj5b9mypQwHHJxWrlxZHrgiLNAvxgn0S3fyy5cvLw9a8RoOvE1Bpe7EE08s3y/mkb/0IwQEDhwcOOsYLw1e/M98xPsyf3RzsAyHG3oQ66quaT3X1yvDYr2mGIcDIuOzziiMR780rNUxPQ6a8b5RkK5TpsE8MF66LAyjH+OyvhivHshSEXpOOeWU4pZbbikeffTRg++fHrzjc495T7enFOuWZUznM94/lqmO9z/99NOLjRs3ltNlHPr9xE/8xJx55/2ZRqx7SgSqGI+/vD4tzBPjMO1+4vvANJkOJabPdALrO5Yxvivx/oxr6FGXGHo08dgB13fEdRFWUrFTj5ATB/A4kDVh+EK1E3VxcIoDQWA6af+m8WLZ0oMJ3emBHfHaGG8UoYf1RHd9vcZ4zFOgu/7e9en1EqEjFQGn/to4KId4j/r66oVwwfgf+chHqj6zTj311OK4444rnnzyyarPfLE9pfMUAaP+2aM+bmAenvvc5855zfbt28vx+RvYpqk9qyPgNfVHbCfpZ9PL2rVry/eoY/oMC7G99FpGQ4+6xNCjiRcHtqYDCJoORiENQxEo+LW+devWxl/CUTvBr/BPfvKTVd/+4pQZO/+0MI10vggRzE96gGbeOMiEfstK/3jtKEJP23lBvRsRmppen2JdMV5qEPPYJEJPvQ3NO97xjrL/N7/5zarP7LT5DHlNFMZJQ3GvdY5e88V0OH1Zx/isC8S2ynjpNkVh+216z1gX6efw3e9+t7j++uvnFLYdsK2n4SbU1z3dTeEI6TxLXWDo0cRb6MDWb3j9oETQIWiwE+c1HEDSgxgHakIMQYThbXbqTJ/pxXvVSxqueO84RcJ7EZbSX+1Rw8GwOvrHAW3coad+wEznLdXr9SnWb31+mOemU4G95rGtCC919enG58DnxXKxjcQ46XL2WudIp5fi/Zu2KcaP/vFerOd4j7TUa3rYxtiWIuCHPXv2lNNJy/79+8th/N/0mdW3D7p5zyaMZ+hRlxh6NPHqB6S6+FWchpdA/16nAjhQRM1OE6YbB4B+pwuipqeNdFniVER62qDXskatSRykFhN66rVL9YNaqL93dPdar+k0690hnV4vTaGn1zwyL+k0Yx7bitqaXjU99957b9nddAqJ7YVx0uXstc6RzmeqTeiJbbrNaSq2DYI023Ld//3f/xU//OEP55TQa95Z7rRmh+Xtt4yGHnWJoUcTLw5s/Q6eHKTqv+Djdf0OHHEQTYNHHcObDughptEUDpowr/wi50BSP1AxHxxw6r/Yo+Yhao2aQg/zGLVIIdZBOv8Rtuqa1nPTvMTr0+Wtv0eoT69JU+iJdVr/7Fhn6QE55rmtCD1NbXq4QWJgnPry0F3v3ys4gHGjPVmqTehB0zZdR+ChRoztqKl2sB8+V8J6+jr+r9cYsbyGHk0KQ48mXhzYmto3xA43DpJcLcV9VWizw847DRUcgJgGbXUYh79xwADvw0GG1zKcEpdg9wtF4CDBeLQBidcynaZTNBxEmDfGbwpkcXCNabGMdKc1D02hJ9ZTug44cFLSA3XUIjBeug7j9fwNEXDq67V+EGSc9D1CfXpNYvnqeA/eKz6P+CzSsBXz3BafL+U5z3lOeYn4VVddVZx55pnlNHisR4hAcNlll5XvzWfBdsJ46XL2Cz2MTwiNdRzbUNvQE8vG+DEf/GUbjnlgm2Cc9LOMstB6J+AQINlGmW58H+iXbu+8l6FHk8LQo4lH7UYcXJpK4GBINzttDjgcENJfsRwE4ooVdtZR4xLjsKOnOw5ujMf4Cx08AqGK9+e1FP5vCgK8T8x7On8pwkZMi/mpT4fTFL/+679efOITn6j6zOJ1Mf+8nnXHeqiHq1gXMR/RjwNsfXljvca8pOErMLwpwMU89MNBk/dtwnuly5MGHsS20RbvQ/n6179enHHGGeV0qeW55JJLqjFm8bmwLcS2wrqKzy1dTuavaX2A8RnGa9L1wPs3BYWm/kyD92ZbZT5im41ppdOvl6bPoy6dfmzv9c+L6fRaxl7LIo2LoUeSJGXB0CNJkrJg6JEkSVkw9EiSpCwYeiRJUhYMPZIkKQuGHkmSlAVDT4Mbb7yxOPLII4uXvvSlFovFYrFkUZ71rGfNuaHpNDL0NLj11lvLu61ec801FovFYrFkUV74wheWd9+eZoaeBv/xH/9R3oZekqRccBd3HjcyzQw9DQw9kqTcGHoyZeiRJOXG0JMpQ48kKTeGnkwZeiRp8b773e9aOlwWYujJlKFHktr54Q9/WOzbt6+48847i2984xuWDpf//M//LB544IHqk5vP0JMpQ48ktfPII48Ud911V/Hoo48WP/jBD8oQZOleeeqpp8rPivDz+OOPV5/eXIaeTBl6JKmde+65p7j//vurLnXd3r17e35ehp5MGXokqR1qeQ4cOFB1qeseeuihMvg0MfRkytAjSe0YeiYLoYfauSaGnkwZeiSpnUkLPRz0//Vf/7Xqyo+hR/MMOvRceunMip5Z02vXVj0kaUpMWuj56Ec/WrzgBS+ouvJj6NE8gw49O3fOhp7Vq6sekjQlDD2TxdCjeQw9ktQO9+eZ9NBDw97NmzeXZSc77JpLL7304HCeQp4ub31Y1xl6NI+hR5LamfTQs2PHjuKoo44qNmzYUGzatKlYvnx5+X/g/5UrV5bDKGvXrj0YjFbP7NTrw7rO0KN5Bh16brttNvSsWFH1kKQp0XR666UvHW15zWuqN26hHnpOPPHEMrCE22Z22Mtmdtj8Bf831f6g37CuMvRonkGHHhB6KJI0TZpCz5FHHtrnjaIcbuhhvtOAE6i92bJlS/k/tTkrZn6xbt26tXG80047rXFYVxl6NI+hR5LaaQo9t9xSFDfdNLryjW9Ub9xCGnqiVqc+/wSdqP1hGAGIU1ec+iLkxPgxjPGZzpo1a+ZNq2sMPZrH0CNJ7Uz61VuEFdr1pAg39X5gOeunw0IMixqirjL0aJ5hhp4J2jdI0oImPfScddZZc2pv6Ca8RDdXZ8X/XOXFqS76oWlYU1jqEkOP5hlG6OHKLULPhLV5k6S+Ji30XHnllXNCD/POFVrU+FA4VZW2z4lTV1HSK7to0xP961d9dZWhR/MYeiSpnUkLPbkz9GgeQ48ktWPomSyGHs1j6JGkdgw9k8XQo3mGEXo41UvoaWj0L0kTy9AzWQw9E4DLA2ldHw3GFoMW9dxifDGvG0boIewYeiRNG0PPZCH0cFxsYujpCC4LpHD/g8WGHlreRwv7tgw9ktSOoWeyGHomCM84WUx4iTtlUlNk6JGkwTP0TBZDzwRZTOjhQ+UGU/ztQujhXlbMwgQ8hFeSWjP0TBZDzwRZTOihhiduB75Q6Ln++uuLZz/72QfLT/3UT5U3mhokrtpiFriKS5KmhaFnshh6Jkjb0BOntcJCoeeJJ54o7rzzzoPl2muvLcPPIBl6JE0jQ89crIvf/d3fLfbv31/16Y3xPvGJT1Rdo2HomSBtQw+Bh2eg8MRbCv/zOv5v8/j/YZzeMvRImkaGnrnuv//+8nizb9++qk9vr3zlK4v3ve99VddoEHq8ZH1CtA09XOlF7U4UQhCv4/9eCTc1jNDD2zLrM/lLkqaGoWcuQ0+3TUToIexs3ry5WL9+fbkx8T8l7Nq1q3j6059e3HvvvVWfuRY6vVU3jNADZmERsyFJnTeJoYdjCjX/HBc4E7B169ZqyOxT1utPSmf8tdVVKPxwXrduXflaCveQY3hYSuhJp818bdy4sRoyi2NZ3HeOv+kDTvsNSxl6JgA1N9TW1Ev493//9+J1r3td8fDDD1d95orXt2XokaR2GkPPrbcWxVe+MroyMw9t0cSBYMBxAQQWrvSNbsICgShF4IkQQTBJQxFtSLnwJdbB4YYeXk/QIXTxf8wXYQbMN+8TASvGQYwbZzLSYXWGHs1j6JGkdhpDz5FHHtrhjaK85jXVGy8sDTAhvfiF4EBoSQME3U3tQRmHMw2ElQhChxt6CF0ElxT9qLUBIYb3iflK8d4Mm/c5NDD0aJ5hhR6uguf72WK7lKSJ0Bh6XvSi0ZZFhB7u0E9AiAtdKJyiol9gnKhhIRDRHVhWXkOtS5x14P8Y/3BDD6+vn5Eg6DCtWL8ENrqZX5p4RH/+8lqGMW/psDpDj+YZVuhheyb09Kh1lKSJM2lteggHcQ+3XqhhiRBE4EnHp5Yo2veENCQNI/SkqOlh/hg3DWroNywYejSPoUeS2pm00ENooaakH5aHsEHY4W+6fASKCDggaDDOUkNPtDVKT18xzfoprxDvm44fmsJSMPRoHkOPJLXDDV0nKfQwr9TMxGkgClcG12t/aFBMcKjX6kQ7G17HVV9MK21wfLihB7xXnLriyi2mE22FmD7zybDLLrusvMorTrvFMOaHwrLFsDpDj+Yx9EhSO5MWegIhh1ofCv/Xa0yoeSFMNDVgJvgQihjO6+iOq6W4w/+5555bPP7442V3P9u3by9uvvnmqmsW02Ke6u/N+9TnOdZ7v2F1hh7NM6zQM7MNl6GHv5I0DSY19OTK0KN5DD2S1I6hp7cbbrih+PjHP95Y2tQEDYOhR/MYeiSpHUNPb29961t7lgcffLAaa7QMPZrH0CNJ7Rh6JouhR/MMK/TQCJ/QU7sYQJImlqFnshh6NM+wQg+N+wk9XMUlSdNgz549xbe//e2qS133rW99q+fDuQ09mTL0SFI7tE35r//6r+J///d/qz7qqqeeeqo8vu3fv7/qM5ehJ1OGHklqh/vSsM/8xje+UdYgWLpZOKXFZ8QdtH/4wx9Wn95chp5MDSv0cO8rQk+PR6JI0kT60Y9+VDz22GPFAw880Fi4S/GgStP0R1ma5mmUpWme2paFLpM39GRqWKEHhB6KJEldYujJlKFHkpQbQ0+mDD2SpNwYejI1zNCzfPls6PG2FpKkLjH0ZGqYoccnrUuSusjQkylDjyQpN4aeTBl6JEm5MfRkytAjScqNoSdTwww9PmldktRFhp5MGXokSbkx9GTK0CNJyo2hJ1OGHklSbgw9mRpm6NmxYzb0rF1b9ZAkqQMMPZkaZujhqi1CD1dxSZLUFYaeTBl6JEm5MfRkytAjScqNoSdTwww9e/fOhp4VK6oekiR1gKEnU8MMPSD0UCRJ6gpDT6YMPZKk3Bh6MmXokSTlxtCTqWGHnuXLZ0PPgQNVD0mSxszQ0xEHZtLBrl27is2bN5eljSuvvLI4++yzi9e97nXFhRdeWNx9993VkIUNO/T4pHVJUtcYejpi9UxKWL58ebFy5cqZsLDwLBOMTj755GL9+vXFxo0bi9NOO6183Y033liN0Z+hR5KUG0NPR1DTg50zKaFN6Ln22mur/2bdeuutxdOe9rTibW97W9WnP0OPJCk3hp6OaRt6mpx66qllzU8bhh5JUm4MPR1zuKHnuuuuK193ySWXVH3muu+++4pLL730YHnf+95XHHPMMdXQwfNJ65KkrjH0dMzhhp4Xv/jFfRtA7969u1i7du3B8pu/+ZvFT/7kT1ZDB8/QI0nqGkNPxxxO6DnllFNaX/EVhn16y9AjSeoaQ0/HLCb0PPjgg8WP//iPF6961auqPu0ZeiRJuTH0dEyv0EObnL/7u78rvvOd75TdXK11wgknFG95y1vK7sUadujZsWM29KxdW/WQJGnMDD0dsWnTpjLs1Eu44YYbyu577rmn7OZ0Vn1cypo1a8rhCxl26JnJbjPzM3sVlyRJXWDo6Yi9e/eWtTz1Eh577LHipptuKp588smyu2ncKG0YeiRJuTH0ZMrQI0nKjaEnU8MOPXv3zoaeFSuqHpIkjZmhJ1PDDj0g9FAkSeoCQ0+mDD2SpNwYejJl6JEk5cbQk6lRhJ7ly2dDT/UAeUmSxsrQk6lRhB6ftC5J6hJDT6YMPZKk3Bh6MmXokSTlxtCTKUOPJCk3hp5MjSL0+KR1SVKXGHoyZeiRJOXG0JMpQ48kKTeGnkwZeiRJuTH0ZGoUoWfHjtnQs3Zt1UOSpDEy9GRqFKGHq7YIPVzFJUnSuBl6MmXokSTlxtCTKUOPJCk3hp5MjSL07N07G3pWrKh6SJI0RoaeTI0i9IDQQ5EkadwMPZky9EiScmPoyZShR5KUG0NPpkYVepYvnw09Bw5UPSRJGhNDT6ZGFXp80rokqSsMPZky9EiScmPoyZShR5KUG0NPpgw9kqTcGHoyNarQ45PWJUldYejJlKFHkpQbQ0+mDD2SpNwYejJl6JEk5cbQk6lRhZ4dO2ZDz9q1VQ9JksbE0JOpUYUertoi9HAVlyRJ42ToyZShR5KUG0NPpgw9kqTcGHqmAAHm+9//ftXVzqhCz969s6FnxYqqhyRJY2Lo6YhLL720WLNmTXHUUUfNhIR2s3zHHXcUL3nJS8rxjzzyyGLz5s3VkIWNKvSAxWm5SJIkDY2hpyM2bdpUli1btrQOPS960YuKM844o9i3b19x8803l8Hnoosuqob2Z+iRJOXG0NMxO3fubBV6br/99nK82267repTFO985zuLVatWVV39GXokSbkx9HRM29Czffv24ogjjqi6ZsVrqflZyChDz/Lls6HnwIGqhyRJY2Do6Zi2oeeDH/xgceqpp1Zds+K1u3fvrvoc8oUvfKF4/vOff7CcdNJJZfuhUfBJ65KkLjD0dEzb0EPbn16h59/+7d+qPofs37+/+NKXvnSwbNu2rTj22GOrocNl6JEkdYGhp2Pahp4dO3b0PL317W9/u+rT2yhPbxl6JEldYOjpmLah57777ivH47L1wCXrr3jFK6qu/gw9kqTcGHo6Yu/evcWuXbuKrVu3lmGG/ymBU1LHH398GXbCS1/60vKSdV7L8Gc+85nFJZdcUg3tb5ShxyetS5K6wNDTEdyccPXq1fNK+OpXv1r82q/9WvHggw9WfWZvTviyl72sDDs0TO7qzQkNPZKkLjD0ZMrQI0nKjaFnSGibc6DDN6Yx9EiScmPoGQBOTXEJeVi5cmXZLof74BB+umiUoWfHjtnQs3Zt1UOSpDEw9AzA2pmjOZeQg7/Lly8va3kIQgzrolGGHnIfoSdpoiRJ0sgZegaABsdRo7Nhw4birLPOKv8n+Jx44onl/11j6JEk5cbQMwCEHJ6QDkIOp7vAw0ANPYYeSVI3GHoGgPvkcEqLdjy054kGzJ7emjWzesrQs2JF1UOSpDEw9AwIQad+xRY1PQSiLhpl6AGhhyJJ0rgYeoaA4MPdlLsaeGDokSTlxtAzAJzGisbLiEvWKXFVV9eMOvQsXz4beqrmTpIkjZyhZwC4estL1vtj9RB6jjqKmrCqpyRJI2ToGYB+l6wTgLpo1KEHXL1F8Fm3ruohSdIIGXoGIC5ZJ+SsWLHCS9Z7oIlTnObq6Fk/SdIUM/QMgJestzezSjzNJUkaC0PPANWfs+Ul683iNFfS9luSpKEz9AwYwYfL1aO2p6vGGXrS01y1nChJ0tAYegaENj08VT0uVaesX7++Gto94ww94KkdhB7u0uxpLknSKBh6BiAuU48GzKDGh/Y9XM3VReMOPZhZPWXw6egqkiRNGUPPAHD1Vhp4AsHntNNOq7q6pQuh57bbZkMPxdNckqRhM/QMAFdoNYUeGjJT29NFXQg9SE9zSZI0TIaeASDwcH8eQk7gqi1qeTy9tbA4zUXwscZHkjQshp4BIdxEA+Zo0Mydmrt6FVeXQg9ZkXs4Enwo3LG5o1f6S5ImmKFngKjdoVEzJa316aIuhZ7Aqa64lJ2bF27dWg2QJGkADD1DRENmanu6qIuhB9TwcBPrqPWhHbinvCRJg2DoGSJDz+Ej6KSnvDZurAZIknSYDD1DZOhZGppDxdVdlDVrvJGhJOnwGXqGyNAzGDSPirY+nO7qeHMpSVJHGXqWgFDDc7Z6la1bt+YTeqiC2bx59rzUEDD5uLSdRs5DehtJ0hQz9CxBXKLer2QTeuI8FOeghoTgw5PZeRtKw/0gJUnqydCTqaHU9ETL4yGnkbSdT4ef6SqpLfYfHIiGWFsswdCTqaG06SHskERG8EwJ3ira+QyxcknSsNA4j5tx8QWOXzFRuEMpQUgaMENPpobWkHlEtT1IGzhv2VL1lNRdfGmpnuWHURpyKDQF4LE96R1Kd+yoXigNhqEnU0MLPeykYoc1gl9q1ISP8O0kHS4CD1/UCDn8QKKRHvuM9MvLHUoJQDEeNUE+l0YDYujJ1NBCD2KHReObERjx20laLEJN1O7whW1z34n0HDZhyefSaAAMPZkaaugZcfVL+nb+IJQ6hn0AN9jiS8p9JxazT2Dc9Lk0BKfLLhvJfkXTydCTqaGGHoy4+iUuZeevpI5YSuBJ8csmfS4Nv3BoG+SVXlokQ0+H3H777cUll1xSXHTRRcX1119f9e3tO9/5zsx3fmdx3nnnFdu2bSvuuuuuasjChh56qL6OHdQIql94ixG+naQ2CCZ8KTlNNYjaGU55pe19KNT+cOrL2h+1YOjpiB07dpQhZP3MTuKcc86Z+S4vK7Zv314NbfaqV72qePWrX138+Z//eXHmmWeWryEEtTH00IMRV7/E23kJu7RE/HLYtavqOExp4GnThmcxmD+u9EprfyjW/GgBhp6OWDNzpN7Mjbkq/L9q1aqqa77HHnts5ju+rLjjjjuqPkXx2te+tpxOGyMJPSOufuGHXrR7dN8nHQa+pxFWKNSibNy4+NBCIOH1wwg8dVz9FbU/tP+R+jD0dMCTTz45831dVtb2hL0zOx/6XX311VWfuR599NFyeFobRM1Pp0IPYuc3ouqXETwNQ5o+/GJIww6lXotC2xxOI/X6AUO4oXaIceI1yT5tqJj/eM8R/MDS5DL0dMDdd989811dVuzZs6fqM4t+tO/pZcuWLcUb3/jG4mMf+1jx3ve+t3j5y19e3HDDDdXQuR566KHi05/+9MHy4Q9/uDjmmGOqoUM04uqX9O1Gtb+VJhZfGGqY0/vncJ44ggNBJr1hYBQCEL8smm4yGGUENyidI67y8k6l6sPQ0wG0wyHg1NEvPeVVd8UVVxTHH3/8zP7ntOLII48szj777DLcNLnpppvKUBSFU2dHsaMbhah+YUfJTnbI2NfyduyPJfXQL+w04VdENJxrKlydxWkmwsc4fnHwnsyHX3z1YejpgDiV1VTTc/nll1ddc9028wuM4buqxoYPP/xw8eY3v7l4wxveUHYvZGSnt5BWv7CTHUGNzwifhiFNnvhlQCGoLOY7yfeZ8SnDbq+zWPHFH8E+RpPJ0NMRBJhrrrmm6iqKffv2lf0+//nPV33mogao3n4naozaXME10tADfkGml5rSOHKItT7xo4+MNYLKJWmysO/gCzJtp4KiDeGIrhjV5DH0dMQZZ5wx51QW/3MaKjz44IPl6awnnnii7P7Upz41891eVuzevbvsRlzq/u1vf7vq09vIQ0+IU10UqqGH+IsszVjs41m9bOu2c1TW+M7xpaD2ddp+EfDlZtn8taMeDD0dcdVVVxXHHXdcGX7injvplVk0UKbfPffcU/Upil/5lV8pVq5cOfPjZkOxbt26mX3Y8uIDH/hANbS/sYUeUCXO+X92TpQk7A0Sb1O/ACUK+8SZVVbe0V7KSjT4ndaH1cWvHc9tq4Ghp0O+9rWvFRdccEFx/vnnF9ddd13Vd9YjjzxSBqPvfe97VZ9Zn/3sZ4tzzz23uPjii+fU+ixkrKEnpLU+NHIeUq1PNEHg7dgf1i9EocLJ8KMsRE0IZVprQqK9EvsUqcbQk6lOhB6QRtLqGM5DDfGUV2Dfz74xfWvDj6ZeXH01zW1eCHPxy8Zz2aox9GSqM6EH7KSohkmrYEYUfmD4URb4nsUl6tMeBiLc0bBZShh6MtWp0BM6GH54+yi0AaL5URS+N9N6hkBTKE4nc4532tGgj2Ul5EkJQ0+mOhl6QlP44fx8v1vgD1A9/CxUmDWuwF/q8xmloYpanhH9iBi7+BJ7a3YlDD2Z6nToCU3hJ1LGCAIQx4a0sO9kdqKkl8SnhVohZk/qjGjcy1WTueAeRCzzYh9Cyn6FXzAUDo5U6/qLZmoYejI1EaEnReLgPH09AMV152Nso0AgoulAehU+xR+Y6gzO1bJREn5ywY+m+DL22j/QnwetRi3YQoVfNIagiWboydTEhZ4UO+6410ha4jzTGDfoqJxidjjO2OZHY0cqZ4PkdE9uej2ElC8m4SX2HWlhPVGNS+H1Tb9ookQI4ofXUoMQ88Q0mBbTZF8WDQqjsI9L35txGL9rjwPpMENPpiY69AR2EhGA6jVAsVPgPNMYdgixj/TiEY0d3wM2xhyfPh7Po+EXSCBQpDU71CC3qSlmf8P0+oUgSlwBQSDhvfoVapni8xlEqb9vTL9eRtQ+sosMPZmaitBTxy9aqlmadkjs5NgB8KtoBNUvcfEIxR9hGpvYEPlRMILtvpOiQTMhIE7zUajJYZ9xuCIEsc/hh1e/INS2RO1SNBxk/tISOxPem+5478VceZGWhdpH8j7UPrHuaEqQBif2pxGuokSNV4e3NUNPpqYy9KT40lELxK+4ph1CesnVkH7x8IMw3ko6bEvZRr1fzaEvYhT2BwSGYeGzikCyUGEfxbiDCAlpEKpPPy292kdGDRDhhf/TgHi4hZ0f06qHI96H7TotIwpKhp5MTX3oqeMXEtX7/Cpq+nJS+HLGzXj4UiwxDPEdjryV45kFHQa2U34tE8jT9hvpNhq/sBcKQxzg4nVDCvYTgWVnHfBlJATokAhAsZ00FWqfCI71AEV3GrAo7F8Zv6m5wWIK23m6rfN+A2LoyVR2oaeOLxFfUr6g/aqGF6r+XUA0KeDsWs7HnYlGeh3GL1I2CHa+7NTZwde3vSicNlnKQYSDWu788vXHNk2IiVNr7LgGsc4iIKXBiEKIYt+bln7bOK8ZEENPprIPPU3SXy9NX8LDDEBRuUQlkjqMHX/a8DP97OuFqv+oFVwoCLG9ME4acHpdIk0AZ4OharD+65b3oB/D4qDR70BBWGIcD/iaJLGdp/vj+ndhCQw9mTL0tNTr/Hecq+7XmK8qD/x/u4rf/oldxeplu4ov/r+H+jcdjDw+jQmnlXq1YeCzj1+jlKZxKG3v9RIlphs79X7BSdJAGHoyZeg5DL0C0ADKY0evKP71mWuKncvWFFuWbSw2/uwni/e+ZW9ZOeCxcMj4XCOwUDsSvzL7ISTxKzRqXGqf55yShiYCzqBOHUhaNENPpgw9SxQHRk41cCBLC8EoDnJJ2f3M1cWuZauLPcevLh5ftbp48ukLh6cDy44qdixbV1z43M3F+1+9q/j0e28r9vxjVVMUp0uicLql6d4gac2TB9u5WD+xvvnclpIwTadS5xl6MmXoGT0qB5I8c7BQEfBnb9hbfPH/qYLUTHA6sHJ18f2nDb5G6WCJG6jF6bn6JaTTHo4IKCx7rA8vr5OyYOjJlKFnPOKWIQQdKhY409EX4WNmpPvesKG476TVxcPPOPFgjRFl87JNB8u6ZTuK9x69pfjkaZuKf1s3E5w2NNQ8He6puWjDVC9LqVEieKRBKy3slGptowaGeYvLwVkfC53KkjQ1DD2ZMvSMz6COsUwnbVbSlGeo0KFCg/ww5+wLB/6YAMGofglpv8v4u1Da3jitKayl7XeofpOUDUNPpgw904ljOGdquOq5HoI41h/2MZ4XEpLqJdox1WuU2oQmxkmDVlpYgHrbqKZpHG5h+rbBkbJj6MmUoScPEYIiMxB8qNyZeNRUtQktvcKapCwZejJl6MkPFSdR0TEVwUeSFsnQkylDT544CxXBh6vbJSknhp5MGXryRS1PBB8aOUtSLgw9mTL05I1mLdHQmQucbNMrKQeGnkwZekQb3wg+ca/CXiVuxbPQLXNoX8xw9inxminfv0iaIIaeTBl6BGp4uF1NnO5abIlQtNCzNhnOqTQDkKRxMvRkytCjQPCpX9FdL3ErnoVumUPNEcO5DU68ph6qIgDNu2GiJA2ZoSdThh4NQoSihZ46wXDuF1QPQJxWW/BRHJI0IIaeTBl6NC5NAYhTZAsFJ0laKkNPpgw96gIun08fl0HD516nvAhFPAyegBTjx6O1uOdQNJqmIbWnzSQ1MfRkytCjriCgpHeL5pQXp8wQQSceir7YEg2tCURMh0BkjZKUL0NPpgw96hqCTnrKq+lB6jSQpnYoanKiTVE0mmY4DanrD1utFx/DIeXJ0JMpQ4+6ivY+EVr4Sy3Q4TZ2JhDxWgJRPLQ9go+P4ZDyY+jJlKFHXUZNTpziGjQfwyHly9CTKUOPckbtT9QmGXykfBh6OuTd7353sWLFiuK4444rzjjjjKpvf+95z3uK5z3veTM772Vl2UyLzRYMPcpd+hgOnz8m5cHQ0xFvf/vbi1NOOaXYvXt3sWfPnmLNmjXFOeecUw1t9id/8ifF8ccfX1x55ZVl986dOw090iIYfKS8GHo64thjjy0+9KEPVV1FccUVV5Q1N4SgJvfee+/Mznp5cdVVV1V9FsfQI80i6MRVYwQfgpCk6WTo6YCHH364DDg333xz1WcW/bZt21Z1zfWJT3yiHH777bcXF1xwQVl6BaQmhh7pkDT48GwwH40hTSdDTwfceuutZYB59NFHqz6z6HfeeedVXXOdf/755XDa/6xfv75405veNLOzPqr4wAc+UI0x10033VS8/OUvP1hWrVpVji9pFsGH+/wQfChc0u7pLmm6GHo6oF/o2cJNSxrQn+EXXnhh1Yed9May3/79+6s+hzz00EPFpz/96YPlwx/+cHHMMcdUQyUFvnKEHoqnu6TpYujpgH6nt7Zv3151zUV/hu/bt6/qUxR33XVX2a/NaS5Pb0m9EXTS0108wkLS5DP0dASnqf7xH/+x6irKBsr9Agz9GZ42ZI7an//+7/+u+vRm6JH649RW+kywdes83SVNOkNPR3BqikvWaXtzxx13FC972cvmXLL+la98pTj11FOL+++/v+pTlMO5tJ1L1a+99tri5JNPLk4//fRqaH+GHqmd9EaG1PoQfuLhpW0QlBjX02TS+Bl6OoQbDZ5wwgllGKnfnPCWW24pXvCCF8w5nQXGY3xe1/YePTD0SO3xZPb0Yahpiae4E2wIQ/xPMKJ/r/FpJL2Y4CRpMAw9mTL0SItH+OHZXZz26hWC6oVaIh50euKJzcMp3hhRGg1DT6YMPdLSEVQ4/bVhw2ywIQzxRHf69XpgKv25QqwenKgdkjRchp5MGXqkbqD2KNoMeVNEabgMPZky9EjdEfcGWrHC01zSMBl6MmXokbqF02MEH06VSRoOQ0+mDD1St3BJe7Tv8fJ2aTgMPZky9EjdQyNoQg9Xc0kaPENPpgw9UvfQnicubScASRosQ0+mDD1SN3FJO6GHuz9zZZekwTH0ZMrQI3VXPPOLuzdLGhxDT6YMPVJ3cZor7t3DHaDrGM4jLHjkRVouu2y2fxQbREtzGXoyZeiRuo2wE6e5CDOEGmp+6KZ/2+IND6VDDD2ZMvRI3Rf37qmXeJ4X9/ShwXMUTovRnxKPuKBfzqgV42o4AiOP+khrxLwRZH4MPZky9EjdR0NmAszatbOhhkbObRs3x31/uMtzzqJheL9CGDIA5cHQkylDjzT94vL3nNv2EBZZB9R4ccqQbkJkvRaN/pp+hp5MGXqk6RdXgfFsr1wRcFgHvdo2pTVB3iJg+hl6MmXokaYfB3oO5tRq5IrTewsFmgiHnObSdDP0ZMrQI00/2qlELUaObVYIOiw7Db/7Yby4RQA1P5pehp5MGXqkPETblRwvXV9MTVe0/fGGkNPN0JMpQ4+Uh7Qhb25i2ds0UqYmLBp+N90QUtPB0JMpQ4+Uh5wvXV9sLVfcEJJ15SXs08nQkylDj5SPaK+S29VJcffqxSx3BKVBXMLO+3ITRI6x8agQGktzCo1gxd8o69cfGoeSU9sigjnraevW2WVnXcQ6GtRnEQw9mTL0SPnI8dL1to2Y6+IS9jZPuY/Hg2zceCi8xIF6EIXpDhs1WiwHd62O9407WLNsg7x7NeEmAiDhL33PfsXQszgzq0x1hh4pH3HahnvW5GIpl+tHSOzVDoog0Cbc0EaI96dw4KYwXwQrAhV/o8SNEyk8XiSmQfgYxqk2jvsEj3R+FyosM8u+GPVA1VR4ZArriPXN8rMuYh0NmqEnU4YeKR8cNOMAkwsOniwvfxeLg23TJez1sEOoYfrUoEV4GdTdr5lWzAPvudTpxikkao/qD60lDKeNtxmXcMayNd29eqHww/bGqap6MCTcMD2my/QHta4Ww9CTKUOPlJd4AGnbRr2TLg7Uh7u8EZriNE8aFAg7aUgYFsJDfG6Uhd6T8eP0UbSL6fVUfqZLWFtMbQrvH1e4Uerhh/cf17pqy9CTKUOPlJc4ZcLfHMSB93BPkXAATw/wFIJUWvMzKnG6jUKYCSwboYN+bU8hEeaWWsPSFH7qNUjjWlcLMfRkytAj5YUDEAcjDo7TjjDAsi62EXMdtURdOYATNNLTXfVTR1Ei2MQpt8MNfW3Uw0+8fxfDTjD0ZMrQI+UnDprDPBB2QRpWlooan66ghiYNGXyetJGJgDMuhB9qo7ocdoKhJ1OGHik/HCA5WHapjcUwRHsc/k4bQhif3zgaAU8DQ0+mDD1SfjhYEgYIP9OMGh6WM5dG22rP0JMpQ4+Un2jrQoPTabbURsyaXoaeTBl6pDxFm5BJaH9xOAbViFnTydCTKUOPlKe4dH0a27tgkI2YNX0MPZky9Eh5ilAwrZeuT3MjZi2doSdThh4pX4QCSpcuxx4UGzGrH0NPh9x+++3FJZdcUlx00UXF9ddfX/VtZ+fOncUuHqzSkqFHytc0X7puI2b1Y+jpiB0zP0sIIevXry/OOeecmS/tsmL79u3V0P54LeNT2jL0SPniZnbsLnjK9jTV9tiIWQsx9HTEmjVris08qa3C/6tWraq6ejsws8c6auanzdqZn26GHkltEA7SRxpMy43ubMSshRh6OuDJJ58sAws1NmHvzF6JfldffXXVp9lZZ51VbNq0qSyGHkltEXTSJ3gnv7kmlo2YtRBDTwfcfffdZWDZs2dP1WcW/Wjf0wvteFay15qxUOjZv39/8aUvfelg2bZtW3HsscdWQyXlKi5hp3BF11LbwhCmaF7YqwyzVslGzFqIoacDCC9NgYV+6SmvFKe1VqxYUb4WC4WeL3zhC8Xzn//8g+Wkk04qT4tJEruRuGkhu4WtW6sBLRBiGJ/2QdGIeKEyrFBiI2YtxNDTAf1qej7ykY9UXXNtmPl5Rgme3pK0FDRo5knZ7EYo1PqsWTMbZvjtVS+9Qg7hiRqXphKn02hHNGg2YlYbhp4OiDY911xzTdWnKPbt21f2u/baa6s+c62e2YMwvKlE7U8/hh5JTaiFiUbObQohh7DE5e9taliiRmnQ7W5sxKw2DD0dcfrpp8/8epr5+VSpX7316KOPFp/5zGeK733ve1WfuazpkTQo1Prw24lCmCCg1EvbkFPHNNlVUUs0yMvlmadhhClNF0NPR1x11VXFcccdV5xxxhnFmWeeWQaY9D49N9xwQ9nvnnvuqfrMZeiRNCni5ojUEA0KNTxM00bM6sfQ0yFf+9rXigsuuKA4//zzi+uuu67qO+uBBx4oLrvssuKJJ56o+szFJe5tTmsFQ4+kcYn2N5RBXc1lI2a1YejJlKFH0jjF6SgaSy9VhCgbMWshhp5MGXokjRPteaLB9FKfAWYjZrVl6MmUoUfSuBF2CCtcwr7YRs3c6LB+6byNmLUQQ0+mDD2SuiAaIC8UWGiyuHHj7Okwxq8XLoUf5t2eNR0MPZky9EjqgvQS9nojZLqpzaEmqB5yuNEh92fl1JaNl9WWoSdThh5JXRF3guZUFTgm8X8acqjJIeQQkgZ5fx/lxdCTKUOPpK6gpiYaNddrdQhE3ntHg2LoyZShR1KXxCXsFGp1tmyxRkeDZ+jJlKFHUtcQdBZxj1Vp0Qw9mTL0SJJyY+jJlKFHkpQbQ0+mDD2SpNwYejJl6JEk5cbQkylDjyQpN4aeTBl6JEm5MfRkytAjScqNoSdThh5JUm4MPZky9EiScmPoyZShR5KUG0NPpgw9kqTcGHoy9bnPfa74sR/7seLnfu7nWpcTTjih+Omf/unGYZZDhXXkelq4uJ7aFdbRz/zMzzQOs8yWn/3Zn3VbalH8zv1cccQRRxTvfe97qyPhdDL0NNi/f3/x0Y9+tLjllltal3/4h38oTj755MZhlkPlT//0T4t169Y1DrMcKr/xG79R/NVf/VXjMMuh8uxnP7vYtm1b4zDLbLn66quLpz/96Y3DLIfKBz7wgWLVqlWNw3Ip//Iv/1Ls3bu3OhJOJ0PPgOzatas49dRTqy71wo7lj//4j6su9fKGN7yhuPjii6su9UIN69e+9rWqS02+9a1vFc94xjOqLvVy5ZVXlj82NN0MPQNi6GnH0NOOoacdQ8/CDD3tGHryYOgZEENPO4aedgw97Rh6FmboacfQkwdDz4AQev75n/+56lIvrCPX08JcT+2wjgw9/RF63JYWRuhxPU0/Q48kScqCoUeSJGXB0CNJkrJg6JEkSVkw9AzA9u3bi3POOac4/fTTi7/8y7+s+k4PblbFrck3b95cliYHDhwo/uIv/qJYs2ZNsWHDhuKOO+6ohhzyT//0T8X69euLN77xjY3T4TW8tt80eB2v7zWNcfriF79YvOtd7yrn/7zzzivuuuuuasghX/nKV4qzzz67eM1rXlO8+93vrvrOxXL99m//dnHmmWeW21bdl7/85fImj/22N6bBTSDZLpumMS4PPPBA8f73v79cR/3WU3ynWIZenzPrj/XI+mS91jGNM844o3j961/fOI3/+Z//Ofh5ve1tb2vc3rrgtttuK+d/586dVZ9D6N/v+5B+p/h+8j2ti2nw3eQ7Wsc0WD9Mg/XFeusK5r2p1PE94fvS6/sQ21u/71TbafTbZjV+hp4luuaaa4ply5aVG/lVV11VfinY0U6T1atXFyeeeGKxcuXKclnrHn300eK0004rl5sdM+vilFNOKfbt21eNMbtz4rUXXXRRccUVVxTHHXfcnB0D0+A1S5nGODHPK1asKD74wQ+WV/IxX0cffXRx4403VmMUxWc+85niyCOPLIexrbCs9W2Fgw/LvWPHjuL8888vl5dxA9vbM5/5zIPTaNrefv/3f7/vNMaJ9cQybtmypZy/d7zjHfO2KeaVfsw747AsLFMq1h3jsi5Yr6zfENOIbYVppNsKB3/uvss00u3t/vvvr8boDr53y5cvLzZt2lT1mcU88x1g+VhOljddRr47sdwsI8vK95TvWjicabDemsLTODBfhDH+piXFPPM9iW2FZeR7FJq2N16TOpxp1LdZdYOhZ4n+6I/+qPiDP/iDqqsobr755nLj50swbdjpsWx17DCPOeaYqmtW7CgD3dyjJ7BzYGcboaZpGjzWoz4NXhfq0xgn1k3dq171qvIXcvizP/uz4rWvfW3VNYtljm2FWhDWbxqUWGfsbAPbWzoNajjS7Y3Lk+vTYPtMp9ElDz/8cHHssceWB9zAvFLrEFgWlollA8ta31ZYJ6zfwDTS7Y1psK1EqOHA1DSN+gFz3AiHa9euLX94pKGHbZ7lSb8PLC/fkcCy8B1Kscx81xDTuOCCC8puNE2jaZtl/XUB80fo6SXCCPvlwPeB71FgW2nah6fbW30arJN+06hvs+oOQ88S/dIv/dK8A15UA0+bXqEnThGk2Bm96EUvKv9/7LHHytfV1xP9uDcG2k6jLp1G13B6Kj14s62wTCmWOQ7Wn/rUp+YtY6zz+HXeNI1XvOIVB7e3ftN45JFHqj7dwrxdfvnl5f/MI93pr2jQj2UD66tpW2HdgHXF+E3bWzxButf29uIXv7jqGj9OK1PDSq1KPfSwzbM8qfic+a6A707T9hbbCtPgmVwEz1CfBuuj3zTGjXljfpjvPXv2VH0PafqcGTe2ldje6tsKr4ntrde2stA06BfTUHfMP4poUfiV2vSF4Y6604bl5ItcR9Vv+qsHl156afGc5zyn/J82Abyu/quHfh/60IfK/5umwY6lPo26dBpdwrriicVpDQbbSvzKDixzVIPzi7v+yzxqbqK9SdM03vKWtxzc3pqmEZ9bl27ix2f7zne+s3yy81lnnXXwIYfMI/Naf+ghyxQ1Eqyv+rbCOmHdoNf2xjQuvPDC8v+m7Y1pPPe5z626xo+gw/co/k9DD9s8y5iqbyt8d+qBhWWOUzdM4+d//ufL/0N9GqyPpm02pjFuLB+nldnnMt+cerv22murobN3Nm8KPbGtxPbWtA+P7a1pGun21msa6Tar7jD0LNHTnva0xi/Mq1/96qpresTBs47TODSSTNG+gl+RuOmmm8rXfe973yu7A/1o1IqmabBDq0+jLp1GV/DLnLYTNGpMsa2kO2SwzLGtsBz1mgbWGctI42U0TeM973lP32nE5xbT6AI+WxrYvvCFLyz+8A//8GDIYR6Z1+9+97tld2CZ4nNmWevbCuuEdYOYRn17S6fRtL0xjdjexo3TWgSdUA89LAfLmIpthe8KWJZ66GGZWXYwjZe85CXl/6FpGk3bbExj3KgRfOKJJ8r/d+/eXe57CT6BbaUp9NS3laZ9eLq91afRtL3Vp5Fub+oOQ88SnXTSSY1fmLe+9a1V1/SIg2cdV89whUyKX5G/8Au/UP5/3333la+7/fbby+5Av/gV2TQNdtj1adSl0+gKfgU3HRTYVtI2GGCZY1v56Ec/WjZ+TrHOWMZ777237G6aBjUf/aYRn1tMo2t+8Rd/8eDBgXlkXr/61a+W3YFlYtnAsta3FdYJ6wYxjfr2xjT6bW9MI7a3cSIAHnXUUcVll11WNoqnEKI5Vcr/YDlYxlRsK3xXwLLUQw/LzLKDadRrtpqm0bTNxjS6Jrb1O++8s+xmW2kKPfVtpWkfnm5v9Wk0bW/1aaTbrLrD0LNEfBnqVZhUe9Z3NtMgdih1LGva+BEc+NPGs7wuvcyTX2X0+/znP192t50Grwv1aXQB89zrs2db4VRU6nnPe97B8WP9pqd2OL1BY9Onnnqq7GYab3/728v/wa9cqvfr04iDFtg+02l0DQdzLvMF88i8nnvuuWU3IvDGQYVlZb2lWK9xYIppxKkhsE7TbaXXNNLtbVy4RJ2anbRw9Rbte/gfLAfLk34f+H7RL7As9dNQfMdiW4lppAfrpmn022a7JvYJcXqO+YzTUIHvQ31badqHp9tbfRp8B/tNo77NqjsMPUtEGwF2JPEl+5u/+ZvyC9CFK4oGLQ6odeyk6R+hJsaLK4rAgS3dAdPINz3ADGIa48a89WvLVd9WWNb6tkIbi/SAUg9Rbba3+jR4+n/aPU7MR8w7+Nzr88v/zHOgO217whVYxx9//MFthemxTqK9DnjN7/3e71Vds939trevf/3r5TTS7a1LCDv1S9ZZHr4DgW0lbTgfVx3FgZdlpZtlD22mwSlI1g+apjFO8fmBBuxvetObyh8BcWozrlDje4Je20q/7W0Q01B3GHoGgB3Fs571rHKjT3ek04KdLctVL6kPf/jDZT9+/cR9ZFJUAf/O7/xOccIJJ5TVwumBO/CaftOInU2/aYxLhLR6iV+DgStBOHVBuwOG17eV66+/vtxZcjqDHS3bVh3T6Le9xTRYPzRmbZrGuMRnzLwxj/zfNH/0Y95jPJYpFQdf1iPrs+lqIqbBATDWRX1bSbfZuH9SVzWFnvg+8F3gO9G0jCwT3yWWkWVlmVP1afAd5buaYhqsn17TGCfmh1qYmDe2h/oVU7Gt8H3he7PQ9sa49e/UYqfRtM2qGww9A8LlktzHIb3x17Tg1AAH9Xqp49JX+tevmkmxk6XdQL2RaeC1/abB63h9v2mMQ7pe6qWO9dlvW+GUFfffaboENwxiGuPCPPVbP4HxWIZoqFrHsrMO0tOBdQttb2222S6gZqVpOdt8H+I7lV6anopp1ANTaqFpjEu6LfXbDmJb6fd9YFi/71TbafTbZjV+hh5JkpQFQ48kScqCoUeSJGXB0CNJkrJg6JEkSVkw9EiSpCwYeiRJUhYMPdKE4r4kPIupfn8SHngaz2gaFu6LUr/x4jjxLKQ3v/nN5Twxb5LUxNAjTai4U3Y9fHDQp/8wjeI92vr+979fzst5551XztdiQ0/T3Y4lTSdDjzShOFCvXLmyfBhleqDPLfQwLzyKgIdALjbwwNAj5cPQI00oDtRxwOZZXaEeSJoO6mm/GP/KK688+EywX/3VXy2H/fVf/3X5HCGeA/b+97+/7Id4Dc8kInjx/+tf//p5j3PgPXg2FsP5u2PHjmrIofnfsGFDOZz/m3C6jodgMg6Fmq144GU8yysd1oQnrsd8UHgmF5iH9PWUwGsYj368duvWrdWQQ8vPQyh5bhXPY6ov/5YtW+a851lnnVUNkTQuhh5pQkVoIBRQ28NBGocbeggQ/M8zmF75yleW3Zwyos3Q5z73ueKII44odu/ePec1v/zLv1w+WJFuAsdv/dZvlcPB9AlE0eYoXhPdETjq81ZHWGA6LGd0R2hBvHc/vE8auNKnhDetH8ZlncZ4/KU7phHLwvvyf8xDPIiS8Rmevk/6v6TxMPRIE4oDddSO8H8EgTggh7ahh4cphne+853F0UcfPechljxh+qKLLir/j9d8/OMfL7tx7bXXlv2uuuqqspv/06CB9H35e+KJJ5b/91OfDuGHfswDInD00zQvodf6oaYmxThr164t/4/l/9jHPlZ2Y9u2bWW//fv3Hww9MY+SusHQI00oDsIcnEEQIEBQ2xMH5NDroB796uODWp56kKCb/ojXcIBPnXzyyQdrhxjeVOJ90/nvpWneQM1POv8LhR7WCzU1nG5at25d+ZrQtH4Ytz7flJjfpuXnKdz0i9qwOG1HGOX0nDU90vgZeqQJVQ8N0QaFGg0OtoHaifpBfVCh59Zbby27EQf9yy+/vOzm/34H+vr8N4nwFKfEQv1U00KhJzBuhJGYt6bQQ6iK04VNmpafmjL6ffOb36z6zGJcPgMCV305JI2WoUeaUE2hgdoeAgAH38B4aSiIIDGI0PO3f/u3ZTe4V84znvGM4qtf/WrZTXAgYPTSJvSAZUpDSbx3BAi624aeELVioI1QfT7p12+aTctPDdcpp5xSdc3H+LxO0vgYeqQJ1RQaOJBzcKWECDmcYiG0cMBPg0QcwFNtQw8Nd+lPoTuGg5oUamQ4nUR/CleZxYG/beiJmquNGzeW06DGpB6CFgoo1IDFPDA/zFeEJtZZnPZiODhdSGhjftPX1dcZ89+0/AxPX8twpidpvAw90oTiwNp0CoYDcxoKQACJ/hzseR2vB3/jYB3iYJ1K+6WvueKKK8r/achcR3jgveK908bEvea/CfMc04j5Dk3zX8f7xutpoBxXggXWT4yTSl/H//E63pOQc9dddxUXX3xxce655xaf/exny2FgvPS1LGf9PSWNnqFHkhYpQo+kyeK3VpIWydAjTSa/tZK0SIQeiqTJYuiRJElZMPRIkqQsGHokSVIWDD2SJCkLhh5JkpQFQ48kScqCoUeSJGXB0CNJkrJg6JEkSVkw9EiSpCwYeiRJUhYMPZIkKQuGHkmSlAVDjyRJyoKhR5IkZcHQI0mSsmDokSRJWTD0SJKkLBh6JElSFgw9kiQpC4YeSZKUBUOPJEnKgqFHkiRlwdAjSZKyYOiRJElZMPRIkqQsGHokSVIWDD2SJCkDRfH/A6MbJSagde/sAAAAAElFTkSuQmCC">
La loss represente des "sauts" à cause de la reprise de l'entrainement à deux reprises. Cela induit une modification du learning rate et explique la forme de la courbe.
## Résultats
Les questions générées sont évaluées sur les métrique BLEU et ROUGE. Ce sont des métriques approximative pour la génération de texte.
<img src="data:image/png;base64,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">
<img src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAArMAAAGJCAYAAACZ7rtNAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsMAAA7DAcdvqGQAAFhNSURBVHhe7d0JuNXUvf5x732ePm2tA7Xyby/iLS2I9aJUxRG1Uue5Tljn4lCqiHVGbK2ibVUUoYJoqVBAEAoKiMiMAjIjg+gBmZEZmQdBHH//864kh7BP9jkBzj575+T78VkPOyvZ2Ul2jnn3ykqynwEAAAAJRZgFAABAYhFmAQAAkFiEWQAAACQWYRYAAACJRZgFAABAYhFmAQAAkFiEWQAAACQWYRYAAACJRZgFAABAYhFmAQAAkFiEWaTS22+/bcOGDfOHKkarVq1cKWR9+/ZNxHLm07Zt26xr1645306F8D0UwjJUhNGjR7v10L8A0ocwi7zbb7/9spZcHZzOPvtsu/TSS/2h+Mo6aGp5CzkYaPkOO+wwa9SokSvZaFz4O6hVq5ZddtllNn78eH+K0u6991479thjrVq1aq7oteoyabtpntm+V42L2oa9e/e2M844w2rXrm3f/e537Sc/+YmdfPLJdvfdd9uSJUv8qTzhZc8s5e1PCxYscNPVqVPHbYdcfp+5nn8gyftsXOXtVwCqNsIs8i44oEaVXB2c9iXMZjtoBstciAYOHOiWe9KkSX5NdkHYDdancePGVrNmTff+yZMn+1N5Vq5caQ0aNCj5Dvv16+eKXqtO4zRNoLzQEcwnLJhXvXr1rEOHDjZ06FB74403rF27dtawYUO3rGHBPKJKeftTixYtXHivDJUZZrNt86jtnURatzjfL4CqiTCLvMvFAXXjxo02bdo0mzdvnn3++ed+7S65CLPZbN682T788ENX9Lo8mveKFSv8oXjK+wxtXy13HEGYDevUqZN7/x133OHXeIL59urVy6/ZRXWZ32152y9z+oULF7q6zOUJfPnll3bffff5Q57MeeyJqHUPaD/S/qT9SvtXeWbOnGlFRUX+UGn6nMxts3btWn+oNI3bk30j2MZlbfPwttq0aZNNnTrVPvvsMzccZc2aNW6abH9X+6q8v9tMUesEIH0Is8i7OOHjlFNOsfPPP98f2kVhQe9//fXX/ZpdASsoOjWt1rywzDAbvCdTuD4IBZklOKDqdeZ6qPUwc3rVhQUBSuFPrYLBdOVtk0B5n6F5Z44va97B8mTS+8L16ltao0aNMn8UaJym0bRSVrCSzGW74oor7LjjjrPt27f7NeUrb/2y0fsyS7Cc2n+0H4XHZX6GhlWvfVHdFKKmCdO21PjHH3/cDjnkkJL5vvTSS/4UHk2jVulgvIq6V2RSvaZVCeYXtW+oBPRa06v1PRinZR8yZIg/xS7XXnttyTQqWqZwWI/az4KS7fsO03KE3xP1dxvsm+F11LzL268AVG2EWeSdDkI6OJXl6aefdtNt2bLFr/Hcdttt9tOf/tQfMhs5cmTJ/DZs2OAOtjpQK1CFT3fvbZgNHzTDRYLPDQwaNMjV6fPVF1MlCA0aFwgO0AoLOnirD2iTJk3cdCNGjPCnihbnM7R8wXpkLnOUYHnCdOGY3t+2bVu/xmzChAmu7vnnn/drStM4TaNpRZ8bLEcUjQtvw1/84hduffZE5jzi0jIF6x5sIxXtN9p/tBzan7RfBdtT+1sgqKtfv35JIM22nqLP0T5Y1ncnmm/nzp1dK7U+u2fPnm6azO9IdZrf1VdfbRMnTnTLqjCrZdC48DoFVK/P1GfMnj3bhdigv3BYsG76V9tD+6n213PPPdefYtffR7hcddVVbn4fffSRP1W0uH+3wfdzwQUXuK4z+lsJPitYRwDpQ5hF3ukglK0Eli9f7oZ1sAv7zne+Yw8//LB7vXPnTqtbt64LE2E63Zv53r0Js1LWQTPzMzR/1c2ZM8evMfdadeHP1sFZdeoHGtApZdXdc889fk20uJ+Rbf2iBIFB71E577zz7Mc//rGdc8457vRv4OWXX3bzVB/ZbDRO02haKS90aFx4Gx544IFZLyQLlzDNI1spT7DuYcG2034UplZ07W/a7ySYTmEzjuB7L++7ixJ8lk77BzSsv4fMrgraPhqXuZ1E9SeeeKI/5FG3DdUHITIImnfeeacbDgQ/pLKt75/+9Cf73ve+Z+PGjfNrou3J322wzRYvXuzXeMpaRwBVX7yjG5BDwQFLB6LMEqZQdcQRR/hD5lqr9N4gZAT9K++66y43HKYr7MMhpTLC7M9//nN3AVQm1WlcQMsVHg6oPrzMUeJ+Rrb1ixJ8blD0vurVq5da52D7q4Usm+DCM00rZW0/0bjwNjzooINK9YkVTRcuYcE89BmZpTzBOodpWPtPposuush9lvY7Cbbx3Llz3XB5NN9s351aM8PUMt68eXPXIhksoz5L/VcDGj7rrLP8oV203hoXtf6qb9asmT/kyZw++J61fpnl9NNPd2cRMnXr1s29J9zCrFCcWaS8v9ubb77ZH/K22eGHH+4P7VLWOgKo+gizyLvgQFme7t27u2mDU+86xRkOA2Ud0IKDf6AywmzmcCBznkE4yZStPizuZ2QOlyXzc3UqV6FF7w+3BCpIqe6FF17wa0rTOE0ThK5g+0X1yVS/Wo0Lr4/6Zep0czZR65U5jz0Rtc01v6jvIfjsYF/Yk20smmfUcmbO59Zbb3XDaqHW/W/1eUE3lPB+qOGo+QXbPDxtIOo9mdNrvH4YBdsmqoSNGjXKvb9Hjx5+jadPnz6uPig/+9nPXH1Zy5c5/6jPk7LmAaDqi/9/XiBHdBCKOghH0WnU3//+9zZ//nz3vn/961/+GLOlS5e6ugceeMCv2UWnq8OhKDPMPvvss+69madoFZhVHyjroKn68HooiEWdLlYA17hAtgN0tvowzSdb6174M7Rc4fUoS9Tn6h6zer/6QAZ01btu2VXWKXGN0zTBFfKLFi1y83nuuefccFjQB1c/WgJ6v05Vhy80CotaLw3H3Z8yRa279hvtP5m0bPos7XeyJ9tY9DnlfXfLli1z88zsbqILwFQf3g+zrfee7LOSOf1rr73mhsPdYLLR96RWfPVxz/TVV1+VKlLe3224e0PU9yNlrSOAqo8wi7yLOqBmo1OOOvX497//3b1v69at/hiPgoD634XpYhhNq9tLBTLDbNC3MzgdLjNmzLD999/f1QeCMBZ1KyrVh9ejadOm7gKWsClTprjpNC6Q7QCdrT5M89H8NN9A1GfsSdDK9rna9ppH+MlpuiBMdXqoQSbVaVz4ojH1j1TYiZr/Qw895KZXy15A97VVXTiYh0Wtl4bj7k+ZotY9uC2Z9qMwfbfhMLon21j0OZo+qs9s8N0F+1vLli3dcOCkk05y9eHwpuGo9Q6CXpx9VjKDoW6VpWF1q4gS9KPWhVunnXZaqdu3xRH37zbq+5HMZQaQLoRZ5F1wQM1WwqZPn+6m1xOm1Ic206xZs9x4taapn6Fush/VepkZZoMr1oNl6dixo5166qnuterC1J9RfRfV11DjgwNo8N5AEAJ0gNcFUCpaFtVpXCDbATpbfVjwGZpvWZ8RtR7ZZPvcDz74wM3jN7/5jV/jCVoJ9RkKTCrB50XdQioIwApsCioDBgwomT7q9mvBOC2Trs4fPHiwe19woZJKmIb1nmylLNnWPWgt1f6k/Ur7lz5H+1tA885clrLoc/QePfhBF1Gp6LXmEf7u1BquaV955RW33hdeeKFdd911brpweAvWO5O6iWhceJ8NRL0nKhiqJV11wXem70BnM7RcwXRaTn2G5pdZwvOKEvfvNtv3E7XMANKDMIu8Cw5QUUUHwkyqP/PMM+3NN9/0a3anU50KBQcccIC7WEQH2XBfT8kMs6IDYfA+fYYOqvp8vQ7TRS1qRVRLlcYFB9Co5dWV3Aqzhx56qCt6nXl1t96X+RmSrT5TnM+IWo9syvpctc5qXLg1UV599VUXRNSlQEWvVZeNlk/9cPUDQq3fupdsly5d/LGl6UEBOg2tz9apZ91dQVfhq4U+8yb/wfJHlaj9KSyYLpP2H+1H2p+0f2g/yez6sCfbWILl0X6mgKowph9Qmd+dTu+rj+yPfvQj9whfvUf7nN4fDm/B/KLoB0Z4nw1EvSdq3hLez3SPVy2LuhOsXr3ajdd7spXMeUWJ83cbzC9TtmUGkA6EWQAAACQWYRYAAACJRZgFAABAYiU2zLZo0cIuv/xyd2W7+prFoT5Z6qenPne6cjZb/zIAAAAkQ2LDrDr7BxdCxA2zumBEF6boynXd8kehNnzbFwAAACRL4rsZxA2zwa1fws9X122DMm/9AgAAgORITZjV7W++//3v+0Oe4L1qqQUAAEDypCbM6kbt9evX94c8wXvDNygP6NGauudjUI4++mgXhsN1FAqFQqFU5XLQQQfZiy++6B8ZgcJEmC1+rx5bmunTTz91T7kJygsvvOD+qMN1FAqFQqFU5aKGnG7duvlHRqAwpSbMltXNIHiCTVk+/vhj99QhAADS4le/+lXWpy0ChSI1YVZhVNOFH0GpuyHEvQCMMAsASBvCLJIgsWFWITYoCqnB68CUKVOsXr16u13cdcopp7hbcy1ZssT1idUzwOPemoswCwBIG8IskiCxYVatqrrXbGYJTJ061Y455hhbtWqVX+M9NKFhw4YuxNauXXuPHppAmAUApA1hFkmQ+G4GlYUwCwBIm3yH2e3bt1MqoXzxxRf+Fk8mwmxMhFkAQNrkI8x+8803rovgvHnzbM6cOZRKKuqCuXHjRv9bSBbCbEyEWQBA2uQjzCpQ6Zi7YcMG+/LLL+2rr76i5Lhs27bNdctcuHCh/y0kC2E2JsIsACBt8hFmly5dypM582DHjh2uhVb/Jg1hNibCLAAgbfIRZufPn+9aZVH55s6da5s3b/aHkoMwGxNhFgCQNoTZdCHMVnGEWQBA2hBm04UwW8URZgEAaUOYTbZevXrZQw89ZBdddNFu9+LPhjBbxRFmAQBpQ5hNNgVYPSAqeFpqeRRmt2zZ4g8lB2E2JsIsACBtCLPlU1BUYNRTRrt06bLb00UHDRpkbdq0sd69e5dap8ynkAbzCZs1a5Z17NjRFc1f4zOnGT58uLVv395NM23aNL92d4RZOIRZAEDaEGbLp6CoFtB69eqVtITKtdde6wJkkyZNrFatWlajRg0bPHiwGyeZ4TIzcGraQw45xL1X89Aj+sPzF73WNDfccIPdcsstVqdOHRdsMxFm4RBmAQBpUyhh9oqOEyq1XP3yRP+Ty6egqCD74osv+jXmWmgVHidPnuzXmF111VUulAbKC7OaVu8JTJo0yY4++uiSMDtjxgzbf//93fsC+lyF5sx7xe5JmKXPbBVGmAUApE2hhNkT/jbSfvrw25VW9jTMHnjggf6Q56abbrKTTz7ZH/K0bt3aateu7Q+VH2br1q3r3hNWvXr1kjDbuXNn11L7+OOP71YOPfRQ1yUhjDALhzALAEibQgmzH67YbNOXbqy0Mm/NVv+Ty6egqFAZdumll1qzZs38Ic/AgQN3C73lhVlN269fP3/Ic/zxx5eEWfXFPfvss+2JJ55wRfVB0bzCCLNwCLMAgLShz2z5osLsww8/7FpRP//8c7/GrGXLlnb66af7Q+b6woZD59NPP71b4NS0zZs394fMVq1a5cYHYVYXl2l4zJgxbrgshFk4hFkAQNoQZssXFWYVCnW6X/1elyxZYl27dnXh9dlnn/Wn8LoiKJiuXLnSzeOSSy7ZLXB26tTJvUfvnTdvnvsMlSDMisZrONw3t2/fvv4rb9mConmHh6MQZqs4wiwAIG0Is+VTMMwMs6L6hg0b2gEHHOAuEGvXrp0/xqPbaDVt2tSFzFNPPdVNn9l6qqBas2ZN9/6zzjrL3dFAITesRYsWri+u3qvSuHFjf4z3/qA+XAizKUWYBQCkDWE2fxYtWuS/8ixYsMAF0ZEjR/o1FY8wW8URZgEAaUOYzR+1nqqVtU+fPq4/rfrgnnjiif7Y3CDMVnGEWQBA2hBm80dhVg9AUL/bK6+8cre+srlCmK3iCLMAgLQhzKYLYbaKI8wCANKGMJsuhNkqjjALAEgbwmy6EGarOMIsACBtCLPpQpit4gizAIC0yUeY1S2oCLP5QZit4gizAIC0IcymC2G2iiPMAgDShjCbfMOGDXOP0R0+fLhfkx1htoojzAIA0oYwm2x6YtgPfvAD97AFvf75z39uI0aM8MeWRpit4gizAIC04QKwZOvZs2fJtly5cqXVq1fPGjZs6IajEGarOMIsACBtCLPl05O6GjVq5J7QVbt2bdcCKmoBPf300+3AAw+0Y489ttQTvPSesGA+YS1atLCaNWva4Ycf7t4ffE5gxYoV7ulgyiea7uabb/bHRNNnVKtWzR8qjTBbxRFmAQBpUzBhdvWHZiunV15ZN8//4PIFIVTB9d1333V1CoWHHnqoexTtkiVLrGvXrlarVi3XdzUQhN6A5hOu07R6j967ePFiu/7660uF2XPPPdcef/xxKyoqstmzZ9sTTzzhwm02eu+ll17qD5VGmK3iCLMAgLQpmDD73BFmjx9UeaXL+f4Hly8Iofo38OCDD9ohhxxiO3bs8GvMmjdvvtsp/vLCrMKx3hNYtWqVGx+EWX0vGlZ9YOjQoa5OXQoy6X01atSwyZMn+zWlEWarOMIsACBtCibMdmpUuWUPw2xm94ALL7zQbr/9dn/I069fPzvggAP8ofLDrLon6D1hxx9/fEmYfeGFF+y4446zE044wU4++WQXlM844wyrW7fubsFadCcDzXvatGl+TTTCbBVHmAUApA19ZssXFWbVveC8887zhzydOnVy3QYCmWF24MCBu9UplLZu3dof8lSvXr0kzGbOL5u+ffu6+erf8hBmqzjCLAAgbQiz5YsKswqcCpCLFi3ya8xuu+02u+iii/wh7wIwBdKAxofDrALxVVdd5Q+ZTZw40Y466qiSMDty5Eg3vfrUZhMsR5wgK4TZKo4wCwBIG8Js+aLC7Pbt210XgOACLgXTzOCpfrV16tSxu+++212U9eSTT+4WZgcPHuz63Woeer+CrLoRBGFWzj//fHe7LdUNGDDA/asuBxJ0W9CyZZZsCLNVHGEWAJA2hNnyKTSGA2ZA66DAqv6zanXVdJmCIKwW2qj5zJo1yzp27OjKRx995C7geu211/yxHt1L9qabbnKtvnq/tp8E84sq2RBmqzjCLAAgbQiz+aMwGt4OGlZLa3kXce0LhdktW7b4Q8lBmI2JMAsASBvCbP4E4VUXgulhDEGXhVwizFZxhFkAQNoQZvNL96lVqFV3g/A9a3OFMFvFEWYBAGlDmE0XwmwVR5gFAKQNYTZdCLNVHGEWAJA2hNl0IcxWcYRZAEDaEGbThTBbxRFmAQBpQ5hNF8JsFUeYBQCkDWE2XQizVRxhFgCQNoTZZNPTxcp64lcmwmwVR5gFAKQNYTbZ9ibM8jjbSqZnHuuJGHpWcePGjf3a7PSFnn/++XbQQQdZvXr13LOS4yLMAgDShjBbPj3UQEWGDBmyW3gcNGiQtWnTxnr37l1qnYL3BMLzCehhCR07drS+ffu64ahphg8fbu3bt3fTafowwmyBa9asmQukekbxggUL3BfWtGlTf2y06tWruy915cqVbgfTY+LifsmEWQBA2hBmy6dwqQxy+umnuwY2vZZrr73W5YwmTZqUNLwNHjzYjRONC9N8wnWa9pBDDnHvVYPdhRdeWCqc6rXec/LJJ7txmn7gwIH+WMJswVMw7dChgz9k1rNnT/eFKtxGydxJJNgJ4iDMAgDSplDC7CPjHqn0Epfyhc74tm3b1q8x69Kli8sXkydP9mvMrrjiChdsA5n5IzOnaNqrrrrKHzKbOHGi/fCHPywJpzNmzLD999/funbt6oZFLbSnnXaaP0SYLWhr164ttZOI6nr16uUPlaZfLsHONm/ePNeye/fdd7vh8hBmAQBpUyhhtlGfRnZ0t6Mrrdw89Gb/k8unEPqd73zHlixZ4teY3XrrrXb00Uf7Qx6FyiOPPNIfKj/M1q1b11q3bu0PeX70ox+VhNNu3bq5XKL3hYvmoVAqhNkCNn36dPdlZe7sqlPflGwWL15sN9xwg9vBfvCDH5TaScLGjx9vJ554Ykk55phjrFq1av5YAACqvkIJsyM+GWFvLXyr0sqElRP8Ty6fAqRCY1hUiAyCZiD8WjLHH3jggdavXz9/yHP88ceXzFf/nnXWWa7o88JF85Ko5SgLYbYSzZw5M2uYVRN7lE2bNrn+LOprO2DAAHv++efL/JLXrVvnOlUHpXPnzq5rAwAAaUGf2fJFhdl77rnHnf0Ne/rpp61Bgwb+kLkGsqKiIn/IXGNcOMwqszRv3twfMlu1apUbH+QWXRT2/e9/3zXwZUOYLWAKpvpCo7oZ9O/f3x/anQKswqi6KARatmy5245TFroZAADSJh9hVt0Akx5mR4wY4fJF0J9VXRAuuOACe+ihh9ywXH755RZ0fdSZY80jnEmeffZZd/GX5rFo0SK75ppr7IwzzigJp7ofrM4yazjo4qDtpmuIAoTZAqdfPGotDejuBAcffLAtXbrUr9mdOmPrSsIwfcHacYK+JWUhzAIA0oYwW76oMCs6U6yMoUCqf3VBV7hfrTKIcskRRxzhxqulNRxmpUWLFlazZk07/PDD7dFHH7UTTjhht+6UupuT7nKg96nVV/8ee+yx/lgvzKouswTdEDIRZivZfffd5wKtru5TM33Dhg3tzjvv9Meaa7VVR2vdhisY1hf40ksvudbZMWPG2CWXXBK5A0YhzAIA0oYwu2+0Hsof4RCbScEyfNY4G7XeKseo8S6TQm22uzntCcJsHqibgH6xKGRmPjTh/ffft+OOO871MQnoyr+LL77Y3UKjdu3a7r60cVplhTALAEgbwmz+KOSqi4Ea33r06GG33367u4BdXS1zhTBbxRFmAQBpwwVg+aMwe+WVV7ozyPpX3RIyrxWqaITZKo4wCwBIG8JsuhBmqzjCLAAgbQiz6UKYreIIswCAtMlXmF2/fr0/hMryzTffuKxDmK3CCLMAgLTJR5hdvXp1mVf/Izd039o5c+bYV1995dckB2E2JsIsACBt8hFmP/vsMxeqFi5caJ9++imlEsqyZcvcNte/SUSYjYkwCwBIm3yEWdmxY4cLWWqhVdE9Viu6BPPe0xI1r8oquVqW5cuXu5ZZdTVIIsJsTIRZAEDa5CvMAnuCMBsTYRYAkDaEWSQBYTYmwiwAIG0Is0gCwmxMhFkAQNoQZpEEhNmYCLMAgLQhzCIJCLMxEWYBAGlDmEUSEGZjIswCANKGMIskIMzGRJgFAKQNYRZJQJiNiTALAEgbwiySgDAbE2EWAJA2hFkkAWE2JsIsACBtCLNIAsJsTIRZAEDaEGaRBITZmAizAIC0IcwiCQizMRFmAQBpQ5hFEhBmYyLMAgDShjCLJCDMxkSYBQCkDWEWSUCYjYkwCwBIG8IskoAwGxNhFgCQNoRZJAFhNibCLAAgbQizSALCbEyEWQBA2hBmkQSE2ZgIswCAtCHMIgkIszERZgEAaUOYRRIQZmMizAIA0oYwiyQgzMZEmAUApA1hFklAmI2JMAsASBvCLJKAMBsTYRYAkDaEWSQBYTYmwiwAIG0Is0gCwmxMhFkAQNoQZpEEhNmYCLMAgLQhzCIJCLMxEWYBAGlDmEUSEGZjIswCANKGMIskIMzGRJgFAKQNYRZJQJiNiTALAEgbwiySgDAbE2EWAJA2hFkkAWE2JsIsACBtCLNIAsJsTIRZAEDaEGaRBITZmAizAIC0IcwiCQizMRFmAQBpQ5hFEhBmYyLMAgDShjCLJCDMxkSYBQCkDWEWSUCYjYkwCwBIG8IskiDRYfbBBx+0WrVqWY0aNaxx48Z+bdlatmxpRx55pO23336utGrVyh9TNsIsACBtCLNIgsSG2WbNmlm9evVs2rRptmDBAmvUqJE1bdrUHxtN4w877DDr16+fGx49ejRhFgCALAizSILEhtnq1atbhw4d/CGznj17upZWhdsoqtf4QYMG+TV7hjALAEgbwiySIJFhdu3atS6YTp482a/xqK5Xr17+0O769u3rxs+aNcu16qpbguriIswCANKGMIskSGSYnT59ugumGzZs8Gs8qmvTpo0/tLt27dq58epf26RJE7vhhhvskEMOydrNYNy4cXbccceVlKOOOsqqVavmjwUAoOojzCIJEhlmZ86cmTXMtm/f3h/aneo1/uWXX/ZrzO69915Xt3HjRr9ml/Xr19s777xTUrp27eq6NgAAkBaEWSRBIsPspk2bXAiN6mbQv39/f2h3qtf4lStX+jVm8+bNc3XZ+tmG0c0AAJA2hFkkQWIvANOdDDp37uwPmbuw6+CDD7alS5f6NbtTvcaHLwALuh4sW7bMr8mOMAsASBvCLJIgsWH2vvvuc4F24sSJVlRUZA0bNrQ777zTH2s2adIkq127tq1YscKvMTdet/DSLbmGDBliderUsUsvvdQfWzbCLAAgbQizSILEhlnRAxBq1qzpQmbmQxPUdeDEE0+01atX+zUeTafp9b6495gVwiwAIG0Is0iCRIfZykSYBQCkDWEWSUCYjYkwCwBIG8IskoAwGxNhFgCQNoRZJAFhNibCLAAgbQizSALCbEyEWQBA2hBmkQR5DbN68pbuOqBbZRU6wiwAIG0Is0iCvIXZpk2bugcWKCAGYVb3ju3UqZN7XWgIswCAtCHMIgnyEmZfeeUVu+6666xPnz722GOP2ZgxY1z9hAkT3MMPChFhFgCQNoRZJEFewqweXNC1a1f3+s9//nNJmN22bZsdcMAB7nWhIcwCANKGMIskyEuYvfLKK13rrOgpXkGYHTp0qHsEbSEizAIA0oYwiyTIS5jVY2Qvuugimz59urVo0cKF2QEDBti1117rhgsRYRYAkDaEWSRB3i4Au/XWW90FYGeffbYde+yx7nWtWrX8sYWHMAsASBvCLJIgb2FWRowY4e5e0KZNGxs0aJBfW5gIswCAtCHMIgnyEmYbNWrkuhokCWEWAJA2hFkkQV7CbLNmzQizAAAUOMIskiAvYXb27NlWr14969y5sy1evNivLWyEWQBA2hBmkQR5CbNqldUFX1FFXRAKEWEWAJA2hFkkQV7CrB5fW1YpRIRZAEDaEGaRBHkJs0lEmAUApA1hFkmQtzBbVFTkuhvoaWB6vK2eBFaorbJCmAUApA1hFkmQlzA7fvx41z+2bt26dsMNN1jTpk2tQYMGri54zG2hIcwCANKGMIskyEuYveeeeyIv9FJLbf369f2hwkKYBQCkDWEWSZCXMKsgm61LgVpnCxFhFgCQNoRZJAEtszERZgEAaUOYRRLQZzYmwiwAIG0Is0iCvJ3TV6C97LLLXAttUHr37u2PLTyEWQBA2hBmkQSF2UG1ABFmAQBpQ5hFEuQlzA4YMMD1j82kuqj6QkCYBQCkDWEWSZCXMPvAAw/Yc8895w/t0rdvX9d3thARZgEAaUOYRRLkJcxyay4AAAofYRZJkJfkeMcdd1izZs38oV3atm3LrbkAACgQhFkkQV7C7OTJk6169erWvHlz10I7c+ZM+9Of/mQ1a9bk1lwAABQIwiySIG/n9NUKq0CrbgVBady4sT+28BBmAQBpQ5hFEuS1g+qOHTusqKjIPvzwQ9u8ebNfW5gIswCAtCHMIgkK4mqr5cuX29y5c/2hwkSYBQCkDWEWSVCpYVZ3MejYsaM/5Dn11FNLuhnUqVPHBg8e7I8pLIRZAEDaEGaRBJUaZg888EBbu3atP2Q2aNAg+9nPfmadO3d23Q3OOecca9KkiT+2sBBmAQBpQ5hFElRamJ03b55rfQ279NJL7fbbb/eHzLp06WJHHnmkP1RYCLMAgLQhzCIJKi3Mzpo1y4XZNWvWuOEFCxa44V69erlh0W26MgNvoSDMAkA67fzqG9vw2Re2dMN2m71yi01ZvMHenfupDZq10npNWWqvvLfI/jFqvv3t7TnWst+H1rzXDLul61T7badJduVLE+yyF8fbRS+8Z+e3G2tntRljZz472ho+846d/PdR1uCvI+2XTwy3eo8NsyMfHWo/ffjt3Yrq/q94XP1Ww+24J0fYiX8baac89Y6d0fpda/TcaDvn+TF2/j/es4vbj3Ofo8/rN32Fv+T7jjCLJKi05Kg7F9SoUcO6d+/uhtu3b2/HHXecex1QmOVxtgBybeP2L+zjVVts7Ly19p+py6z9Owvskf4f2m3d3rfLOoy3E4oDg4KEgsI9/5lpnccttvc/2ei/O702bf/Slm/cYR+v3mrTirfHe/PX2YjZa1yoe33acus5+RO3rV58d4E9P2KeC3ePvvmRPfT6LLu7OOD9/tVpdlOXKS7kXf/KZPf61m5TXf2dPae7ae7rM9NN/0hxKNR7n3hrtpvPM0M/tjbD59qzw+bak4Nm25/6f2T39/3A7nptut3e/X27sfNka/zPiXZJcajT96awp+B3THEIzAyIVb28NGah/43tO8IskqBSm0FbtWrlWl51IZj+7d27tz/Go/FRTwYrBIRZFILtX3xta7futMXrPrMPV2y2yYvW26R9KApz73z8qQ0vWm2DP1xlb32w0rXq9H1/mb02Zam9OvET6zJ+sf3rvUXuAKnQ127kfNcSpfCiIDN1yQabt2arfVq8XPmybedXtqI4ZBWt3LLbNpm4cL09URx8FHiufnminV4ccKIO/ntS1ML25wEfue3zUfF3kBTrtu20BZ9ucyF01JxP7Y3py61LcfBsN3KeC4cPvzGreDvNsN/9e4oLhRf84z0XCNUaGLUd0lbUeqrW1PPajnX7koK4AviDxcG71VtF1rY4vL9c/DfSY9In1n/GCve3of0v/PdWWWXV5h3+t77vCLNIgko/pz98+HAXWlesKH0aRPV9+/b1hwoLYRYBBUqdctQBQ6FSLXwzl23ap2CpgPiXNz+y+/t8YE1fnWY3vDLZnTI8+/kxdspTo+zox4dFHmALsei0qE6f6qB/TadJrtVNB3y1rik47U1RK9/jA4tcK6nC1hUdJ7hTrHsbtP7vL8Pc+6/71yQ3T7X6KbQr0Ie/l/EL1lmnsYusWXFo+dWz2YOwTu0qNOuHgIL9vvj8y69dC+jqLZ/bkvXF+9fqrfbB8k3u1HZ42cJFoempwXOsRXEg/UOPaa7lU6eedTr6qL+UPnW9t0X7ofZHtXz+pnj/1PbT6XSFOrWoqjVVLanPDPk48nusiPLCqPnuO+le/EOrT/GProHFP8D0Y2xM8Q8zbaMPiv8Wtc207bQN1Qq/o3ibYu8QZpEEhdlBtQARZpPns51f2Zrig9mitZ+5MKCDvlpL1Gqi1hO1NCokqVVFYeuO4hBwY+cprtVF/c8UJE975h3Xp61QwmTdPw9xAU4tZmo5UwuaAmM+i4LcvgTLiihBgFafRC2TArROVStA63tWyFTgW7h2m/sxsrfUAvze/HXW4Z0FrkvCScWfGbU8hViOfWKE66t5ecfxdnOXKe6Uvn5AqUVR4fC1yUvtzZkrXautQqFanfVjTS26hMH0IswiCQizMRFm80/h9JP122360o0ulOrCC/XNUxjVBRdqJVJr4PF/zW2oUqDUxRgKMmqt02de2mGcC5ZaBp1+VGtV0A9Qy6ZWKwXmoB+gltn1Axzi9QPU6Xu1DOr0vloHdfpfp4PnrtnqTp+rpS4pwqf81bKp9dEp+ZdGL4xsaYtX5ru+mIXUtSGwftsXbpn0PerHUEX00dSPJ/XbVbcItYKqH6j2L7XYK0Sr24T6i6rfqLoItB421/0tqM/qoFmrXOBWN5RlG7bblh3J2XdQeAizSALCbEyE2Yqlfp/qv6eLakbOWRPqvzd/txCjVtJTn37HBciog35Z5RePDnWtqmqNUj9HnXrVxSZqkVKo/PvgOe4KZIWk3lOXuYtYRs/1T1Uu3+SCpMKzWnc3FwcCXdEMAGlCmEUSEGZjIszuGbVWDflotQuMaq30+u+N2uf+e+oDqFYq9ZtUy5RapHRxkk4jj1uwzrUGqp8cAGDfEWaRBITZmAizZdMpzX+PX+KuhtYVv5khVH0adfo/aCVVv0adig9aSXXKPbOVVPdx1Kl2XcihU9cAgMpFmEUSEGZjIszuoquD1TVAV4ArlOp0fji4/qzlYLuwOLCqX6j6S2p6AKiqtn+13TZ+vtFWfbbKPtnyic3dMNdmrZ1lU1dPtYkrJ9q4FePs3WXv2shPRtrQJUPt7UVv25sL37Q35r9hfeb2sdfmvGbdZ3e3Lh91sX99+C97+YOXXdFr1Wlczzk9rffc3vb6vNdtwIIB9tbCt2zw4sE2fMlwG7V0lI1eNtp9jj5v5baV/pLtO8IskiDRYfbBBx+0WrVquYcxNG7c2K8t35IlS9x9bvfkaWNpDrNzVm1xF5boIia1rIaDa1B0hbTCra6E1oVaQJIs2bzEJq2a5IJCh5kdrOV7Le2hsQ/ZsCXDXDjB3nEhb+dGW/3Z6t1C3vtr3rfpn053r2evn+3qF2xa4KZZvnW5m37tjrXuvVu/2OrmE7bty222fsd6F9oWb15sczbMsZmfzrTJqybbmOVjXMBT2NP32WNOD+v8YWfrOLPjXhXtDzcOudEav9XYLh1wqZ33xnn2q//8yk557RQ7utvRBVle+fAVf0vtO8IskiCxYVYPV6hXr55NmzbNPRpXD2Jo2rSpP7ZsCr7BAxziSkuY3fr5V+4iKN2ySldO6zGKmcFVXQbUD1YXaE1YuJ4LoxJux1c7XGBQcFi3Y50LEiu2rXDBYuGmhS5oKHDMWjfLBRAFkTnr57ggoVBR6LZ8scU+3vCxjV0+1rVutZ3W1u4bfZ9dP/h6O+M/Z0SGgcxy8msn2+3Db7cXpr/gWtjWbPcey11V6Hv8dPunLhh+uO5Dm7J6ir2z9B0XCHt93MuFQYW6p6Y8ZX8e92e7d/S99vsRvy8V8s7sc2ZBh7xclQY9GthpvU+zs/qeZRf1u8iuGHiFXfv2tfa7ob+z24bfZn8Y8Qe7a9Rd9sd3/2j3j7nfWoxtYY+Me8T+MuEv1mpiK/v75L/bM1Oesefef87aTW/nhegZHdzrNu+3sdZTW7tpnpz0pD0+4XH3HTz83sP24JgH3b589zt3W7NRzdzn6PPUYltRCLNIgsSG2erVq1uHDh38IbOePXu6cKpwW5auXbvar3/9a8KsT7c3Uh9V3ZNTzwzPDK4qun+o7gKg+3WqDysqngLi8E+Gu9OHClydZnVyoeuvk/7qWgl1sNJB6reDfmuXDLjEHTRP6nlS5IE1H+WM3mfYxf0vthsG32B3jLzDHWh18NVBWS1jAxcOdC1mOu26N0WhSttDYUoHc81fB/E7R91ptw671ZoMbRJZFCailjezKIhpegWM9jPal7TKKUxou5/Y88RS71EQVoD456x/2viV411oLhTbvtjmfoyotVKnoPvO6+uWU9+JwpS22WUDLrPTe59ear2SUPR9aJ875/Vz3H531VtX2fVvX+/WS/ufwrb2Ee0rWufnpz1vL8580W0DBfNuRd3cfhmctu8/v/+u0/bFf4cK8tpf9b2qtVc/4orWF7nWY7Ucq9VYPwDTgDCLJEhkmF27dq0LopMnT/ZrPKrr1auXP1Ta6tWr7Sc/+YlNnz49tWFWt5zS7a900/R6Ea2uKroZv55T/8a05e6m6VWVDvhqiVQrpFog1fo449MZ7uA14pMRexQs1XIXddCl5L+o1Uwth2q1UiuYTsEOWjTIBRT1cYxLLdQKPwq8ClBRn3X+G+fb/aPvd8EpCMSVUdSK13hQYzu779mRy1VeUThs1KeRW69rBl3jQmHzd5pbi/dauNZAhcF/fvBPe3X2q9Zvfj93Gl/9M6eviQ55O7/O//1/UTEIs0iCRIZZhVEF0Q0bNvg1HtW1adPGHyrtlltusbvuusu9Li/Mvvfee3bMMceUlLp161q1atX8scmzfOMO+2PvGaWCq27wrttc6U4CurXVZ18UXn9Xnf5W4Jy3cZ4LmzqIqi+jDqo6uOogq1NxT0x8wh18dRDWwfi6t6+zKwde6Q7QOsirFeqEnidEHswruigcnNrrVNd6pJCgFiQFHS3Lb978jVsuhQa1Jt085GYXjnUK8k/j/uRakhROdPGHTvHqQhFdOKKQrT6GCt7qClBILUMKMVoufT86Da9l1kUrCnVaH51WVZi8Zdgt7vS0Ws/0o0AtaA+MecC1ounUqVrSFJ70nmenPutClE7tK7AphOr7ViujWnoVqNR6pu2iz1XXh0WbF7nuD/qRoh8rubRp5yZ30Y1ab9UCrNActS/ko2hZ1NqslnK1Hj86/lG3HfWdKMirf7C6XqhrAVAWwiySIJFhdubMmVnDbPv27f2h3fXt29cOO+ww277du5CgvDCreY8dO7akvPrqq65rQ9Js+OwL90z7cIB9+I1Z9p+py/b5GfJ7Qqdgl21d5lpxdLXtkMVD7D9z/+NaPRVadLBVuFEouPzNy11rZ67DgVpTdar43NfPdS2sV791tTv4xw6W6wozWCJ/Plr3kWvNz2w53ZMSnArvWtTVnQrXflfqVHhxkM88Fa4fOmodzbxYCtgXhFkkQSLD7KZNm1wQjepm0L9/f39od7pYTAE2XDS9/h09erQ/VXZJ62agOwroAq3/+4vXleDnjwx2j1NdtXmHP0XuqMVHB1+dllfrY1SQ3JOiFk61bKpFUy2ZatlTi55a8tSHUhem6OCvg74O9jrI6wKWD9Z+4JZlyeYlrrVuw+cb7LMvq263CQCoaIRZJEEiw6wonHbu3NkfMhs0aJAdfPDBtnTpUr9md5lBVqUqhtmvvv7WPeNfF20FLbG3d3/fPTo2F9QyqdOWal3Vlc1RYVRFFyupBVStn2r51MU76r+oU7S6j6LCry68UAhVAFX4JHgCQH4RZpEEiQ2z9913nwu0EydOtKKiImvYsKHdeeed/lhz9T/96U9txYoVfs3ugjAbV6GH2W+/NfdI19NCT9+66qUJNnPZJn+KfafT6TrFHlzhfUqv6FvwqIuAugyoC8GElRNy3ncRAJAbhFkkQWLDrLRs2dJq1qzpQmbmQxN0kZgCru5gEEVhVvemjauQw+w7H39q57cbWxJiz//He65uX3xb/N/8jfPd1du6l+Gv+/w6Mrjq3opNRzR1tzPShT9cUAIAVQdhFkmQ6DBbmQoxzKrVVa2vQYhVq2z/GStcK+2e+ubbb9zFK7r/olpVFVIzg6uu0L956M2uS4Eu4NIFXQCAqoswiyQgzMZUSGFW9369rdv7JSFW/WO7T9yzR25+8fUXNm3NNHfltC6oiroBvwKtgq0Crp4KpMALAEgPwiySgDAbUyGF2Uvaj3Mh9hePDrU2w+fa9i++9seUbenWpfaP6f+wm4bcVCq4quh+qLrnp+4KoHu6AgDSjTCLJCDMxlQoYXbr51+5IHvUX4a6e8jGoacc6RngmeFVN/HXvVT18AE9+hIAgDDCLJKAMBtToYTZEbPXuDDb+J8T/Zrs9FQmPU0pHGB1Oyw9PUtPSAIAoCyEWSQBYTamQgmzTwya7cJs2xHZuwHo8a+6SCt4gtYvu//ShVjuNAAA2BOEWSQBYTamQgmzF73wnguzExau92t20SNV1SdWdx1QiK3fvb49Mu4RW7Et+l67AACUhTCLJCDMxlQIYfazL76yn7UcbLUfGWw7v9p1ZwE9i/2fH/yz5CEGx3Q7xu4ffb97jCsAAHuLMIskIMzGVAhhNrO/7M6vd1rXoq52Ru8zSvrENhvVzD3sAACAfUWYRRIQZmMqhDD7pN9f9tnhs93ts8JP5dLjZYvWF/lTAgCw7wizSALCbEyFEGYv9u8ve/1bTUtC7PWDr7epq6f6UwAAUHEIs0gCwmxM+Q6z4f6yx/do4O5UMHb5WH8sAAAVjzCLJCDMxpTvMDtyjtdf9qJ/dnctspe/ebk/BgCA3CDMIgkIszHlO8z+7e05Lsze3O8pF2b/Oumv/hgAAHKDMIskIMzGlO8we4nfX/aaN29xYXbI4iH+GAAAcoMwiyQgzMaUzzC7q7/s23ZSz5NcmNWjagEAyCXCLJKAMBtTPsPsqDmf+v1le7kge0G/C/wxAADkDmEWSUCYjSmfYfbvg73+srf0a+PC7J/H/dkfAwBA7hBmkQSE2ZjyGWYv7eD1l73xrTtcmO0/v78/BgCA3CHMIgkIszHlK8yG7y/bsNdpLswu3brUHwsAQO4QZpEEhNmY8hVm3/nY6y978ctvuCB7Wu/T/DEAAOQWYRZJQJiNKV9h9im/v+xt/du7MPvAmAf8MQAA5BZhFklAmI0pX2H2sg7jvYu/3r7XhdleH/fyxwAAkFuEWSQBYTamfITZoL+sSqM+jVyYnbdxnj8WAIDcIswiCQizMeUjzI6eu9brL/vSQBdk9cAEAAAqC2EWSUCYjSkfYfbpIR+7MNu0fycXZu8adZc/BgCA3CPMIgkIszHlI8xe9qLXX7bpkIddmP33R//2xwAAkHuEWSQBYTamyg6z4f6yF/e/xIXZWWtn+WMBAMg9wiySgDAbU2WH2THzvP6yl3Qc7oJsgx4N7Otvv/bHAgCQe4RZJAFhNqbKDrPP+P1l7xzQ1YXZ24bf5o8BACTKlzvMdm4127HRbNunZltWmm1aarZhkdm6eWZrisxWzTJbMc1s2WSzT8abLRm392XTMv+D9x1hFklAmI2pssPsb/z+sncNe8yF2Zc+eMkfAyCnPltn9ukcL1DoNaoOhUoFSoVJBUmFSAVIfdcLRhX/j36Q2Yevm8141WzKv8wmvGA29lmzUU+YDXvEbNC9ZgPuMOv7O7Ne15q9+huzLuebdfqV2Ysnmf3jGLM2dc2ePtzs8YPyV95r46/wviPMIgkIszFVZpjd+dU3Jf1lrxx4lQuzU1ZP8ccC2CNffGa2cYnZ8qnFf8hvm03rVhxQnjMb0sLs9VvMul9q1vEUs2drRwcDBZQ3bvPCzaoP/JmmlIKgWhQVAtWCqABYNMALf5OKf3Ar+I34ixf6+t3uBb5uF3thr/3xXtD7+0+it3NVL3//Hy/ktv6Ztx3aHlW8b9X3tkvHk81ebuhtp1fO9gJy14v2vnzY1//C9h1hFklAmI2pMsPsWL+/7MUvjrL63evbL7v/0nZ+vdMfi0T7fIvZtjXF4eoTs7Ufm62cabZ0ktmiMWZzhxQHg/5mM18ze79LcTjoaPbe82bv/t1s+KNmgx8we/MuL1j95wazHld6B67O55r969dm/zzD7KVTzV480eyFY83aHW32/C/MnqtTfACtZfbUYWZ/K96How60e1qeqlE8/3rFn3m6Fwb73GT21h/NRrUyG9/ObHpxYJzzlnfKc/VHZptXeKFyXyhIqTVtxXSzhe94LWhTX/EClFrNBtzphafwQf3fF0Yvf1RR0FCoCL//X41KT6dtqHHvPGk2b1jxcm3yFzABtm8o3u/mei2R2temdDIb/bS3/QbeXRzum5j1vMrbbgpXClsKX5nbIMlF37PCpH6kqDVVAfLfF3itrNp/1Oqq1lcFcm0XtcpqH5vQ3vtBo+CufU+tuArz2pZq3VXA1/6pVl/tq2oFrgIIs0gCwmxMlRlmnxnq9Ze9a8BrrlX2hsHFwQUVa+e24oC1vDhofWi2+D2z2QO9FrtxxUFs9FNmIx/3AuTQh70QOegeL0j2/0NxmLy1+IB3sxcoX7vGC5UKdC78FIdKBcr2x3lBUiGyogIkZe/KX/+fF8oU0BTW9L0qmCiQfDLBCyBlUShZPNb7UaHvOOoz9ANCAUg/QhRq8kE/hqZ3934ADW3p/ehRQNP++NwR0cu9p+Xpmt5+rRD4ylnF+/1lxX8HN3rrPvjB4oD/1+K/obZe6Pugl/eDRttOYU8hWkFPP+iQGIRZJAFhNqbKDLOXd/T6y9474m8uzLadVnxwwC5ffW722driELLYC6MKJGoh++gNL5BO7OC1Nuk0cr/fFwfOxl7rZYcG2U8lV1ZRi6aWoe3/ecvz8mnFy3aOWbdLvOXsUxwMtMxv3e0t/8jHvHVRyJ78srd+s/7jhe/5w70grvXXKfSVM7xWUPX3XL/AO7WuwL51tdf38/PNXuuott++0sUsCia6eEUtpVoOtVR90NtrLVWgefdvxevwkNdiqsCjYBW1TeIWtSxXdmtwNksnev0SX708+rS5WjWD1t09LcEPIrWuKzg+89OK+UGkeekUtn6EKXjqR5srxfuX+oa+/2/v9LRCsVob1aVi/ULvTEIVaWXEniPMIgkIszFVVphVf9naj3j9Za9/+wYXZscuH+uPTTC1hG5dVRx+5nuhS4FD/RcVzKZ29sKaCz/FAe7NZl6oU1BQ3zH1Z1T4e+Z/ow/Se1MUDtoc6c1bLXa9r/M+d9ifvIO7gooCkoLx5H96y6iwNLOnt8wf9fMC5dzBXqhc+K63TsumeAFb66kgqdCtdUfVpkCvfaX39RW7n0aV4AeRgn2HE7xwrx9E6h4wsLnXgqxgqv1SgVRhFNhLhFkkAWE2psoKs+/N9/vLtn/X9ZVVn9ntX233xyaEbjEzravZ2/d5B9qoA/K+FrVW6WCuFiy1uva4wmupU1cAdQ3QAV3hQuFToVktaWqxVKAGKssXxX+7ahHfvt5rIVdrsfpLq+VcfabVgqx+02pZ1z6qH0R6rfqSH0TrctfCDJSDMIskIMzGVFlhtvWwuS7M/nHA665V9qq3rvLHFCgdkNVSqT56CpVRwVNFLaG6EEmnThVwdTpV/U3V/1Sniof/2WzMM8UB9EWvBVQtn/NHeAd4Hdh1ylyBAABQaQizSALCbEyVFWav6DjBhdkHRz7nwuxTU57yxxQAtSqpT6IujlKfRV0MkhlaVafuAeqLpyvO1bcSAJBIhFkkAWE2psoIs+H+srcMvdWF2eFLhvtjK5luLaOWUbWWqgVVraqZwbXVwd5FTOprqguTPp1t9u23/gwAAElHmEUSEGZjqoww+978dX5/2bHWoEcDF2Y379zsj80h9etTX73x//BuOaX7k2YGVxXd9FtX3ev+mrq4JEn31wQA7DHCLJKAMBtTZYTZZ/3+svcMGOiC7MX9L/bHVKCvvzBb/r53s3TdM1X3x2xVrXRwffJQs05nehdxzejht7p+488EAJAGhFkkAWE2psoIs1e95PWXfeSd9i7MPjbhMX9MBVg02ru/5BM/LB1cFWb15CPdD1T3CNVthhR6AQCpRphFEhBmY8p1mA33l71j5F0uzA5cONAfu5d0SyDdHUDP/g6HVz0NSU9C0lOQdHN0bogOAIhAmEUSEGZjynWYHbfA6y974Qvv2Uk9T3JhdsW2Ff7YPaSb9ut+q+GnBv21uvdkKd3UHwCAGAizSALCbEy5DrPPDff6y97/5lAXZH/d59f+mJi+2mk28zXvUZi7tcIe4z3JSncnAABgDxBmkQSE2ZhyHWavfnmiC7OPje7kwmyLsS38MeXQwwSGP2rWutauAKs+sK819u44wK2yAAB7iTCLJEh8mN2wYYOtXr3aHyrfkiVLbNasWf5QfLkMs0F/WXcng3fvd2G2z9w+/tgIuqvA3CFmPa707vUahNjWPzMb+ZjZpmX+hAAA7D3CLJIg0WH2wQcftP3228+Vxo0b+7XRWrVqZSeffHLJ9HXq1LG7777bH1u+XIbZ8X5/2Qv+8Z7rXqAwu2DTAn9sBF28FQRYFT1GdlYZ4RcAgL1AmEUSJDbMNmvWzOrVq2fTpk2zBQsWWKNGjaxp06b+2NIUZtu3b29FRUW2du1a69ixowu1qo8jl2H2+RHzXJh9aMA7Lsie1vs0f0yEndu8APvX/2f21h/NVn/kjwAAoGIRZpEEiQ2z1atXtw4dOvhDZj179nThVOE2rtq1a9vll1/uD5Utl2G28T+9/rJ/HdPNhdk/vlscUrMpGuCF2Vd/41cAAJAbhFkkQSLDrFpWFVwnT57s13hU16tXL3+ofAqn6qoQR67CbLi/7ENj/uTCbPfZ3f2xEfrd7oVZPcELAIAcIswiCRIZZqdPn+6Cqy7+ClNdmzZt/KGytW7d2vWhzZxHYOzYsXbUUUeVFLXiVqtWzR9bcSYsXO+C7Pn/eM8u6HeBC7NF64v8sRm++drsqcO8MLt1lV8JAEBuEGaRBIkMszNnzswaZtUvtjzDhg1z0w4aNMivKW3jxo02YcKEkqIWX3VtqGht/f6yDw8Y74Jsgx4N7BvdrSDK4ve8IPtyGX1qAQCoIIRZJEEiw+ymTZtcGI3qZtC/f39/KJou+NJ0ffv29WviyVU3g6C/7NNje7kw23RE9ovYbGhLL8yOfsqvAAAgdwizSILEXgCmOxl07tzZHzLXynrwwQfb0qVL/ZrS9jbISi7CbLi/7KPjnnBhttOsMvrCtjvaC7OrPvArAADIHcIskiCxYfa+++5zgXbixInudlsNGza0O++80x9rrmvAYYcdZsuXL3fDPXr0KAmyo0eP3q3EkYswO9HvL3te27F2+ZuXuzA7bU2WuzF8OtsLsm2O9CsAAMgtwiySILFhVlq2bGk1a9Z0ITPzoQkzZsxwf4Rr1qxxwxqve9FGlThyEWbbjfT6y7Z8c7ILsr/s/kv7+tuv/bEZxj7nhdm37/crAADILcIskiDRYbYy5SLMXtNpkguzbcb1d2H25iE3+2Mi/OvXXphd+I5fAQBAbhFmkQSE2ZgqOsyG+8v+bWJrF2ZfmP6CPzbD9g1ekP1b8efr9lwAAFQCwiySgDAbU0WH2SmLN7gge27bsXbt29e6MDtuxTh/bIb3/+2F2b6/8ysAAMg9wiySgDAbU0WH2b7vL3Nh9k8DZlj97vVd2f7Vdn9shp5Xe2F2Vh+/AgCA3CPMIgkIszHlos/sji+/tqELx7pW2WsGXePXZvhyh9mTh5o98UOznVv9SgAAco8wiyQgzMaUizArHWZ2cGG29dTWfk2GOW95rbLdLvYrAACoHIRZJAFhNqZchdkmQ5u4MDtq6Si/JkP/P3hhdtJLfgUAAJWDMIskIMzGlIswq3vKNujRwIXZzTs3+7Uh335r9tRhXpjdlP3JZgAA5AJhFklAmI0pF2F2xqczXJC9bMBlfk2GT8Z7QfalU/0KAAAqD2EWSUCYjSkXYfaVD19xYbbVxFZ+TYbhf/bC7DtP+hUAAFQewiySgDAbUy7C7B0j73Bh9u1Fb/s1Gdod7YXZFdP9CgAAKg9hFklAmI2posPsN99+Yyf1PMmF2bU71vq1IWs/9oJs61p+BQAAlYswiyQgzMZU0WF2zvo5Lsie98Z5fk2GcW29MDvoHr8CAIDKRZhFEhBmY6roMPvanNdcmH1k3CN+TYZXzvbC7PzhfgUAAJWLMIskIMzGVNFhdufXO23iyon24boP/ZqQ7RvMWh1s9rfiz/v6C78SAIDKRZhFEhBmY8rFBWBZTe/mtcr+50a/AgCAykeYRRIQZmOq1DDb67demP2gl18BAEDlI8wiCQizMVVamP1yh9mTh3rdDHZu9SsBAKh8hFkkAWE2pkoLsx8P8lpl/32hXwEAQH4QZpEEhNmYKi3MvtnMC7MT2vsVAADkB2EWSUCYjalSwuy333oPSVCY3bTUrwQAID8Is0gCwmxMlRJml07yguyLJ/oVAADkD2EWSUCYjalSwuyIv3hhduTjfgUAAPlDmEUSEGZjqpQw2+5oL8wum+JXAACQP4RZJAFhNqach9l1870gqz6z6jsLAECeEWaRBITZmHIeZsf/wwuzA5v7FQAA5BdhFklAmI0p52G2y3lemJ07xK8AACC/CLNIAsJsTDkNs9s3eE/8+lvx/L/+wq8EACC/CLNIAsJsTDkNszN6eK2yva/zKwAAyD/CLJKAMBtTTsOsQqzC7IxX/QoAAPKPMIskIMzGlLMwq24F6l6gbgbqbgAAQIEgzCIJCLMx5SzM6oIvtcp2PtevAACgMBBmkQSE2ZhyFmYH3u2F2fHt/AoAAAoDYRZJQJiNKSdhVg9H0EMSFGbXzfMrAQAoDIRZJAFhNqachNnlU70gq8fYAgBQYAizSALCbEw5CbMjH/fC7PBH/QoAAAoHYRZJQJiNKSdh9sUTvTC7dKJfAQBA4SDMIgkIszFVeJjdtNQLsuozq76zAAAUGMIskoAwG1OFh9mJL3phdsCdfgUAAIWFMIskIMzGVOFh9ssdZvOHm63+0K8AAKCwEGaRBITZmHLSZxYAgAJGmEUSEGZjIswCANKGMIskIMzGRJgFAKQNYRZJQJiNiTALAEgbwiySgDAbE2EWAJA2hFkkAWE2JsIsACBtCLNIAsJsTIRZAEDaEGaRBIkPs2vXrrUVK1b4Q/HMmjXLduzY4Q/FQ5gFAKQNYRZJkOgw27JlS9tvv/1cady4sV+bXVFRkZ144olu+h/84AfWqlUrf0z5CLMAgLQhzCIJEhtm7733XqtXr55NmzbNFixYYI0aNbKmTZv6Y6OdcsopLvSuXLnSJk+ebAceeKB16tTJH1s2wiwAIG0Is0iCxIbZGjVqWIcOHfwhs549e7oWV4XbKAqjGj9z5ky/xuzuu++2Bg0a+ENlI8wCANKGMIskSGSYVT9ZBVO1roaprm/fvv7Q7lT//e9/3x/yjB492r1n9erVfk12hFkAQNoQZpEEiQyz06dPdyF0w4YNfo1Hde3atfOHdte2bVurX7++P+QJwuyMGTP8ml007ogjjigphx12mP33f//3bnX7Wv7nf/7HDj/88MhxFG/7qESNo3hF26dWrVqR4yjsQ3GKts/Pf/7zyHEU9iE1Aj355JP+kREoTIkMs+oqkC3Mtm/f3h/anUJutjD7wQcf+DW7bNq0yaZMmVJSRo0aZS+99NJudftaFJC7d+8eOY4yxR555BE7++yzI8dRvKJ+3/37948cR5lid911l11++eWR4yhe+a//+i8bOXJk5DjKFPvd735nN998c+S4NJQ+ffrYkiVL/CMjUJgSGWYVNBVCo7oZ6MAeZcCAAVm7Gaxbt86vqVxqDcnWxxfmLs6Lc5eKNPvhD39oixYt8oeQ6Zlnnin3wtC0U5jNbBjALrprzsMPP+wPAShEib0ATHcyyLwA7OCDD7alS5f6Nbtbvny5C66ZF4Cdc845/lDlI8yWjTBbPsJs2Qiz5SPMlo0wCxS+xIbZ++67zwVa9Z8Nbs115513+mPNxo8f7y7YUogNnHrqqS4crVq1quTWXF26dPHHVj7CbNkIs+UjzJaNMFs+wmzZCLNA4UtsmJVu3bq5W2sdffTRpTqoqwX2rLPOsjVr1vg15v6HrXCkAKBxr732mj8mP3r37k2YLYO2jwqy0/YhzGbHPlQ+bR/CbHbsQ0DhS3SYBQAAQLoRZgEAAJBYhFkAAAAkFmEWAAAAiUWYzRNdHXvppZda8+bNbdKkSX5t8umCu+HDh9tzzz1nrVq18mtLe/DBB+3CCy+02267zaZOnerX7qLHD99000120UUXRc5n8+bN7o4WuovFHXfcYUVFRf6YXXSBny74u/LKK8tclsrWr18/t95XXXWVde3a1davX++P2aW8ZY+z/q+++qo1adLErrvuush56D333HOPm4f+jZpHPmj7aJlUtJ9of8qke023aNGizGXXOmvd92X9y5tHvmnf0XJFLVt5y14R6x9nHvmgZY0qYZW1/uXNA8C+I8zmgUKKguygQYPc/9wOOOAAGzx4sD822bQ+derUcffw1X19o2j9VYL11y3Shg4d6o81V6/36mluetiFbsGmA0FAQUZ3sdA89OALzUPT6JZrAdVpHrq9l+5BrPGqyzctgw58nTt3doH9kksuseOOO862b9/uT1H+su/N+teoUWO3eejqdb0ncx4rV670p8gfbZ8HHnjAffd6cp/uPPL000/7Y71lP/bYY8tc9opY//LmUQhuvPFGt71Uwipj/VesWOHeo7rwPFSfb9oeWp7MEgiWvbz11zpr3aPWX9OWt/7heWg7anuG5wGgYhBmK1kQ1MKtkZdddpndfvvt/lDVoP+5az0zaf0PPfRQf8ij9Vd4CSjot27d2h/y7hmsea1evdoNK+REzSN8kFDYy5yHDirhwJcPeixy2Nq1a926qRU1UN6yx1l/HVTD8+jYsaObR3Cw1sE1cx76EVKIB1rd51PLHoha9iBUBDLXXz+M9nT9NY/w47Ez55FvujXhKaecEhlm46y/1jcsvP47duyw/fffv9T6q07jRC3jUfNQfb4FYTabbMu+J+uvactaf21rbfPMeei7AVCxCLOVTKeGM586pv8pHn/88f5Q1ZAtzCq0RrUiBeuvFiO9T+8PU93AgQPd6+D0epjmcfLJJ7vXW7ZsyTqPN9980x8qHLVr1y55eEecZd+b9Vdo/t73vudO4Uu2eZx44on+UOF4/PHH3Q+cQHnLXtY2jLv+wTwyheeRT2r900Nh9MRDrUd4XeKuv9Y3LLz+em+29Q/mq8+Mmkfmds2HYNm0rFEtxdmWfU/WX9OWtf7a1tnmoe8IQMUp/ZeGnFI/yVtvvdUf8qiVpHr16v5Q1ZDtYHDttdeWaoUOr7/6nOl9mQcgtXiodVF0ajBqHj/5yU/c67Lm8fLLL/tDhUEtNVrW4OEZcZZ9b9e/bt26JY+AjpqHDsQKSIVAy6KiftX6oTNy5Eh/TPSyq+9xsOzZ1l91cdc/mEem8Dzy6frrr3f9iUXhKRwg466/1jcsvP5xwly2QBhelnzRMvzv//6vHXPMMW6ZNTxkyBB/bPZl35P117Rlrb+2dbZ56DsCUHFK/6Uhpy644AJ32jRM/5P97ne/6w9VDdkOBlr/zNOQ4fXXxXB6X3AqL6BWx6Df5Pnnnx85D7U8ysSJE8udRyEItlH4gBhn2fd2/XVKuqx5aDmCeeSblkVF+4tCyb///W9/TPSyq891eeuvurjrH8wjU3ge+aJW/F/84hf+UOkwG3f9w/udhNc/TpjLFgjDy5IvEyZM8F9566KWfXXfCWRb9j1Zf01b1vprW2ebh74jABWn9F8acuoPf/iDa50MU+ucTjVXJdkOBlp/XaEfFl5/nTbV+2bNmuWGA4cccoj16NHDvdadAKLmERzgly9fnnUeasEsBNu2bSvVz1PiLPverr9absuah5YlHJIKxTPPPOMuEvz888/dcNSyqxWsvPVXXdz1D+aRKTyPfPnpT3/q/r6CovCkEoSsuOsfFcSC9c/296u64HOyBcIgzBUSLVd4fbIt+56sv6Yta/21rbPNQ98RgIpT+i8NOaX/2dWqVcs+++wzv8asWbNmBXkA2BfZDgZa/yOPPNIf8qjbRbD+X3zxhbtoQqeNA0uWLNntIJJtHuF+lZo+ah6ZF2DlQ3AaeNiwYX7N7spb9rjrr7slBILvIzyPzAtRdOo5PI9CEaz/jBkz3HCcZc9cf3Xj2NP11/RB9w/JnEe+6G8lWwnEWX+tb1h4/YP9JWr9g79DfV7UPMLLUSjatGnj/r8SyLbse7L+mras9de2zjYPABWLv6pKFlzhqvuwioKNDqqF1pdzXwUHg0y6Iv+www4rOdBGrb8OtNdcc40/5A2rv2dg5syZbt7BPGbPnu3moTslBHR/1fCBRvMIB5V80TJr2XVwzaa8Zd+b9dc9e8uaR/B9heeRL+Fl0AWBWhf9AAxuXxZn2Sti/cubR6FQeMoMkBWx/urfnzkP1QXUj1nv0XslmEe4f3M+aDmCZRL9GPrtb3/r+l8HgmUvb/21zoHM9de05a2/tnl4Htqe+m4AVCzCbB7of6D6n179+vXtoIMOclcWVxUKXlq3zBI+uATrrz5s1apVi1x//U9foV/3E1WQzTxAvvTSS24eOojrFLQ+N0zdFS6++GIXgvR+hb1CuOhCyxveLkEJL3+cZY+7/jVr1nRdOKLmEXxXmofudZw5j3zRMukHT7Ct1Aod3MkiEF7/qGUPfiTty/rHmUch0PKrhFXW+us9em+2eeRDECr196P/x+j1mWeeWerhG3HXX+u+t+sfnoe2Y9Q8AOw7wmyeqMVp8uTJrtWgKtGBJFsJi7P+CxYscPfjDXfJCNPtpjTfzKu2w3TgUN/BzIth8iVzm4RLpvKWvSLWX+8tbx75EN4u69at82t3V976a5217vuy/nHmkW/BdspUWetf3jwqm/pWf/zxxyXbpazlqoz1jzMPAPuGMAsAAIDEIswCAAAgsQizAAAASCzCLAAAABKLMAsAAIDEIswCAAAgsQizAAAASCzCLFCFBPfWXLRokV/jCepzSfPPvHl/PrVv395uvPFGt0y5XncAQP4QZoEqRE8gOuOMM+zqq6/2azwKc3raUS5VxmfEpRvna1n02GAt156GWQXgzKc5AQAKE2EWqEIUwBTEFOSGDRvm16YvzGpZqlevbs2aNdvjICuEWQBIDsIsUIUEYfaee+6xk046ya8tHTSjwlq4Lpi+X79+Jc+3P+2009y4J554wurWrWs1atSwp59+2tVJ8J6+ffvaL3/5S/f6yiuvLPWYz1deecXq16/vxp966qnWqVMnf8yu5de/RxxxhHudTYsWLdzz7n/84x/v9jl6r+YdlGzz0PJefvnlVq1aNTednpsvme9XCeg9WmbVaR1at27tj9m1/s8995x7Fv9BBx1Uav0ff/xx+9WvflUy37LWDwAQD2EWqEKCMLhy5Ur7/ve/by+++KKrD4JWIAiMYeG6YHoN63VRUZGdd955blin7pcsWWIjRoxwnzFt2rTd3nPCCSfYyJEj3bDmee6557rx0rhxY7viiitsyJAhblhBVu8JhoPlv+6662zx4sXuc6IoyGo6fcakSZPsmmuu2e1zgs8ui8a/+uqrtmXLFjes9wQ0LnP7dO7c2S2rwrro38MPP7xkOFj/YLmCZdA6BzR+zJgxtnXrVjfctWtX9y8AYO8RZoEqJAiDwetatWrZzp07S4JWICqsheuC6SdPnuyG5e6777ZDDjnEduzY4deYa50MWlaD97zxxhtuWBRSVTdo0CDbsGGDe92xY0d/rCf8ufpX06xfv94NR9m4caObplevXn6NudbP4HMkCJJl0XiF2TVr1vg1u0Rtn4YNG5aqu/baa10ruATr//rrr7th0TKqTsusHxh6remCMAsA2HeEWaAKUdgKhzh1BXjsscdKglYgKqyF6zKnl8x5S9R7FNzC6tSp41pzp0yZ4sZHlWAeUZ+RSS3Beo/CcZiCtT5HtCzlzUefdcopp7h5KagHrcMSXq9AeHnDJficqPUPAnzQet28eXO3PfS93H///TZu3DhXDwDYe4RZoArJDIPqZrD//vvHCrPHH398Sd2+hNnp06e7YQnCXPfu3d3twvRap9mzifqMTMF8wq3G8qMf/ch9jsQJs4HBgwfb7373OzfPiRMnurqo7aN+wuovnE3U+msZVZd5q7TevXvb7bff7ra5Ws4BAHuPMAtUIVFhUBc23XHHHS5UBdRXUxd2Bd5//303XWYwDYuadzj0Be95/vnn3bDoXq8K0zNmzHDD+ox7773XvQ4Lwl7UZ0TRxWhBK6wohIY/p7wwu3DhQv/VLpo+6LqguyA0bdrUvQ40adIkcp5Bt4uo9dcyBheWZQZa0fTz5s3zhwAAe4MwC1QhUWGwf//+LjSFw+mcOXPsF7/4hbuoS8FNp+j1vooIs7rgSfUqGg7Gi0LnD37wA3dFv06za5ym13sl6jOiDBw40PXf1bTnn39+qc8pL8xq/GWXXebeEyzD6aef7o/1Lu7SPBVgw/PVXRqOO+44+/3vf+/qte2C8cH6n3nmmZHrr/H6HA0/9NBDdsEFF9g555zjxgEA9h5hFqhCFJSC8BQWVa+LkNQSqQuydLeC8DQKXpnTR80j23t69uzpXof7oQbUkqnT7LqtV+Y04fmVZ9asWW7Z1fo7fPhwv9YTtfyZdBeE4PN0T97grgYB1T377LOl5qMfB7oll+p1m7FAEGbV0qp6vTd8r1/Ruup9arHV9FwIBgD7jjALABUgCLMAgMrF/3kBoAIQZgEgP/g/LwBUAIVZFQBA5SLMAgAAILEIswAAAEgswiwAAAASizALAACAxCLMAgAAILEIswAAAEgswiwAAAASizALAACAxCLMAgAAILEIswAAAEgswiwAAAASizALAACAxCLMAgAAILEIswAAAEgswiwAAAASizALAACAxCLMAgAAILEIswAAAEgswiwAAAASizALAACAxCLMAgAAILEIswAAAEgswiwAAAASizALAACAxCLMAgAAILEIswAAAEgswiwAAAASizALAACAxCLMAgAAILEIswAAAEgos/8PAKXZKAZp7rQAAAAASUVORK5CYII=">
## Tokenizer
Le tokenizer de départ est [BarthezTokenizer](https://huggingface.co/transformers/model_doc/barthez.html) auquel ont été rajouté les tokens spéciaux \<sep\> et \<hl\>.
## Utilisation
_Le modèle est un POC, nous garantissons pas ses performances_
```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from transformers import Text2TextGenerationPipeline
model_name = 'lincoln/barthez-squadFR-fquad-piaf-question-generation'
loaded_model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
loaded_tokenizer = AutoTokenizer.from_pretrained(model_name)
nlp = Text2TextGenerationPipeline(model=loaded_model, tokenizer=loaded_tokenizer)
nlp("Les projecteurs peuvent être utilisées pour <hl>illuminer<hl> des terrains de jeu extérieurs")
# >>> [{'generated_text': 'À quoi servent les projecteurs sur les terrains de jeu extérieurs?'}]
```
```py
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from transformers import Text2TextGenerationPipeline
model_name = 'lincoln/barthez-squadFR-fquad-piaf-question-generation'
loaded_model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
loaded_tokenizer = AutoTokenizer.from_pretrained(model_name)
text = "Les Etats signataires de la convention sur la diversité biologique des Nations unies doivent parvenir, lors de la COP15, qui s’ouvre <hl>lundi<hl>, à un nouvel accord mondial pour enrayer la destruction du vivant au cours de la prochaine décennie."
inputs = loaded_tokenizer(text, return_tensors='pt')
out = loaded_model.generate(
input_ids=inputs.input_ids,
attention_mask=inputs.attention_mask,
num_beams=16,
num_return_sequences=16,
length_penalty=10
)
questions = []
for question in out:
questions.append(loaded_tokenizer.decode(question, skip_special_tokens=True))
for q in questions:
print(q)
# Quand se tient la conférence des Nations Unies sur la diversité biologique?
# Quand a lieu la conférence des Nations Unies sur la diversité biologique?
# Quand se tient la conférence sur la diversité biologique des Nations unies?
# Quand se tient la conférence de la diversité biologique des Nations unies?
# Quand a lieu la conférence sur la diversité biologique des Nations unies?
# Quand a lieu la conférence de la diversité biologique des Nations unies?
# Quand se tient la conférence des Nations unies sur la diversité biologique?
# Quand a lieu la conférence des Nations unies sur la diversité biologique?
# Quand se tient la conférence sur la diversité biologique des Nations Unies?
# Quand se tient la conférence des Nations Unies sur la diversité biologique?
# Quand se tient la conférence de la diversité biologique des Nations Unies?
# Quand la COP15 a-t-elle lieu?
# Quand la COP15 a-t-elle lieu?
# Quand se tient la conférence sur la diversité biologique?
# Quand s'ouvre la COP15,?
# Quand s'ouvre la COP15?
```
## Citation
Model based on:
paper: https://arxiv.org/abs/2010.12321 \
github: https://github.com/moussaKam/BARThez
```
@article{eddine2020barthez,
title={BARThez: a Skilled Pretrained French Sequence-to-Sequence Model},
author={Eddine, Moussa Kamal and Tixier, Antoine J-P and Vazirgiannis, Michalis},
journal={arXiv preprint arXiv:2010.12321},
year={2020}
}
```
| {"language": ["fr"], "license": "mit", "tags": ["seq2seq", "barthez"], "datasets": ["squadFR", "fquad", "piaf"], "metrics": ["bleu", "rouge"], "pipeline_tag": "text2text-generation", "widget": [{"text": "La science des donn\u00e9es est un domaine interdisciplinaire qui utilise des m\u00e9thodes, des processus, des algorithmes et des syst\u00e8mes scientifiques pour extraire des connaissances et des id\u00e9es de nombreuses donn\u00e9es structurelles et non structur\u00e9es.Elle est souvent associ\u00e9e aux <hl>donn\u00e9es massives et \u00e0 l'analyse des donn\u00e9es<hl>."}]} | lincoln/barthez-squadFR-fquad-piaf-question-generation | null | [
"transformers",
"pytorch",
"mbart",
"text2text-generation",
"seq2seq",
"barthez",
"fr",
"dataset:squadFR",
"dataset:fquad",
"dataset:piaf",
"arxiv:2010.12321",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2010.12321"
] | [
"fr"
] | TAGS
#transformers #pytorch #mbart #text2text-generation #seq2seq #barthez #fr #dataset-squadFR #dataset-fquad #dataset-piaf #arxiv-2010.12321 #license-mit #autotrain_compatible #endpoints_compatible #region-us
|
# Génération de question à partir d'un contexte
Le modèle est _fine tuné_ à partir du modèle moussaKam/barthez afin de générer des questions à partir d'un paragraphe et d'une suite de token. La suite de token représente la réponse sur laquelle la question est basée.
Input: _Les projecteurs peuvent être utilisées pour \<hl\>illuminer\<hl\> des terrains de jeu extérieurs_
Output: _À quoi servent les projecteurs sur les terrains de jeu extérieurs?_
## Données d'apprentissage
La base d'entrainement est la concatenation des bases SquadFR, fquad, piaf. L'input est le context et nous avons entouré à l'aide du token spécial \<hl\> les réponses.
Volumétrie (nombre de triplet contexte/réponse/question):
* train: 98 211
* test: 12 277
* valid: 12 776
## Entrainement
L'apprentissage s'est effectué sur une carte Tesla V100.
* Batch size: 20
* Weight decay: 0.01
* Learning rate: 3x10-5 (décroit linéairement)
* < 24h d'entrainement
* Paramètres par défaut de la classe TrainingArguments
* Total steps: 56 000
<img src="data:image/png;base64,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">
La loss represente des "sauts" à cause de la reprise de l'entrainement à deux reprises. Cela induit une modification du learning rate et explique la forme de la courbe.
## Résultats
Les questions générées sont évaluées sur les métrique BLEU et ROUGE. Ce sont des métriques approximative pour la génération de texte.
<img src="data:image/png;base64,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">
<img src="data:image/png;base64,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">
## Tokenizer
Le tokenizer de départ est BarthezTokenizer auquel ont été rajouté les tokens spéciaux \<sep\> et \<hl\>.
## Utilisation
_Le modèle est un POC, nous garantissons pas ses performances_
Model based on:
paper: URL \
github: URL
| [
"# Génération de question à partir d'un contexte\n\nLe modèle est _fine tuné_ à partir du modèle moussaKam/barthez afin de générer des questions à partir d'un paragraphe et d'une suite de token. La suite de token représente la réponse sur laquelle la question est basée.\n\nInput: _Les projecteurs peuvent être utili... | [
"TAGS\n#transformers #pytorch #mbart #text2text-generation #seq2seq #barthez #fr #dataset-squadFR #dataset-fquad #dataset-piaf #arxiv-2010.12321 #license-mit #autotrain_compatible #endpoints_compatible #region-us \n",
"# Génération de question à partir d'un contexte\n\nLe modèle est _fine tuné_ à partir du modèle... |
token-classification | transformers |
# Extraction de réponse
Ce modèle est _fine tuné_ à partir du modèle [camembert-base](https://huggingface.co/camembert-base) pour la tâche de classification de tokens.
L'objectif est d'identifier les suites de tokens probables qui pourrait être l'objet d'une question.
## Données d'apprentissage
La base d'entrainement est la concatenation des bases SquadFR, [fquad](https://huggingface.co/datasets/fquad), [piaf](https://huggingface.co/datasets/piaf).
Les réponses de chaque contexte ont été labelisées avec le label "ANS".
Volumétrie (nombre de contexte):
* train: 24 652
* test: 1 370
* valid: 1 370
## Entrainement
L'apprentissage s'est effectué sur une carte Tesla K80.
* Batch size: 16
* Weight decay: 0.01
* Learning rate: 2x10-5 (décroit linéairement)
* Paramètres par défaut de la classe [TrainingArguments](https://huggingface.co/transformers/main_classes/trainer.html#trainingarguments)
* Total steps: 1 000
Le modèle semble sur apprendre au delà :
## Critiques
Le modèle n'a pas de bonnes performances et doit être corrigé après prédiction pour être cohérent. La tâche de classification n'est pas évidente car le modèle doit identifier des groupes de token _sachant_ qu'une question peut être posée.
## Utilisation
_Le modèle est un POC, nous garantissons pas ses performances_
```python
from transformers import AutoTokenizer, AutoModelForTokenClassification
import numpy as np
model_name = "lincoln/camembert-squadFR-fquad-piaf-answer-extraction"
loaded_tokenizer = AutoTokenizer.from_pretrained(model_path)
loaded_model = AutoModelForTokenClassification.from_pretrained(model_path)
text = "La science des données est un domaine interdisciplinaire qui utilise des méthodes, des processus,\
des algorithmes et des systèmes scientifiques pour extraire des connaissances et des idées de nombreuses données structurelles et non structurées.\
Elle est souvent associée aux données massives et à l'analyse des données."
inputs = loaded_tokenizer(text, return_tensors="pt", return_offsets_mapping=True)
outputs = loaded_model(inputs.input_ids).logits
probs = 1 / (1 + np.exp(-outputs.detach().numpy()))
probs[:, :, 1][0] = np.convolve(probs[:, :, 1][0], np.ones(2), 'same') / 2
sentences = loaded_tokenizer.tokenize(text, add_special_tokens=False)
prob_answer_tokens = probs[:, 1:-1, 1].flatten().tolist()
offset_start_mapping = inputs.offset_mapping[:, 1:-1, 0].flatten().tolist()
offset_end_mapping = inputs.offset_mapping[:, 1:-1, 1].flatten().tolist()
threshold = 0.4
entities = []
for ix, (token, prob_ans, offset_start, offset_end) in enumerate(zip(sentences, prob_answer_tokens, offset_start_mapping, offset_end_mapping)):
entities.append({
'entity': 'ANS' if prob_ans > threshold else 'O',
'score': prob_ans,
'index': ix,
'word': token,
'start': offset_start,
'end': offset_end
})
for p in entities:
print(p)
# {'entity': 'O', 'score': 0.3118681311607361, 'index': 0, 'word': '▁La', 'start': 0, 'end': 2}
# {'entity': 'O', 'score': 0.37866950035095215, 'index': 1, 'word': '▁science', 'start': 3, 'end': 10}
# {'entity': 'ANS', 'score': 0.45018652081489563, 'index': 2, 'word': '▁des', 'start': 11, 'end': 14}
# {'entity': 'ANS', 'score': 0.4615934491157532, 'index': 3, 'word': '▁données', 'start': 15, 'end': 22}
# {'entity': 'O', 'score': 0.35033443570137024, 'index': 4, 'word': '▁est', 'start': 23, 'end': 26}
# {'entity': 'O', 'score': 0.24779987335205078, 'index': 5, 'word': '▁un', 'start': 27, 'end': 29}
# {'entity': 'O', 'score': 0.27084410190582275, 'index': 6, 'word': '▁domaine', 'start': 30, 'end': 37}
# {'entity': 'O', 'score': 0.3259460926055908, 'index': 7, 'word': '▁in', 'start': 38, 'end': 40}
# {'entity': 'O', 'score': 0.371802419424057, 'index': 8, 'word': 'terdisciplinaire', 'start': 40, 'end': 56}
# {'entity': 'O', 'score': 0.3140853941440582, 'index': 9, 'word': '▁qui', 'start': 57, 'end': 60}
# {'entity': 'O', 'score': 0.2629334330558777, 'index': 10, 'word': '▁utilise', 'start': 61, 'end': 68}
# {'entity': 'O', 'score': 0.2968383729457855, 'index': 11, 'word': '▁des', 'start': 69, 'end': 72}
# {'entity': 'O', 'score': 0.33898216485977173, 'index': 12, 'word': '▁méthodes', 'start': 73, 'end': 81}
# {'entity': 'O', 'score': 0.3776060938835144, 'index': 13, 'word': ',', 'start': 81, 'end': 82}
# {'entity': 'O', 'score': 0.3710060119628906, 'index': 14, 'word': '▁des', 'start': 83, 'end': 86}
# {'entity': 'O', 'score': 0.35908180475234985, 'index': 15, 'word': '▁processus', 'start': 87, 'end': 96}
# {'entity': 'O', 'score': 0.3890596628189087, 'index': 16, 'word': ',', 'start': 96, 'end': 97}
# {'entity': 'O', 'score': 0.38341325521469116, 'index': 17, 'word': '▁des', 'start': 101, 'end': 104}
# {'entity': 'O', 'score': 0.3743852376937866, 'index': 18, 'word': '▁', 'start': 105, 'end': 106}
# {'entity': 'O', 'score': 0.3943936228752136, 'index': 19, 'word': 'algorithme', 'start': 105, 'end': 115}
# {'entity': 'O', 'score': 0.39456743001937866, 'index': 20, 'word': 's', 'start': 115, 'end': 116}
# {'entity': 'O', 'score': 0.3846966624259949, 'index': 21, 'word': '▁et', 'start': 117, 'end': 119}
# {'entity': 'O', 'score': 0.367380827665329, 'index': 22, 'word': '▁des', 'start': 120, 'end': 123}
# {'entity': 'O', 'score': 0.3652925491333008, 'index': 23, 'word': '▁systèmes', 'start': 124, 'end': 132}
# {'entity': 'O', 'score': 0.3975735306739807, 'index': 24, 'word': '▁scientifiques', 'start': 133, 'end': 146}
# {'entity': 'O', 'score': 0.36417365074157715, 'index': 25, 'word': '▁pour', 'start': 147, 'end': 151}
# {'entity': 'O', 'score': 0.32438698410987854, 'index': 26, 'word': '▁extraire', 'start': 152, 'end': 160}
# {'entity': 'O', 'score': 0.3416857123374939, 'index': 27, 'word': '▁des', 'start': 161, 'end': 164}
# {'entity': 'O', 'score': 0.3674810230731964, 'index': 28, 'word': '▁connaissances', 'start': 165, 'end': 178}
# {'entity': 'O', 'score': 0.38362061977386475, 'index': 29, 'word': '▁et', 'start': 179, 'end': 181}
# {'entity': 'O', 'score': 0.364640474319458, 'index': 30, 'word': '▁des', 'start': 182, 'end': 185}
# {'entity': 'O', 'score': 0.36050117015838623, 'index': 31, 'word': '▁idées', 'start': 186, 'end': 191}
# {'entity': 'O', 'score': 0.3768993020057678, 'index': 32, 'word': '▁de', 'start': 192, 'end': 194}
# {'entity': 'O', 'score': 0.39184248447418213, 'index': 33, 'word': '▁nombreuses', 'start': 195, 'end': 205}
# {'entity': 'ANS', 'score': 0.4091200828552246, 'index': 34, 'word': '▁données', 'start': 206, 'end': 213}
# {'entity': 'ANS', 'score': 0.41234123706817627, 'index': 35, 'word': '▁structurelle', 'start': 214, 'end': 226}
# {'entity': 'ANS', 'score': 0.40243157744407654, 'index': 36, 'word': 's', 'start': 226, 'end': 227}
# {'entity': 'ANS', 'score': 0.4007353186607361, 'index': 37, 'word': '▁et', 'start': 228, 'end': 230}
# {'entity': 'ANS', 'score': 0.40597623586654663, 'index': 38, 'word': '▁non', 'start': 231, 'end': 234}
# {'entity': 'ANS', 'score': 0.40272021293640137, 'index': 39, 'word': '▁structurée', 'start': 235, 'end': 245}
# {'entity': 'O', 'score': 0.392631471157074, 'index': 40, 'word': 's', 'start': 245, 'end': 246}
# {'entity': 'O', 'score': 0.34266412258148193, 'index': 41, 'word': '.', 'start': 246, 'end': 247}
# {'entity': 'O', 'score': 0.26178646087646484, 'index': 42, 'word': '▁Elle', 'start': 255, 'end': 259}
# {'entity': 'O', 'score': 0.2265639454126358, 'index': 43, 'word': '▁est', 'start': 260, 'end': 263}
# {'entity': 'O', 'score': 0.22844195365905762, 'index': 44, 'word': '▁souvent', 'start': 264, 'end': 271}
# {'entity': 'O', 'score': 0.2475772500038147, 'index': 45, 'word': '▁associée', 'start': 272, 'end': 280}
# {'entity': 'O', 'score': 0.3002186715602875, 'index': 46, 'word': '▁aux', 'start': 281, 'end': 284}
# {'entity': 'O', 'score': 0.3875720798969269, 'index': 47, 'word': '▁données', 'start': 285, 'end': 292}
# {'entity': 'ANS', 'score': 0.445063054561615, 'index': 48, 'word': '▁massive', 'start': 293, 'end': 300}
# {'entity': 'ANS', 'score': 0.4419114589691162, 'index': 49, 'word': 's', 'start': 300, 'end': 301}
# {'entity': 'ANS', 'score': 0.4240635633468628, 'index': 50, 'word': '▁et', 'start': 302, 'end': 304}
# {'entity': 'O', 'score': 0.3900952935218811, 'index': 51, 'word': '▁à', 'start': 305, 'end': 306}
# {'entity': 'O', 'score': 0.3784807324409485, 'index': 52, 'word': '▁l', 'start': 307, 'end': 308}
# {'entity': 'O', 'score': 0.3459452986717224, 'index': 53, 'word': "'", 'start': 308, 'end': 309}
# {'entity': 'O', 'score': 0.37636008858680725, 'index': 54, 'word': 'analyse', 'start': 309, 'end': 316}
# {'entity': 'ANS', 'score': 0.4475618302822113, 'index': 55, 'word': '▁des', 'start': 317, 'end': 320}
# {'entity': 'ANS', 'score': 0.43845775723457336, 'index': 56, 'word': '▁données', 'start': 321, 'end': 328}
# {'entity': 'O', 'score': 0.3761221170425415, 'index': 57, 'word': '.', 'start': 328, 'end': 329}
```
| {"language": ["fr"], "license": "mit", "tags": ["camembert", "answer extraction"], "datasets": ["squadFR", "fquad", "piaf"]} | lincoln/camembert-squadFR-fquad-piaf-answer-extraction | null | [
"transformers",
"pytorch",
"camembert",
"token-classification",
"answer extraction",
"fr",
"dataset:squadFR",
"dataset:fquad",
"dataset:piaf",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"fr"
] | TAGS
#transformers #pytorch #camembert #token-classification #answer extraction #fr #dataset-squadFR #dataset-fquad #dataset-piaf #license-mit #autotrain_compatible #endpoints_compatible #region-us
|
# Extraction de réponse
Ce modèle est _fine tuné_ à partir du modèle camembert-base pour la tâche de classification de tokens.
L'objectif est d'identifier les suites de tokens probables qui pourrait être l'objet d'une question.
## Données d'apprentissage
La base d'entrainement est la concatenation des bases SquadFR, fquad, piaf.
Les réponses de chaque contexte ont été labelisées avec le label "ANS".
Volumétrie (nombre de contexte):
* train: 24 652
* test: 1 370
* valid: 1 370
## Entrainement
L'apprentissage s'est effectué sur une carte Tesla K80.
* Batch size: 16
* Weight decay: 0.01
* Learning rate: 2x10-5 (décroit linéairement)
* Paramètres par défaut de la classe TrainingArguments
* Total steps: 1 000
Le modèle semble sur apprendre au delà :
!Loss
## Critiques
Le modèle n'a pas de bonnes performances et doit être corrigé après prédiction pour être cohérent. La tâche de classification n'est pas évidente car le modèle doit identifier des groupes de token _sachant_ qu'une question peut être posée.
!Performances
## Utilisation
_Le modèle est un POC, nous garantissons pas ses performances_
| [
"# Extraction de réponse\n\nCe modèle est _fine tuné_ à partir du modèle camembert-base pour la tâche de classification de tokens. \nL'objectif est d'identifier les suites de tokens probables qui pourrait être l'objet d'une question.",
"## Données d'apprentissage\n\nLa base d'entrainement est la concatenation des... | [
"TAGS\n#transformers #pytorch #camembert #token-classification #answer extraction #fr #dataset-squadFR #dataset-fquad #dataset-piaf #license-mit #autotrain_compatible #endpoints_compatible #region-us \n",
"# Extraction de réponse\n\nCe modèle est _fine tuné_ à partir du modèle camembert-base pour la tâche de clas... |
text-classification | transformers |
# Classification d'articles de presses avec Flaubert
Ce modèle se base sur le modèle [`flaubert/flaubert_base_cased`](https://huggingface.co/flaubert/flaubert_base_cased) et à été fine-tuné en utilisant des articles de presse issus de la base de données MLSUM.
Dans leur papier, les équipes de reciTAL et de la Sorbonne ont proposé comme ouverture de réaliser un modèle de détection de topic sur les articles de presse.
Les topics ont été extrait à partir des URL et nous avons effectué une étape de regroupement de topics pour éliminer ceux avec un trop faible volume et ceux qui paraissaient redondants.
Nous avons finalement utilisé la liste de topics avec les regroupements suivants:
* __Economie__: economie, argent, emploi, entreprises, economie-francaise, immobilier, crise-financiere, evasion-fiscale, economie-mondiale, m-voiture, smart-cities, automobile, logement, flottes-d-entreprise, import, crise-de-l-euro, guide-des-impots, le-club-de-l-economie, telephonie-mobile
* __Opinion__: idees, les-decodeurs, tribunes
* __Politique__: politique, election-presidentielle-2012, election-presidentielle-2017, elections-americaines, municipales, referendum-sur-le-brexit, elections-legislatives-2017, elections-regionales, donald-trump, elections-regionales-2015, europeennes-2014, elections-cantonales-2011, primaire-parti-socialiste, gouvernement-philippe, elections-departementales-2015, chroniques-de-la-presidence-trump, primaire-de-la-gauche, la-republique-en-marche, elections-americaines-mi-mandat-2018, elections, elections-italiennes, elections-senatoriales
* __Societe__: societe, sante, attaques-a-paris, immigration-et-diversite, religions, medecine, francaises-francais, mobilite
* __Culture__: televisions-radio, musiques, festival, arts, scenes, festival-de-cannes, mode, bande-dessinee, architecture, vins, photo, m-mode, fashion-week, les-recettes-du-monde, tele-zapping, critique-litteraire, festival-d-avignon, m-gastronomie-le-lieu, les-enfants-akira, gastronomie, culture, livres, cinema, actualite-medias, blog, m-gastronomie
* __Sport__: sport, football, jeux-olympiques, ligue-1, tennis, coupe-du-monde, mondial-2018, rugby, euro-2016, jeux-olympiques-rio-2016, cyclisme, ligue-des-champions, basket, roland-garros, athletisme, tour-de-france, euro2012, jeux-olympiques-pyeongchang-2018, coupe-du-monde-rugby, formule-1, voile, top-14, ski, handball, sports-mecaniques, sports-de-combat, blog-du-tour-de-france, sport-et-societe, sports-de-glisse, tournoi-des-6-nations
* __Environement__: planete, climat, biodiversite, pollution, energies, cop21
* __Technologie__: pixels, technologies, sciences, cosmos, la-france-connectee, trajectoires-digitales
* __Education__: campus, education, bac-lycee, enseignement-superieur, ecole-primaire-et-secondaire, o21, orientation-scolaire, brevet-college
* __Justice__: police-justice, panama-papers, affaire-penelope-fillon, documents-wikileaks, enquetes, paradise-papers
Les thèmes ayant moins de 100 articles n'ont pas été pris en compte.
Nous avons également mis de côté les articles faisant référence à des topics geographiques, ce qui a donné lieu à un nouveau modèle de classification.
Après nettoyage, la base MLSUM a été réduite à 293 995 articles. Le corps d'un article en moyenne comporte 694 tokens.
Nous avons entrainé le modèle sur 20% de la base nettoyée. En moyenne, le nombre d'articles par classe est de ~4K.
## Entrainement
Nous avons benchmarké différents modèles en les entrainant sur différentes parties des articles (titre, résumé, corps et titre+résumé) et avec des échantillons d'apprentissage de tailles différentes.
Les modèles ont été entrainé sur le cloud Azure avec des Tesla V100.
## Modèle
Le modèle partagé sur HF est le modéle qui prend en entrée le corps d'un article. Nous l'avons entrainé sur 20% du jeu de donnée nettoyé.
## Résulats
*Les lignes correspondent aux labels prédits et les colonnes aux véritables topics. Les pourcentages sont calculés sur les colonnes.*
_Nous garantissons pas les résultats sur le long terme. Modèle réalisé dans le cadre d'un POC._
## Utilisation
```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from transformers import TextClassificationPipeline
model_name = 'lincoln/flaubert-mlsum-topic-classification'
loaded_tokenizer = AutoTokenizer.from_pretrained(model_name)
loaded_model = AutoModelForSequenceClassification.from_pretrained(model_name)
nlp = TextClassificationPipeline(model=loaded_model, tokenizer=loaded_tokenizer)
nlp("Le Bayern Munich prend la grenadine.", truncation=True)
```
## Citation
```bibtex
@article{scialom2020mlsum,
title={MLSUM: The Multilingual Summarization Corpus},
author={Thomas Scialom and Paul-Alexis Dray and Sylvain Lamprier and Benjamin Piwowarski and Jacopo Staiano},
year={2020},
eprint={2004.14900},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["fr"], "license": "mit", "tags": ["text-classification", "flaubert"], "datasets": ["MLSUM"], "pipeline_tag": "text-classification", "widget": [{"text": "La bourse de paris en forte baisse apr\u00e8s que des canards ont envahit le parlement."}]} | lincoln/flaubert-mlsum-topic-classification | null | [
"transformers",
"pytorch",
"tf",
"flaubert",
"text-classification",
"fr",
"dataset:MLSUM",
"arxiv:2004.14900",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2004.14900"
] | [
"fr"
] | TAGS
#transformers #pytorch #tf #flaubert #text-classification #fr #dataset-MLSUM #arxiv-2004.14900 #license-mit #autotrain_compatible #endpoints_compatible #has_space #region-us
|
# Classification d'articles de presses avec Flaubert
Ce modèle se base sur le modèle 'flaubert/flaubert_base_cased' et à été fine-tuné en utilisant des articles de presse issus de la base de données MLSUM.
Dans leur papier, les équipes de reciTAL et de la Sorbonne ont proposé comme ouverture de réaliser un modèle de détection de topic sur les articles de presse.
Les topics ont été extrait à partir des URL et nous avons effectué une étape de regroupement de topics pour éliminer ceux avec un trop faible volume et ceux qui paraissaient redondants.
Nous avons finalement utilisé la liste de topics avec les regroupements suivants:
* __Economie__: economie, argent, emploi, entreprises, economie-francaise, immobilier, crise-financiere, evasion-fiscale, economie-mondiale, m-voiture, smart-cities, automobile, logement, flottes-d-entreprise, import, crise-de-l-euro, guide-des-impots, le-club-de-l-economie, telephonie-mobile
* __Opinion__: idees, les-decodeurs, tribunes
* __Politique__: politique, election-presidentielle-2012, election-presidentielle-2017, elections-americaines, municipales, referendum-sur-le-brexit, elections-legislatives-2017, elections-regionales, donald-trump, elections-regionales-2015, europeennes-2014, elections-cantonales-2011, primaire-parti-socialiste, gouvernement-philippe, elections-departementales-2015, chroniques-de-la-presidence-trump, primaire-de-la-gauche, la-republique-en-marche, elections-americaines-mi-mandat-2018, elections, elections-italiennes, elections-senatoriales
* __Societe__: societe, sante, attaques-a-paris, immigration-et-diversite, religions, medecine, francaises-francais, mobilite
* __Culture__: televisions-radio, musiques, festival, arts, scenes, festival-de-cannes, mode, bande-dessinee, architecture, vins, photo, m-mode, fashion-week, les-recettes-du-monde, tele-zapping, critique-litteraire, festival-d-avignon, m-gastronomie-le-lieu, les-enfants-akira, gastronomie, culture, livres, cinema, actualite-medias, blog, m-gastronomie
* __Sport__: sport, football, jeux-olympiques, ligue-1, tennis, coupe-du-monde, mondial-2018, rugby, euro-2016, jeux-olympiques-rio-2016, cyclisme, ligue-des-champions, basket, roland-garros, athletisme, tour-de-france, euro2012, jeux-olympiques-pyeongchang-2018, coupe-du-monde-rugby, formule-1, voile, top-14, ski, handball, sports-mecaniques, sports-de-combat, blog-du-tour-de-france, sport-et-societe, sports-de-glisse, tournoi-des-6-nations
* __Environement__: planete, climat, biodiversite, pollution, energies, cop21
* __Technologie__: pixels, technologies, sciences, cosmos, la-france-connectee, trajectoires-digitales
* __Education__: campus, education, bac-lycee, enseignement-superieur, ecole-primaire-et-secondaire, o21, orientation-scolaire, brevet-college
* __Justice__: police-justice, panama-papers, affaire-penelope-fillon, documents-wikileaks, enquetes, paradise-papers
Les thèmes ayant moins de 100 articles n'ont pas été pris en compte.
Nous avons également mis de côté les articles faisant référence à des topics geographiques, ce qui a donné lieu à un nouveau modèle de classification.
Après nettoyage, la base MLSUM a été réduite à 293 995 articles. Le corps d'un article en moyenne comporte 694 tokens.
Nous avons entrainé le modèle sur 20% de la base nettoyée. En moyenne, le nombre d'articles par classe est de ~4K.
## Entrainement
Nous avons benchmarké différents modèles en les entrainant sur différentes parties des articles (titre, résumé, corps et titre+résumé) et avec des échantillons d'apprentissage de tailles différentes.
!Performance
Les modèles ont été entrainé sur le cloud Azure avec des Tesla V100.
## Modèle
Le modèle partagé sur HF est le modéle qui prend en entrée le corps d'un article. Nous l'avons entrainé sur 20% du jeu de donnée nettoyé.
## Résulats
!Matrice de confusion
*Les lignes correspondent aux labels prédits et les colonnes aux véritables topics. Les pourcentages sont calculés sur les colonnes.*
_Nous garantissons pas les résultats sur le long terme. Modèle réalisé dans le cadre d'un POC._
## Utilisation
| [
"# Classification d'articles de presses avec Flaubert\n\nCe modèle se base sur le modèle 'flaubert/flaubert_base_cased' et à été fine-tuné en utilisant des articles de presse issus de la base de données MLSUM. \nDans leur papier, les équipes de reciTAL et de la Sorbonne ont proposé comme ouverture de réaliser un m... | [
"TAGS\n#transformers #pytorch #tf #flaubert #text-classification #fr #dataset-MLSUM #arxiv-2004.14900 #license-mit #autotrain_compatible #endpoints_compatible #has_space #region-us \n",
"# Classification d'articles de presses avec Flaubert\n\nCe modèle se base sur le modèle 'flaubert/flaubert_base_cased' et à été... |
summarization | transformers |
# Résumé automatique d'article de presses
Ce modèles est basé sur le modèle [`facebook/mbart-large-50`](https://huggingface.co/facebook/mbart-large-50) et été fine-tuné en utilisant des articles de presse issus de la base de données MLSUM. L'hypothèse à été faite que les chapeaux des articles faisaient de bon résumés de référence.
## Entrainement
Nous avons testé deux architecture de modèles (T5 et BART) avec des textes en entrée de 512 ou 1024 tokens. Finallement c'est le modèle BART avec 512 tokens qui à été retenu.
Il a été entrainé sur 2 epochs (~700K articles) sur une Tesla V100 (32 heures d'entrainement).
## Résultats
Nous avons comparé notre modèle (`mbart-large-512-full` sur le graphique) à deux références:
* MBERT qui correspond aux performances du modèle entrainé par l'équipe à l'origine de la base d'articles MLSUM
* Barthez qui est un autre modèle basé sur des articles de presses issus de la base de données OrangeSum
On voit que le score de novelty (cf papier MLSUM) de notre modèle n'est pas encore comparable à ces deux références et encore moins à une production humaine néanmoins les résumés générés sont dans l'ensemble de bonne qualité.
## Utilisation
```python
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
from transformers import SummarizationPipeline
model_name = 'lincoln/mbart-mlsum-automatic-summarization'
loaded_tokenizer = AutoTokenizer.from_pretrained(model_name)
loaded_model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
nlp = SummarizationPipeline(model=loaded_model, tokenizer=loaded_tokenizer)
nlp("""
« La veille de l’ouverture, je vais faire venir un coach pour les salariés qui reprendront le travail.
Cela va me coûter 300 euros, mais après des mois d’oisiveté obligatoire, la reprise n’est pas simple.
Certains sont au chômage partiel depuis mars 2020 », raconte Alain Fontaine, propriétaire du restaurant Le Mesturet,
dans le quartier de la Bourse, à Paris. Cette date d’ouverture, désormais, il la connaît. Emmanuel Macron a, en effet,
donné le feu vert pour un premier accueil des clients en terrasse, mercredi 19 mai. M. Fontaine imagine même faire venir un orchestre ce jour-là pour fêter l’événement.
Il lui reste toutefois à construire sa terrasse. Il pensait que les ouvriers passeraient samedi 1er mai pour l’installer, mais, finalement, le rendez-vous a été décalé.
Pour l’instant, le tas de bois est entreposé dans la salle de restaurant qui n’a plus accueilli de convives depuis le 29 octobre 2020,
quand le couperet de la fermeture administrative est tombé.M. Fontaine, président de l’Association française des maîtres restaurateurs,
ne manquera pas de concurrents prêts à profiter de ce premier temps de réouverture des bars et restaurants. Même si le couvre-feu limite le service à 21 heures.
D’autant que la Mairie de Paris vient d’annoncer le renouvellement des terrasses éphémères installées en 2020 et leur gratuité jusqu’à la fin de l’été.
""")
```
## Citation
```bibtex
@article{scialom2020mlsum,
title={MLSUM: The Multilingual Summarization Corpus},
author={Thomas Scialom and Paul-Alexis Dray and Sylvain Lamprier and Benjamin Piwowarski and Jacopo Staiano},
year={2020},
eprint={2004.14900},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` | {"language": ["fr"], "license": "mit", "tags": ["summarization", "mbart", "bart"], "datasets": ["MLSUM"], "pipeline_tag": "summarization", "widget": [{"text": "\u00ab La veille de l\u2019ouverture, je vais faire venir un coach pour les salari\u00e9s qui reprendront le travail. Cela va me co\u00fbter 300 euros, mais apr\u00e8s des mois d\u2019oisivet\u00e9 obligatoire, la reprise n\u2019est pas simple. Certains sont au ch\u00f4mage partiel depuis mars 2020 \u00bb, raconte Alain Fontaine, propri\u00e9taire du restaurant Le Mesturet, dans le quartier de la Bourse, \u00e0 Paris. Cette date d\u2019ouverture, d\u00e9sormais, il la conna\u00eet. Emmanuel Macron a, en effet, donn\u00e9 le feu vert pour un premier accueil des clients en terrasse, mercredi 19 mai. M. Fontaine imagine m\u00eame faire venir un orchestre ce jour-l\u00e0 pour f\u00eater l\u2019\u00e9v\u00e9nement. Il lui reste toutefois \u00e0 construire sa terrasse. Il pensait que les ouvriers passeraient samedi 1er mai pour l\u2019installer, mais, finalement, le rendez-vous a \u00e9t\u00e9 d\u00e9cal\u00e9. Pour l\u2019instant, le tas de bois est entrepos\u00e9 dans la salle de restaurant qui n\u2019a plus accueilli de convives depuis le 29 octobre 2020, quand le couperet de la fermeture administrative est tomb\u00e9.M. Fontaine, pr\u00e9sident de l\u2019Association fran\u00e7aise des ma\u00eetres restaurateurs, ne manquera pas de concurrents pr\u00eats \u00e0 profiter de ce premier temps de r\u00e9ouverture des bars et restaurants. M\u00eame si le couvre-feu limite le service \u00e0 21 heures. D\u2019autant que la Mairie de Paris vient d\u2019annoncer le renouvellement des terrasses \u00e9ph\u00e9m\u00e8res install\u00e9es en 2020 et leur gratuit\u00e9 jusqu\u2019\u00e0 la fin de l\u2019\u00e9t\u00e9."}]} | lincoln/mbart-mlsum-automatic-summarization | null | [
"transformers",
"pytorch",
"tf",
"mbart",
"text2text-generation",
"summarization",
"bart",
"fr",
"dataset:MLSUM",
"arxiv:2004.14900",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2004.14900"
] | [
"fr"
] | TAGS
#transformers #pytorch #tf #mbart #text2text-generation #summarization #bart #fr #dataset-MLSUM #arxiv-2004.14900 #license-mit #autotrain_compatible #endpoints_compatible #region-us
|
# Résumé automatique d'article de presses
Ce modèles est basé sur le modèle 'facebook/mbart-large-50' et été fine-tuné en utilisant des articles de presse issus de la base de données MLSUM. L'hypothèse à été faite que les chapeaux des articles faisaient de bon résumés de référence.
## Entrainement
Nous avons testé deux architecture de modèles (T5 et BART) avec des textes en entrée de 512 ou 1024 tokens. Finallement c'est le modèle BART avec 512 tokens qui à été retenu.
Il a été entrainé sur 2 epochs (~700K articles) sur une Tesla V100 (32 heures d'entrainement).
## Résultats
!Score de novelty
Nous avons comparé notre modèle ('mbart-large-512-full' sur le graphique) à deux références:
* MBERT qui correspond aux performances du modèle entrainé par l'équipe à l'origine de la base d'articles MLSUM
* Barthez qui est un autre modèle basé sur des articles de presses issus de la base de données OrangeSum
On voit que le score de novelty (cf papier MLSUM) de notre modèle n'est pas encore comparable à ces deux références et encore moins à une production humaine néanmoins les résumés générés sont dans l'ensemble de bonne qualité.
## Utilisation
| [
"# Résumé automatique d'article de presses\n\nCe modèles est basé sur le modèle 'facebook/mbart-large-50' et été fine-tuné en utilisant des articles de presse issus de la base de données MLSUM. L'hypothèse à été faite que les chapeaux des articles faisaient de bon résumés de référence.",
"## Entrainement\n\nNous ... | [
"TAGS\n#transformers #pytorch #tf #mbart #text2text-generation #summarization #bart #fr #dataset-MLSUM #arxiv-2004.14900 #license-mit #autotrain_compatible #endpoints_compatible #region-us \n",
"# Résumé automatique d'article de presses\n\nCe modèles est basé sur le modèle 'facebook/mbart-large-50' et été fine-tu... |
summarization | transformers |
## `bart-large-samsum`
This model was trained using Microsoft's [`Azure Machine Learning Service`](https://azure.microsoft.com/en-us/services/machine-learning). It was fine-tuned on the [`samsum`](https://huggingface.co/datasets/samsum) corpus from [`facebook/bart-large`](https://huggingface.co/facebook/bart-large) checkpoint.
## Usage (Inference)
```python
from transformers import pipeline
summarizer = pipeline("summarization", model="linydub/bart-large-samsum")
input_text = '''
Henry: Hey, is Nate coming over to watch the movie tonight?
Kevin: Yea, he said he'll be arriving a bit later at around 7 since he gets off of work at 6. Have you taken out the garbage yet?
Henry: Oh I forgot. I'll do that once I'm finished with my assignment for my math class.
Kevin: Yea, you should take it out as soon as possible. And also, Nate is bringing his girlfriend.
Henry: Nice, I'm really looking forward to seeing them again.
'''
summarizer(input_text)
```
## Fine-tune on AzureML
[](https://portal.azure.com/#create/Microsoft.Template/uri/https%3A%2F%2Fraw.githubusercontent.com%2Flinydub%2Fazureml-greenai-txtsum%2Fmain%2F.cloud%2Ftemplate-hub%2Flinydub%2Farm-bart-large-samsum.json) [](http://armviz.io/#/?load=https://raw.githubusercontent.com/linydub/azureml-greenai-txtsum/main/.cloud/template-hub/linydub/arm-bart-large-samsum.json)
More information about the fine-tuning process (including samples and benchmarks):
**[Preview]** https://github.com/linydub/azureml-greenai-txtsum
## Resource Usage
These results were retrieved from [`Azure Monitor Metrics`](https://docs.microsoft.com/en-us/azure/azure-monitor/essentials/data-platform-metrics). All experiments were ran on AzureML low priority compute clusters.
| Key | Value |
| --- | ----- |
| Region | US West 2 |
| AzureML Compute SKU | STANDARD_ND40RS_V2 |
| Compute SKU GPU Device | 8 x NVIDIA V100 32GB (NVLink) |
| Compute Node Count | 1 |
| Run Duration | 6m 48s |
| Compute Cost (Dedicated/LowPriority) | $2.50 / $0.50 USD |
| Average CPU Utilization | 47.9% |
| Average GPU Utilization | 69.8% |
| Average GPU Memory Usage | 25.71 GB |
| Total GPU Energy Usage | 370.84 kJ |
*Compute cost ($) is estimated from the run duration, number of compute nodes utilized, and SKU's price per hour. Updated SKU pricing could be found [here](https://azure.microsoft.com/en-us/pricing/details/machine-learning).
### Carbon Emissions
These results were obtained using [`CodeCarbon`](https://github.com/mlco2/codecarbon). The carbon emissions are estimated from training runtime only (excl. setup and evaluation runtimes).
| Key | Value |
| --- | ----- |
| timestamp | 2021-09-16T23:54:25 |
| duration | 263.2430217266083 |
| emissions | 0.029715544634717518 |
| energy_consumed | 0.09985062041235725 |
| country_name | USA |
| region | Washington |
| cloud_provider | azure |
| cloud_region | westus2 |
## Hyperparameters
- max_source_length: 512
- max_target_length: 90
- fp16: True
- seed: 1
- per_device_train_batch_size: 16
- per_device_eval_batch_size: 16
- gradient_accumulation_steps: 1
- learning_rate: 5e-5
- num_train_epochs: 3.0
- weight_decay: 0.1
## Results
| ROUGE | Score |
| ----- | ----- |
| eval_rouge1 | 55.0234 |
| eval_rouge2 | 29.6005 |
| eval_rougeL | 44.914 |
| eval_rougeLsum | 50.464 |
| predict_rouge1 | 53.4345 |
| predict_rouge2 | 28.7445 |
| predict_rougeL | 44.1848 |
| predict_rougeLsum | 49.1874 |
| Metric | Value |
| ------ | ----- |
| epoch | 3.0 |
| eval_gen_len | 30.6027 |
| eval_loss | 1.4327096939086914 |
| eval_runtime | 22.9127 |
| eval_samples | 818 |
| eval_samples_per_second | 35.701 |
| eval_steps_per_second | 0.306 |
| predict_gen_len | 30.4835 |
| predict_loss | 1.4501988887786865 |
| predict_runtime | 26.0269 |
| predict_samples | 819 |
| predict_samples_per_second | 31.467 |
| predict_steps_per_second | 0.269 |
| train_loss | 1.2014821151207233 |
| train_runtime | 263.3678 |
| train_samples | 14732 |
| train_samples_per_second | 167.811 |
| train_steps_per_second | 1.321 |
| total_steps | 348 |
| total_flops | 4.26008990669865e+16 |
| {"language": ["en"], "license": "apache-2.0", "tags": ["summarization", "azureml", "azure", "codecarbon", "bart"], "datasets": ["samsum"], "metrics": ["rouge"], "widget": [{"text": "Henry: Hey, is Nate coming over to watch the movie tonight?\nKevin: Yea, he said he'll be arriving a bit later at around 7 since he gets off of work at 6. Have you taken out the garbage yet?\nHenry: Oh I forgot. I'll do that once I'm finished with my assignment for my math class.\nKevin: Yea, you should take it out as soon as possible. And also, Nate is bringing his girlfriend.\nHenry: Nice, I'm really looking forward to seeing them again."}], "model-index": [{"name": "bart-large-samsum", "results": [{"task": {"type": "abstractive-text-summarization", "name": "Abstractive Text Summarization"}, "dataset": {"name": "SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization", "type": "samsum"}, "metrics": [{"type": "rouge-1", "value": 55.0234, "name": "Validation ROGUE-1"}, {"type": "rouge-2", "value": 29.6005, "name": "Validation ROGUE-2"}, {"type": "rouge-L", "value": 44.914, "name": "Validation ROGUE-L"}, {"type": "rouge-Lsum", "value": 50.464, "name": "Validation ROGUE-Lsum"}, {"type": "rouge-1", "value": 53.4345, "name": "Test ROGUE-1"}, {"type": "rouge-2", "value": 28.7445, "name": "Test ROGUE-2"}, {"type": "rouge-L", "value": 44.1848, "name": "Test ROGUE-L"}, {"type": "rouge-Lsum", "value": 49.1874, "name": "Test ROGUE-Lsum"}]}]}]} | linydub/bart-large-samsum | null | [
"transformers",
"pytorch",
"tensorboard",
"bart",
"text2text-generation",
"summarization",
"azureml",
"azure",
"codecarbon",
"en",
"dataset:samsum",
"license:apache-2.0",
"model-index",
"autotrain_compatible",
"endpoints_compatible",
"has_space",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"en"
] | TAGS
#transformers #pytorch #tensorboard #bart #text2text-generation #summarization #azureml #azure #codecarbon #en #dataset-samsum #license-apache-2.0 #model-index #autotrain_compatible #endpoints_compatible #has_space #region-us
| 'bart-large-samsum'
-------------------
This model was trained using Microsoft's 'Azure Machine Learning Service'. It was fine-tuned on the 'samsum' corpus from 'facebook/bart-large' checkpoint.
Usage (Inference)
-----------------
Fine-tune on AzureML
--------------------
:
[Preview] URL
Resource Usage
--------------
These results were retrieved from 'Azure Monitor Metrics'. All experiments were ran on AzureML low priority compute clusters.
\*Compute cost ($) is estimated from the run duration, number of compute nodes utilized, and SKU's price per hour. Updated SKU pricing could be found here.
### Carbon Emissions
These results were obtained using 'CodeCarbon'. The carbon emissions are estimated from training runtime only (excl. setup and evaluation runtimes).
Hyperparameters
---------------
* max\_source\_length: 512
* max\_target\_length: 90
* fp16: True
* seed: 1
* per\_device\_train\_batch\_size: 16
* per\_device\_eval\_batch\_size: 16
* gradient\_accumulation\_steps: 1
* learning\_rate: 5e-5
* num\_train\_epochs: 3.0
* weight\_decay: 0.1
Results
-------
| [
"### Carbon Emissions\n\n\nThese results were obtained using 'CodeCarbon'. The carbon emissions are estimated from training runtime only (excl. setup and evaluation runtimes).\n\n\n\nHyperparameters\n---------------\n\n\n* max\\_source\\_length: 512\n* max\\_target\\_length: 90\n* fp16: True\n* seed: 1\n* per\\_dev... | [
"TAGS\n#transformers #pytorch #tensorboard #bart #text2text-generation #summarization #azureml #azure #codecarbon #en #dataset-samsum #license-apache-2.0 #model-index #autotrain_compatible #endpoints_compatible #has_space #region-us \n",
"### Carbon Emissions\n\n\nThese results were obtained using 'CodeCarbon'. T... |
fill-mask | transformers | # CLIN-X-EN: a pre-trained language model for the English clinical domain
Details on the model, the pre-training corpus and the downstream task performance are given in the paper: "CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain" by Lukas Lange, Heike Adel, Jannik Strötgen and Dietrich Klakow.
The paper can be found [here](https://arxiv.org/abs/2112.08754).
In case of questions, please contact the authors as listed on the paper.
Please cite the above paper when reporting, reproducing or extending the results.
@misc{lange-etal-2021-clin-x,
author = {Lukas Lange and
Heike Adel and
Jannik Str{\"{o}}tgen and
Dietrich Klakow},
title = {CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain},
year={2021},
eprint={2112.08754},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2112.08754}
}
## Training details
The model is based on the multilingual XLM-R transformer `(xlm-roberta-large)`, which was trained on 100 languages and showed superior performance in many different tasks across languages and can even outperform monolingual models in certain settings (Conneau et al. 2020).
We train the CLIN-X model on clinical Pubmed abstracts (850MB) filtered
following Haynes et al. (2005). Pubmed is used with the courtesy of the U.S. National Library of Medicine
We initialize CLIN-X using the pre-trained XLM-R weights and train masked language modeling (MLM) on the Spanish clinical corpus for 3 epochs which roughly corresponds to 32k steps. This allows researchers and practitioners to address
the English clinical domain with an out-of-the-box tailored model.
## Results for Spanish concept extraction
We apply CLIN-X-EN to five different English sequence labeling tasks from i2b2 in a standard sequence labeling architecture similar to Devlin et al. 2019 and compare to BERT and ClinicalBERT. In addition, we perform experiments with an improved architecture `(+ OurArchitecture)` as described in the paper linked above. The code for our model architecture can be found [here](https://github.com/boschresearch/clin_x).
| | i2b2 2006 | i2b2 2010 | i2b2 2012 (Concept) | i2b2 2012 (Time) | i2b2 2014 |
|-------------------------------|-----------|-----------|---------------------|------------------|-----------|
| BERT | 94.80 | 82.25 | 76.51 | 75.28 | 94.86 |
| ClinicalBERT | 94.8 | 87.8 | 78.9 | 76.6 | 93.0 |
| CLIN-X (EN) | 96.25 | 88.10 | 79.58 | 77.70 | 96.73 |
| CLIN-X (EN) + OurArchitecture | **98.49** | **89.23** | **80.62** | **78.50** | **97.60** |
## Purpose of the project
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
## License
The CLIN-X models are open-sourced under the CC-BY 4.0 license.
See the [LICENSE](LICENSE) file for details. | {} | llange/xlm-roberta-large-english-clinical | null | [
"transformers",
"pytorch",
"xlm-roberta",
"fill-mask",
"arxiv:2112.08754",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2112.08754"
] | [] | TAGS
#transformers #pytorch #xlm-roberta #fill-mask #arxiv-2112.08754 #autotrain_compatible #endpoints_compatible #region-us
| CLIN-X-EN: a pre-trained language model for the English clinical domain
=======================================================================
Details on the model, the pre-training corpus and the downstream task performance are given in the paper: "CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain" by Lukas Lange, Heike Adel, Jannik Strötgen and Dietrich Klakow.
The paper can be found here.
In case of questions, please contact the authors as listed on the paper.
Please cite the above paper when reporting, reproducing or extending the results.
```
@misc{lange-etal-2021-clin-x,
author = {Lukas Lange and
Heike Adel and
Jannik Str{\"{o}}tgen and
Dietrich Klakow},
title = {CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain},
year={2021},
eprint={2112.08754},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={URL
}
```
Training details
----------------
The model is based on the multilingual XLM-R transformer '(xlm-roberta-large)', which was trained on 100 languages and showed superior performance in many different tasks across languages and can even outperform monolingual models in certain settings (Conneau et al. 2020).
We train the CLIN-X model on clinical Pubmed abstracts (850MB) filtered
following Haynes et al. (2005). Pubmed is used with the courtesy of the U.S. National Library of Medicine
We initialize CLIN-X using the pre-trained XLM-R weights and train masked language modeling (MLM) on the Spanish clinical corpus for 3 epochs which roughly corresponds to 32k steps. This allows researchers and practitioners to address
the English clinical domain with an out-of-the-box tailored model.
Results for Spanish concept extraction
--------------------------------------
We apply CLIN-X-EN to five different English sequence labeling tasks from i2b2 in a standard sequence labeling architecture similar to Devlin et al. 2019 and compare to BERT and ClinicalBERT. In addition, we perform experiments with an improved architecture '(+ OurArchitecture)' as described in the paper linked above. The code for our model architecture can be found here.
Purpose of the project
----------------------
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
License
-------
The CLIN-X models are open-sourced under the CC-BY 4.0 license.
See the LICENSE file for details.
| [] | [
"TAGS\n#transformers #pytorch #xlm-roberta #fill-mask #arxiv-2112.08754 #autotrain_compatible #endpoints_compatible #region-us \n"
] |
fill-mask | transformers | # CLIN-X-ES: a pre-trained language model for the Spanish clinical domain
Details on the model, the pre-training corpus and the downstream task performance are given in the paper: "CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain" by Lukas Lange, Heike Adel, Jannik Strötgen and Dietrich Klakow.
The paper can be found [here](https://arxiv.org/abs/2112.08754).
In case of questions, please contact the authors as listed on the paper.
Please cite the above paper when reporting, reproducing or extending the results.
@misc{lange-etal-2021-clin-x,
author = {Lukas Lange and
Heike Adel and
Jannik Str{\"{o}}tgen and
Dietrich Klakow},
title = {CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain},
year={2021},
eprint={2112.08754},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2112.08754}
}
## Training details
The model is based on the multilingual XLM-R transformer `(xlm-roberta-large)`, which was trained on 100 languages and showed superior performance in many different tasks across languages and can even outperform monolingual models in certain settings (Conneau et al. 2020).
Even though XLM-R was pre-trained on 53GB of Spanish documents, this was only 2% of the overall training data. To steer this model towards the Spanish clinical domain, we sample documents from the Scielo archive (https://scielo.org/)
and the MeSpEn resources (Villegas et al. 2018). The resulting corpus has a size of 790MB and is highly specific for the clinical domain.
We initialize CLIN-X using the pre-trained XLM-R weights and train masked language modeling (MLM) on the Spanish clinical corpus for 3 epochs which roughly corresponds to 32k steps. This allows researchers and practitioners to address
the Spanish clinical domain with an out-of-the-box tailored model.
## Results for Spanish concept extraction
We apply CLIN-X-ES to five Spanish concept extraction tasks from the clinical domain in a standard sequence labeling architecture similar to Devlin et al. 2019 and compare to a Spanish BERT model called BETO. In addition, we perform experiments with an improved architecture `(+ OurArchitecture)` as described in the paper linked above. The code for our model architecture can be found [here](https://github.com/boschresearch/clin_x).
| | Cantemist | Meddocan | Meddoprof (NER) | Meddoprof (CLASS) | Pharmaconer |
|------------------------------------------|-----------|----------|-----------------|-------------------|-------------|
| BETO (Spanish BERT) | 81.30 | 96.81 | 79.19 | 74.59 | 87.70 |
| CLIN-X (ES) | 83.22 | 97.08 | 79.54 | 76.95 | 90.05 |
| CLIN-X (ES) + OurArchitecture | **88.24** | **98.00** | **81.68** | **80.54** | **92.27** |
### Results for English concept extraction
As the CLIN-X-ES model is based on XLM-R, the model is still multilingual and we demonstrate the positive impact of cross-language domain adaptation by applying this model to five different English sequence labeling tasks from i2b2.
We found that further transfer from related concept extraction is particularly helpful in this cross-language setting. For a detailed description of the transfer process and all other models, we refer to our paper.
| | i2b2 2006 | i2b2 2010 | i2b2 2012 (Concept) | i2b2 2012 (Time) | i2b2 2014 |
|------------------------------------------|-----------|-----------|---------------|---------------|-----------|
| BERT | 94.80 | 85.25 | 76.51 | 75.28 | 94.86 |
| ClinicalBERT | 94.8 | 87.8 | 78.9 | 76.6 | 93.0 |
| CLIN-X (ES) | 95.49 | 87.94 | 79.58 | 77.57 | 96.80 |
| CLIN-X (ES) + OurArchitecture | 98.30 | 89.10 | 80.42 | 78.48 | **97.62** |
| CLIN-X (ES) + OurArchitecture + Transfer | **89.50** | **89.74** | **80.93** | **79.60** | 97.46 |
## Purpose of the project
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
## License
The CLIN-X models are open-sourced under the CC-BY 4.0 license.
See the [LICENSE](LICENSE) file for details. | {} | llange/xlm-roberta-large-spanish-clinical | null | [
"transformers",
"pytorch",
"xlm-roberta",
"fill-mask",
"arxiv:2112.08754",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [
"2112.08754"
] | [] | TAGS
#transformers #pytorch #xlm-roberta #fill-mask #arxiv-2112.08754 #autotrain_compatible #endpoints_compatible #region-us
| CLIN-X-ES: a pre-trained language model for the Spanish clinical domain
=======================================================================
Details on the model, the pre-training corpus and the downstream task performance are given in the paper: "CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain" by Lukas Lange, Heike Adel, Jannik Strötgen and Dietrich Klakow.
The paper can be found here.
In case of questions, please contact the authors as listed on the paper.
Please cite the above paper when reporting, reproducing or extending the results.
```
@misc{lange-etal-2021-clin-x,
author = {Lukas Lange and
Heike Adel and
Jannik Str{\"{o}}tgen and
Dietrich Klakow},
title = {CLIN-X: pre-trained language models and a study on cross-task transfer for concept extraction in the clinical domain},
year={2021},
eprint={2112.08754},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={URL
}
```
Training details
----------------
The model is based on the multilingual XLM-R transformer '(xlm-roberta-large)', which was trained on 100 languages and showed superior performance in many different tasks across languages and can even outperform monolingual models in certain settings (Conneau et al. 2020).
Even though XLM-R was pre-trained on 53GB of Spanish documents, this was only 2% of the overall training data. To steer this model towards the Spanish clinical domain, we sample documents from the Scielo archive (URL
and the MeSpEn resources (Villegas et al. 2018). The resulting corpus has a size of 790MB and is highly specific for the clinical domain.
We initialize CLIN-X using the pre-trained XLM-R weights and train masked language modeling (MLM) on the Spanish clinical corpus for 3 epochs which roughly corresponds to 32k steps. This allows researchers and practitioners to address
the Spanish clinical domain with an out-of-the-box tailored model.
Results for Spanish concept extraction
--------------------------------------
We apply CLIN-X-ES to five Spanish concept extraction tasks from the clinical domain in a standard sequence labeling architecture similar to Devlin et al. 2019 and compare to a Spanish BERT model called BETO. In addition, we perform experiments with an improved architecture '(+ OurArchitecture)' as described in the paper linked above. The code for our model architecture can be found here.
### Results for English concept extraction
As the CLIN-X-ES model is based on XLM-R, the model is still multilingual and we demonstrate the positive impact of cross-language domain adaptation by applying this model to five different English sequence labeling tasks from i2b2.
We found that further transfer from related concept extraction is particularly helpful in this cross-language setting. For a detailed description of the transfer process and all other models, we refer to our paper.
Purpose of the project
----------------------
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
License
-------
The CLIN-X models are open-sourced under the CC-BY 4.0 license.
See the LICENSE file for details.
| [
"### Results for English concept extraction\n\n\nAs the CLIN-X-ES model is based on XLM-R, the model is still multilingual and we demonstrate the positive impact of cross-language domain adaptation by applying this model to five different English sequence labeling tasks from i2b2.\n\n\nWe found that further transfe... | [
"TAGS\n#transformers #pytorch #xlm-roberta #fill-mask #arxiv-2112.08754 #autotrain_compatible #endpoints_compatible #region-us \n",
"### Results for English concept extraction\n\n\nAs the CLIN-X-ES model is based on XLM-R, the model is still multilingual and we demonstrate the positive impact of cross-language do... |
fill-mask | transformers | # Spanish XLM-R (from NLNDE-MEDDOPROF)
This Spanish language model was created for the MEDDOPROF shared task as part of the **NLNDE** team submission and outperformed all other participants in both sequence labeling tasks.
Details on the model, the pre-training corpus and the downstream task performance are given in the paper: "Boosting Transformers for Job Expression Extraction and Classification in a Low-Resource Setting" by Lukas Lange, Heike Adel and Jannik Strötgen.
The paper can be found [here](http://ceur-ws.org/Vol-2943/meddoprof_paper1.pdf).
In case of questions, please contact the authors as listed on the paper.
Please cite the above paper when reporting, reproducing or extending the results.
@inproceedings{lange-etal-2021-meddoprof,
author = {Lukas Lange and
Heike Adel and
Jannik Str{\"{o}}tgen},
title = {Boosting Transformers for Job Expression Extraction and Classification in a Low-Resource Setting},
year={2021},
booktitle= {{Proceedings of The Iberian Languages Evaluation Forum (IberLEF 2021)}},
series = {{CEUR} Workshop Proceedings},
url = {http://ceur-ws.org/Vol-2943/meddoprof_paper1.pdf},
}
## Training details
We use XLM-R (`xlm-roberta-large`, Conneau et al. 2020) as the main component of our models. XLM-R is a pretrained multilingual transformer model for 100 languages, including Spanish. It shows superior performance in different tasks across languages, and can even outperform
monolingual models in certain settings. It was pretrained on a large-scale corpus,
and Spanish documents made up only 2% of this data.
Thus, we explore further pretraining of this model and tune it towards Spanish
documents by pretraining a medium-size Spanish corpus with general
domain documents. For this, we use the [spanish corpus](https://github.com/josecannete/spanish-corpora) used to train the BETO model.
We use masked language modeling for pretraining and trained for three epochs
over the corpus, which roughly corresponds to 685k steps using a batch-size of 4.
## Performance
This model was trained in the context of the Meddoprof shared tasks and outperformed all other participants in both sequence labeling tasks. Our results (F1) in comparison with the standard XLM-R and the second-best system of the shared task are given in the Table.
More information on the shared task and other participants is given in this paper [here](http://journal.sepln.org/sepln/ojs/ojs/index.php/pln/article/view/6393/3813).
The code for our NER models can be found [here](https://github.com/boschresearch/nlnde-meddoprof).
| | Meddoprof Task 1 (NER) | Meddoprof Task 2 (CLASS) |
|---------------------------------|------------------------|--------------------------|
| Second-best System | 80.0 | 76.4 |
| XLM-R (our baseline) | 79.2 | 77.6 |
| Our Spanish XLM-R (best System) | **83.2** | **79.1** |
## Purpose of the project
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
## License
The CLIN-X models are open-sourced under the CC-BY 4.0 license.
See the [LICENSE](LICENSE) file for details. | {} | llange/xlm-roberta-large-spanish | null | [
"transformers",
"pytorch",
"xlm-roberta",
"fill-mask",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #xlm-roberta #fill-mask #autotrain_compatible #endpoints_compatible #region-us
| Spanish XLM-R (from NLNDE-MEDDOPROF)
====================================
This Spanish language model was created for the MEDDOPROF shared task as part of the NLNDE team submission and outperformed all other participants in both sequence labeling tasks.
Details on the model, the pre-training corpus and the downstream task performance are given in the paper: "Boosting Transformers for Job Expression Extraction and Classification in a Low-Resource Setting" by Lukas Lange, Heike Adel and Jannik Strötgen.
The paper can be found here.
In case of questions, please contact the authors as listed on the paper.
Please cite the above paper when reporting, reproducing or extending the results.
```
@inproceedings{lange-etal-2021-meddoprof,
author = {Lukas Lange and
Heike Adel and
Jannik Str{\"{o}}tgen},
title = {Boosting Transformers for Job Expression Extraction and Classification in a Low-Resource Setting},
year={2021},
booktitle= {{Proceedings of The Iberian Languages Evaluation Forum (IberLEF 2021)}},
series = {{CEUR} Workshop Proceedings},
url = {URL
}
```
Training details
----------------
We use XLM-R ('xlm-roberta-large', Conneau et al. 2020) as the main component of our models. XLM-R is a pretrained multilingual transformer model for 100 languages, including Spanish. It shows superior performance in different tasks across languages, and can even outperform
monolingual models in certain settings. It was pretrained on a large-scale corpus,
and Spanish documents made up only 2% of this data.
Thus, we explore further pretraining of this model and tune it towards Spanish
documents by pretraining a medium-size Spanish corpus with general
domain documents. For this, we use the spanish corpus used to train the BETO model.
We use masked language modeling for pretraining and trained for three epochs
over the corpus, which roughly corresponds to 685k steps using a batch-size of 4.
Performance
-----------
This model was trained in the context of the Meddoprof shared tasks and outperformed all other participants in both sequence labeling tasks. Our results (F1) in comparison with the standard XLM-R and the second-best system of the shared task are given in the Table.
More information on the shared task and other participants is given in this paper here.
The code for our NER models can be found here.
Meddoprof Task 1 (NER): Second-best System, Meddoprof Task 2 (CLASS): 80.0
Meddoprof Task 1 (NER): XLM-R (our baseline), Meddoprof Task 2 (CLASS): 79.2
Meddoprof Task 1 (NER): Our Spanish XLM-R (best System), Meddoprof Task 2 (CLASS): 83.2
Purpose of the project
----------------------
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
License
-------
The CLIN-X models are open-sourced under the CC-BY 4.0 license.
See the LICENSE file for details.
| [] | [
"TAGS\n#transformers #pytorch #xlm-roberta #fill-mask #autotrain_compatible #endpoints_compatible #region-us \n"
] |
text-classification | transformers |
## long-covid-classification
We fine-tuned bert-base-cased using a [manually curated dataset](https://huggingface.co/llangnickel/long-covid-classification-data) to train a Sequence Classification model able to distinguish between long COVID and non-long COVID-related documents.
## Used hyper parameters
|Parameter|Value|
|---|---|
|Learning rate|3e-5|
|Batch size|16|
|Number of epochs|4|
|Sequence Length|512|
## Metrics
|Precision [%]|Recall [%]|F1-score [%]|
|---|---|---|
|91.18|91.18|91.18|
## How to load the model
```
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("llangnickel/long-covid-classification", use_auth_token=True)
label_dict = {0: "nonLongCOVID", 1: "longCOVID"}
model = AutoModelForSequenceClassification.from_pretrained("llangnickel/long-covid-classification", use_auth_token=True, num_labels=len(label_dict))
```
## Citation
@article{10.1093/database/baac048,
author = {Langnickel, Lisa and Darms, Johannes and Heldt, Katharina and Ducks, Denise and Fluck, Juliane},
title = "{Continuous development of the semantic search engine preVIEW: from COVID-19 to long COVID}",
journal = {Database},
volume = {2022},
year = {2022},
month = {07},
issn = {1758-0463},
doi = {10.1093/database/baac048},
url = {https://doi.org/10.1093/database/baac048},
note = {baac048},
eprint = {https://academic.oup.com/database/article-pdf/doi/10.1093/database/baac048/44371817/baac048.pdf},
} | {"license": "mit"} | llangnickel/long-covid-classification | null | [
"transformers",
"pytorch",
"bert",
"text-classification",
"license:mit",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #bert #text-classification #license-mit #autotrain_compatible #endpoints_compatible #region-us
| long-covid-classification
-------------------------
We fine-tuned bert-base-cased using a manually curated dataset to train a Sequence Classification model able to distinguish between long COVID and non-long COVID-related documents.
Used hyper parameters
---------------------
Metrics
-------
Precision [%]: 91.18, Recall [%]: 91.18, F1-score [%]: 91.18
How to load the model
---------------------
@article{10.1093/database/baac048,
author = {Langnickel, Lisa and Darms, Johannes and Heldt, Katharina and Ducks, Denise and Fluck, Juliane},
title = "{Continuous development of the semantic search engine preVIEW: from COVID-19 to long COVID}",
journal = {Database},
volume = {2022},
year = {2022},
month = {07},
issn = {1758-0463},
doi = {10.1093/database/baac048},
url = {URL
note = {baac048},
eprint = {URL
}
| [] | [
"TAGS\n#transformers #pytorch #bert #text-classification #license-mit #autotrain_compatible #endpoints_compatible #region-us \n"
] |
text-generation | transformers |
# Game of thrones DialoGPT | {"tags": ["conversational"]} | cosmicroxks/DialoGPT-small-scott | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"conversational",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
# Game of thrones DialoGPT | [
"# Game of thrones DialoGPT"
] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"# Game of thrones DialoGPT"
] |
text-generation | transformers |
# harry potter DialogGPT Model | {"tags": ["conversational"]} | logube/DialogGPT_small_harrypotter | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"conversational",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
# harry potter DialogGPT Model | [
"# harry potter DialogGPT Model"
] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"# harry potter DialogGPT Model"
] |
token-classification | flair |
Published with ❤️ from [londogard](https://londogard.com).
## Swedish NER in Flair (SUC 3.0)
F1-Score: **85.6** (SUC 3.0)
Predicts 8 tags:
|**Tag**|**Meaning**|
|---|---|
| PRS| person name |
| ORG | organisation name|
| TME | time unit |
| WRK | building name |
| LOC | location name |
| EVN | event name |
| MSR | measurement unit |
| OBJ | object (like "Rolls-Royce" is a object in the form of a special car) |
Based on [Flair embeddings](https://www.aclweb.org/anthology/C18-1139/) and LSTM-CRF.
---
### Demo: How to use in Flair
Requires: **[Flair](https://github.com/flairNLP/flair/)** (`pip install flair`)
```python
from flair.data import Sentence
from flair.models import SequenceTagger
# load tagger
tagger = SequenceTagger.load("londogard/flair-swe-ner")
# make example sentence
sentence = Sentence("Hampus bor i Skåne och har levererat denna model idag.")
# predict NER tags
tagger.predict(sentence)
# print sentence
print(sentence)
# print predicted NER spans
print('The following NER tags are found:')
# iterate over entities and print
for entity in sentence.get_spans('ner'):
print(entity)
```
This yields the following output:
```
Span [0]: "Hampus" [− Labels: PRS (1.0)]
Span [3]: "Skåne" [− Labels: LOC (1.0)]
Span [9]: "idag" [− Labels: TME(1.0)]
```
So, the entities "_Hampus_" (labeled as a **PRS**), "_Skåne_" (labeled as a **LOC**), "_idag_" (labeled as a **TME**) are found in the sentence "_Hampus bor i Skåne och har levererat denna model idag._".
---
**Please mention londogard if using this models.** | {"language": "sv", "tags": ["flair", "token-classification", "sequence-tagger-model"], "datasets": ["SUC 3.0"], "widget": [{"text": "Hampus bor i Sk\u00e5ne och har levererat denna model idag."}]} | londogard/flair-swe-ner | null | [
"flair",
"pytorch",
"token-classification",
"sequence-tagger-model",
"sv",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [
"sv"
] | TAGS
#flair #pytorch #token-classification #sequence-tagger-model #sv #region-us
| Published with ️ from londogard.
Swedish NER in Flair (SUC 3.0)
------------------------------
F1-Score: 85.6 (SUC 3.0)
Predicts 8 tags:
Based on Flair embeddings and LSTM-CRF.
---
### Demo: How to use in Flair
Requires: Flair ('pip install flair')
This yields the following output:
So, the entities "*Hampus*" (labeled as a PRS), "*Skåne*" (labeled as a LOC), "*idag*" (labeled as a TME) are found in the sentence "*Hampus bor i Skåne och har levererat denna model idag.*".
---
Please mention londogard if using this models.
| [
"### Demo: How to use in Flair\n\n\nRequires: Flair ('pip install flair')\n\n\nThis yields the following output:\n\n\nSo, the entities \"*Hampus*\" (labeled as a PRS), \"*Skåne*\" (labeled as a LOC), \"*idag*\" (labeled as a TME) are found in the sentence \"*Hampus bor i Skåne och har levererat denna model idag.*\"... | [
"TAGS\n#flair #pytorch #token-classification #sequence-tagger-model #sv #region-us \n",
"### Demo: How to use in Flair\n\n\nRequires: Flair ('pip install flair')\n\n\nThis yields the following output:\n\n\nSo, the entities \"*Hampus*\" (labeled as a PRS), \"*Skåne*\" (labeled as a LOC), \"*idag*\" (labeled as a T... |
text-generation | transformers |
# Joshua DialoGPT Model | {"tags": ["conversational"]} | lonewanderer27/DialoGPT-small-Joshua | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"conversational",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
# Joshua DialoGPT Model | [
"# Joshua DialoGPT Model"
] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"# Joshua DialoGPT Model"
] |
text-generation | transformers |
# Camp Buddy - Keitaro - DialoGPTSmall Model | {"tags": ["conversational"]} | lonewanderer27/KeitaroBot | null | [
"transformers",
"pytorch",
"gpt2",
"text-generation",
"conversational",
"autotrain_compatible",
"endpoints_compatible",
"text-generation-inference",
"region:us"
] | null | 2022-03-02T23:29:05+00:00 | [] | [] | TAGS
#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us
|
# Camp Buddy - Keitaro - DialoGPTSmall Model | [
"# Camp Buddy - Keitaro - DialoGPTSmall Model"
] | [
"TAGS\n#transformers #pytorch #gpt2 #text-generation #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n",
"# Camp Buddy - Keitaro - DialoGPTSmall Model"
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.