Comments (6)
已私发邮箱,请查收
from vibertgrid-pytorch.
已私发邮箱,请查收
Thanks!
from vibertgrid-pytorch.
can you please share the example config for funsd dataset?
from vibertgrid-pytorch.
comment: " FUNSD resnet-34-pretrained "
############################################
device: 'cuda'
syncBN: True
start_epoch: 0
end_epoch: 33
batch_size: 2
optimizer_cnn_hyp: # SGD -> CNN
learning_rate: 0.008
min_learning_rate: 0.00001
warm_up_epoches: 1
warm_up_init_lr: 0.00001
momentum: 0.9
weight_decay: 0.005
min_weight_decay: 0.005
optimizer_bert_hyp: # AdamW -> BERT
learning_rate: 0.000005
min_learning_rate: 0.0000001
warm_up_epoches: 1
warm_up_init_lr: 0.0000001
beta1: 0.9
beta2: 0.999
epsilon: 0.00000001
weight_decay: 0.01
min_weight_decay: 0.01
#############################################
num_hard_positive_main_1: 16
num_hard_negative_main_1: 16
num_hard_positive_main_2: 32
num_hard_negative_main_2: 32
loss_aux_sample_list:
- 256
- 512
- 256
num_hard_positive_aux: 256
num_hard_negative_aux: 256
ohem_random: True
#############################################
##################################
classifier_mode: "simp"
eval_mode: "seqeval"
tag_mode: "B"
#################################
save_top: "./weights/"
save_log: "./log/"
amp: True
weights: ''
num_workers: 0
#########################################################
data_root: "/dir/to/FUNSD"
#########################################################
# FUNSD
# #######################################
num_classes: 4
########################################
image_mean:
- 0.9480
- 0.9480
- 0.9480
image_std:
- 0.1840
- 0.1840
- 0.1840
image_min_size:
- 320
- 416
- 512
- 608
- 704
image_max_size: 800
test_image_min_size: 512
#########################################################
bert_version: "bert-base-uncased" # FUNSD
backbone: "resnet_34_fpn_pretrained"
########################################################
grid_mode: "mean"
early_fusion_downsampling_ratio: 8
roi_shape: 7
p_fuse_downsampling_ratio: 4
roi_align_output_reshape: False
late_fusion_fuse_embedding_channel: 1024
layer_mode: "single"
add_pos_neg: True
###########################
loss_weights:
# - 1
# - 1
# - 1.5
# - 1
# - 1.5
# - 0
# - 4.906
# - 5.372
# - 2.002
# - 5.373
loss_control_lambda: 1
from vibertgrid-pytorch.
Thank you!
Do you have any pretrained model with FUNSD dataset?
from vibertgrid-pytorch.
I'm sorry that I can't get the pre-trained model on FUNSD for you. The weights are stored on a server that I don't have access to currently due to position changes. You can train the model based on the configuration mentioned above, and it won't take up a long time.
from vibertgrid-pytorch.
Related Issues (10)
- SROIE dataset issues. HOT 5
- Model Training. HOT 2
- FUNSD dataset - empty key_dict HOT 2
- Validation in CRF mode HOT 2
- For Inference Pre-trained weights are not available. Inference running giving errors. HOT 2
- No predictions in inference. HOT 16
- About SROIE annotations HOT 10
- I need help about customize entities of SROIE dataset HOT 6
- Training on custom dataset HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from vibertgrid-pytorch.