From c9e50644ba14ae09aec5a44c12f8116bada26bab Mon Sep 17 00:00:00 2001
From: Gustaf Rydholm <gustaf.rydholm@gmail.com>
Date: Mon, 11 Oct 2021 22:11:24 +0200
Subject: Update Barlow Twins config

---
 training/conf/experiment/barlow_twins.yaml | 84 ++++++++++--------------------
 1 file changed, 27 insertions(+), 57 deletions(-)

(limited to 'training/conf/experiment')

diff --git a/training/conf/experiment/barlow_twins.yaml b/training/conf/experiment/barlow_twins.yaml
index 4901e18..cb5035b 100644
--- a/training/conf/experiment/barlow_twins.yaml
+++ b/training/conf/experiment/barlow_twins.yaml
@@ -1,3 +1,5 @@
+# @package _global_
+
 defaults:
   - override /criterion: null
   - override /datamodule: null
@@ -7,9 +9,15 @@ defaults:
   - override /optimizers: null
 
 
+print_config: true
 epochs: &epochs 1000
 summary: [[1, 1, 56, 1024]]
 
+criterion:
+  _target_: text_recognizer.criterions.barlow_twins.BarlowTwinsLoss
+  dim: 512
+  lambda_: 5.1e-3
+
 callbacks:
   stochastic_weight_averaging:
     _target_: pytorch_lightning.callbacks.StochasticWeightAveraging
@@ -22,21 +30,20 @@ callbacks:
 optimizers:
   madgrad:
     _target_: madgrad.MADGRAD
-    lr: 2.0e-4
+    lr: 3.0e-4
     momentum: 0.9
-    weight_decay: 0
+    weight_decay: 1.0e-6
     eps: 1.0e-6
-
     parameters: network
 
 lr_schedulers:
   network:
     _target_: torch.optim.lr_scheduler.OneCycleLR
-    max_lr: 2.0e-4
+    max_lr: 3.0e-4
     total_steps: null
     epochs: *epochs
-    steps_per_epoch: 632
-    pct_start: 0.3
+    steps_per_epoch: 45
+    pct_start: 0.03
     anneal_strategy: cos
     cycle_momentum: true
     base_momentum: 0.85
@@ -51,23 +58,18 @@ lr_schedulers:
     monitor: val/loss
 
 datamodule:
-  _target_: text_recognizer.data.iam_extended_paragraphs.IAMExtendedParagraphs
-  batch_size: 4
-  num_workers: 12
-  train_fraction: 0.8
-  augment: true
-  pin_memory: true
-  word_pieces: false
-  resize: null
+    _target_: text_recognizer.data.iam_lines.IAMLines
+    batch_size: 16
+    num_workers: 12
+    train_fraction: 0.8
+    pin_memory: false
+    transform: transform/iam_lines_barlow.yaml
+    test_transform: transform/iam_lines_barlow.yaml
+    mapping: 
+      _target_: text_recognizer.data.mappings.emnist_mapping.EmnistMapping
 
 network:
-  _target_: text_recognizer.networks.conv_transformer.ConvTransformer
-  input_dims: [1, 576, 640]
-  hidden_dim: &hidden_dim 128
-  encoder_dim: 1280
-  dropout_rate: 0.2
-  num_classes: *num_classes
-  pad_index: *ignore_index
+  _target_: text_recognizer.networks.barlow_twins.network.BarlowTwins
   encoder:
     _target_: text_recognizer.networks.encoders.efficientnet.EfficientNet
     arch: b0
@@ -75,44 +77,12 @@ network:
     stochastic_dropout_rate: 0.2
     bn_momentum: 0.99
     bn_eps: 1.0e-3
-  decoder:
-    _target_: text_recognizer.networks.transformer.Decoder
-    dim: *hidden_dim
-    depth: 3 
-    num_heads: 4
-    attn_fn: text_recognizer.networks.transformer.attention.Attention
-    attn_kwargs:
-      dim_head: 32
-      dropout_rate: 0.2
-    norm_fn: text_recognizer.networks.transformer.norm.ScaleNorm
-    ff_fn: text_recognizer.networks.transformer.mlp.FeedForward
-    ff_kwargs:
-      dim_out: null
-      expansion_factor: 4
-      glu: true
-      dropout_rate: 0.2
-    cross_attend: true
-    pre_norm: true
-    rotary_emb:
-      _target_: text_recognizer.networks.transformer.positional_encodings.rotary_embedding.RotaryEmbedding
-      dim: 32
-  pixel_pos_embedding:
-    _target_: text_recognizer.networks.transformer.positional_encodings.PositionalEncoding2D
-    hidden_dim: *hidden_dim 
-    max_h: 18
-    max_w: 20
-  token_pos_embedding:
-    _target_: text_recognizer.networks.transformer.positional_encodings.PositionalEncoding
-    hidden_dim: *hidden_dim 
-    dropout_rate: 0.2
-    max_len: *max_output_len
+  projector:
+    _target_: text_recognizer.networks.barlow_twins.projector.Projector
+    dims: [1280, 512, 512, 512]
 
 model:
-  _target_: text_recognizer.models.transformer.TransformerLitModel
-  max_output_len: *max_output_len
-  start_token: <s>
-  end_token: <e>
-  pad_token: <p>
+  _target_: text_recognizer.models.barlow_twins.BarlowTwinsLitModel
 
 trainer:
   _target_: pytorch_lightning.Trainer
-- 
cgit v1.2.3-70-g09d2