Add char htr experiment, rename from ocr to htr, vqvae loss collapses

12 files changed, 96 insertions, 24 deletions

diff --git a/notebooks/05c-test-model-end-to-end.ipynb b/notebooks/05c-test-model-end-to-end.ipynb
index 850d205..913eafd 100644
--- a/notebooks/05c-test-model-end-to-end.ipynb
+++ b/notebooks/05c-test-model-end-to-end.ipynb
@@ -133,6 +133,7 @@
       "  _target_: text_recognizer.models.vqvae.VQVAELitModel\n",
       "  interval: step\n",
       "  monitor: val/loss\n",
+      "  latent_loss_weight: 0.25\n",
       "network:\n",
       "  _target_: text_recognizer.networks.vqvae.VQVAE\n",
       "  in_channels: 1\n",
@@ -174,7 +175,7 @@
       "print_config: true\n",
       "ignore_warnings: true\n",
       "\n",
-      "{'callbacks': {'model_checkpoint': {'_target_': 'pytorch_lightning.callbacks.ModelCheckpoint', 'monitor': 'val/loss', 'save_top_k': 1, 'save_last': True, 'mode': 'min', 'verbose': False, 'dirpath': 'checkpoints/', 'filename': '{epoch:02d}'}, 'learning_rate_monitor': {'_target_': 'pytorch_lightning.callbacks.LearningRateMonitor', 'logging_interval': 'step', 'log_momentum': False}, 'watch_model': {'_target_': 'callbacks.wandb_callbacks.WatchModel', 'log': 'all', 'log_freq': 100}, 'upload_code_as_artifact': {'_target_': 'callbacks.wandb_callbacks.UploadCodeAsArtifact', 'project_dir': '${work_dir}/text_recognizer'}, 'upload_ckpts_as_artifact': {'_target_': 'callbacks.wandb_callbacks.UploadCheckpointsAsArtifact', 'ckpt_dir': 'checkpoints/', 'upload_best_only': True}, 'log_image_reconstruction': {'_target_': 'callbacks.wandb_callbacks.LogReconstuctedImages', 'num_samples': 8}}, 'criterion': {'_target_': 'torch.nn.MSELoss', 'reduction': 'mean'}, 'datamodule': {'_target_': 'text_recognizer.data.iam_extended_paragraphs.IAMExtendedParagraphs', 'batch_size': 32, 'num_workers': 12, 'train_fraction': 0.8, 'augment': True, 'pin_memory': False, 'word_pieces': True}, 'logger': {'wandb': {'_target_': 'pytorch_lightning.loggers.wandb.WandbLogger', 'project': 'text-recognizer', 'name': None, 'save_dir': '.', 'offline': False, 'id': None, 'log_model': False, 'prefix': '', 'job_type': 'train', 'group': '', 'tags': []}}, 'lr_scheduler': {'_target_': 'torch.optim.lr_scheduler.OneCycleLR', 'max_lr': 0.001, 'total_steps': None, 'epochs': 64, 'steps_per_epoch': 624, 'pct_start': 0.3, 'anneal_strategy': 'cos', 'cycle_momentum': True, 'base_momentum': 0.85, 'max_momentum': 0.95, 'div_factor': 25.0, 'final_div_factor': 10000.0, 'three_phase': True, 'last_epoch': -1, 'verbose': False}, 'mapping': {'_target_': 'text_recognizer.data.word_piece_mapping.WordPieceMapping', 'num_features': 1000, 'tokens': 'iamdb_1kwp_tokens_1000.txt', 'lexicon': 'iamdb_1kwp_lex_1000.txt', 'data_dir': None, 'use_words': False, 'prepend_wordsep': False, 'special_tokens': ['<s>', '<e>', '<p>'], 'extra_symbols': ['\\n']}, 'model': {'_target_': 'text_recognizer.models.vqvae.VQVAELitModel', 'interval': 'step', 'monitor': 'val/loss'}, 'network': {'_target_': 'text_recognizer.networks.vqvae.VQVAE', 'in_channels': 1, 'res_channels': 32, 'num_residual_layers': 2, 'embedding_dim': 64, 'num_embeddings': 512, 'decay': 0.99, 'activation': 'mish'}, 'optimizer': {'_target_': 'madgrad.MADGRAD', 'lr': 0.01, 'momentum': 0.9, 'weight_decay': 0, 'eps': 1e-06}, 'trainer': {'_target_': 'pytorch_lightning.Trainer', 'stochastic_weight_avg': False, 'auto_scale_batch_size': 'binsearch', 'auto_lr_find': False, 'gradient_clip_val': 0, 'fast_dev_run': False, 'gpus': 1, 'precision': 16, 'max_epochs': 64, 'terminate_on_nan': True, 'weights_summary': 'top', 'limit_train_batches': 1.0, 'limit_val_batches': 1.0, 'limit_test_batches': 1.0, 'resume_from_checkpoint': None}, 'seed': 4711, 'tune': False, 'train': True, 'test': True, 'logging': 'INFO', 'work_dir': '${hydra:runtime.cwd}', 'debug': False, 'print_config': True, 'ignore_warnings': True}\n"
+      "{'callbacks': {'model_checkpoint': {'_target_': 'pytorch_lightning.callbacks.ModelCheckpoint', 'monitor': 'val/loss', 'save_top_k': 1, 'save_last': True, 'mode': 'min', 'verbose': False, 'dirpath': 'checkpoints/', 'filename': '{epoch:02d}'}, 'learning_rate_monitor': {'_target_': 'pytorch_lightning.callbacks.LearningRateMonitor', 'logging_interval': 'step', 'log_momentum': False}, 'watch_model': {'_target_': 'callbacks.wandb_callbacks.WatchModel', 'log': 'all', 'log_freq': 100}, 'upload_code_as_artifact': {'_target_': 'callbacks.wandb_callbacks.UploadCodeAsArtifact', 'project_dir': '${work_dir}/text_recognizer'}, 'upload_ckpts_as_artifact': {'_target_': 'callbacks.wandb_callbacks.UploadCheckpointsAsArtifact', 'ckpt_dir': 'checkpoints/', 'upload_best_only': True}, 'log_image_reconstruction': {'_target_': 'callbacks.wandb_callbacks.LogReconstuctedImages', 'num_samples': 8}}, 'criterion': {'_target_': 'torch.nn.MSELoss', 'reduction': 'mean'}, 'datamodule': {'_target_': 'text_recognizer.data.iam_extended_paragraphs.IAMExtendedParagraphs', 'batch_size': 32, 'num_workers': 12, 'train_fraction': 0.8, 'augment': True, 'pin_memory': False, 'word_pieces': True}, 'logger': {'wandb': {'_target_': 'pytorch_lightning.loggers.wandb.WandbLogger', 'project': 'text-recognizer', 'name': None, 'save_dir': '.', 'offline': False, 'id': None, 'log_model': False, 'prefix': '', 'job_type': 'train', 'group': '', 'tags': []}}, 'lr_scheduler': {'_target_': 'torch.optim.lr_scheduler.OneCycleLR', 'max_lr': 0.001, 'total_steps': None, 'epochs': 64, 'steps_per_epoch': 624, 'pct_start': 0.3, 'anneal_strategy': 'cos', 'cycle_momentum': True, 'base_momentum': 0.85, 'max_momentum': 0.95, 'div_factor': 25.0, 'final_div_factor': 10000.0, 'three_phase': True, 'last_epoch': -1, 'verbose': False}, 'mapping': {'_target_': 'text_recognizer.data.word_piece_mapping.WordPieceMapping', 'num_features': 1000, 'tokens': 'iamdb_1kwp_tokens_1000.txt', 'lexicon': 'iamdb_1kwp_lex_1000.txt', 'data_dir': None, 'use_words': False, 'prepend_wordsep': False, 'special_tokens': ['<s>', '<e>', '<p>'], 'extra_symbols': ['\\n']}, 'model': {'_target_': 'text_recognizer.models.vqvae.VQVAELitModel', 'interval': 'step', 'monitor': 'val/loss', 'latent_loss_weight': 0.25}, 'network': {'_target_': 'text_recognizer.networks.vqvae.VQVAE', 'in_channels': 1, 'res_channels': 32, 'num_residual_layers': 2, 'embedding_dim': 64, 'num_embeddings': 512, 'decay': 0.99, 'activation': 'mish'}, 'optimizer': {'_target_': 'madgrad.MADGRAD', 'lr': 0.01, 'momentum': 0.9, 'weight_decay': 0, 'eps': 1e-06}, 'trainer': {'_target_': 'pytorch_lightning.Trainer', 'stochastic_weight_avg': False, 'auto_scale_batch_size': 'binsearch', 'auto_lr_find': False, 'gradient_clip_val': 0, 'fast_dev_run': False, 'gpus': 1, 'precision': 16, 'max_epochs': 64, 'terminate_on_nan': True, 'weights_summary': 'top', 'limit_train_batches': 1.0, 'limit_val_batches': 1.0, 'limit_test_batches': 1.0, 'resume_from_checkpoint': None}, 'seed': 4711, 'tune': False, 'train': True, 'test': True, 'logging': 'INFO', 'work_dir': '${hydra:runtime.cwd}', 'debug': False, 'print_config': True, 'ignore_warnings': True}\n"
      ]
     }
    ],
@@ -196,7 +197,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "2021-08-04 04:49:04.188 | DEBUG    | text_recognizer.data.word_piece_mapping:__init__:37 - Using data dir: /home/aktersnurra/projects/text-recognizer/data/downloaded/iam/iamdb\n"
+      "2021-08-04 05:07:26.480 | DEBUG    | text_recognizer.data.word_piece_mapping:__init__:37 - Using data dir: /home/aktersnurra/projects/text-recognizer/data/downloaded/iam/iamdb\n"
      ]
     }
    ],
@@ -206,7 +207,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 35,
+   "execution_count": 5,
    "id": "969ba3be-d78f-4b1e-b522-ea8a42669e86",
    "metadata": {},
    "outputs": [],
@@ -216,7 +217,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 36,
+   "execution_count": 6,
    "id": "6147cd3e-0ad1-490f-917d-21be9bb8ce1c",
    "metadata": {},
    "outputs": [],
@@ -226,7 +227,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 37,
+   "execution_count": 7,
    "id": "a0ecea0c-abaf-4d5d-a13d-c085c1e4d282",
    "metadata": {},
    "outputs": [
@@ -236,7 +237,7 @@
        "torch.Size([1, 64, 144, 160])"
       ]
      },
-     "execution_count": 37,
+     "execution_count": 7,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -247,7 +248,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 38,
+   "execution_count": 8,
    "id": "a7b9f249-7e5e-4f31-bbe1-cfd6d3701cf0",
    "metadata": {},
    "outputs": [
@@ -260,20 +261,83 @@
       "torch.Size([512])\n",
       "torch.Size([512])\n"
      ]
-    },
+    }
+   ],
+   "source": [
+    "t, l = network(x)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "9a9450d2-f45d-4823-adac-68a8ea05ed1d",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "tensor(0.0188, grad_fn=<AddBackward0>)"
+      ]
+     },
+     "execution_count": 9,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "l"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "93b8c90f-788a-4095-aa7a-55b34f0ddaaf",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from torch.nn import functional as F\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "c9983788-2dae-4375-a821-a64cd1c68edf",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "tensor(0.5669, grad_fn=<AddBackward0>)"
+      ]
+     },
+     "execution_count": 15,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "F.mse_loss(x, t) + l"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "29b128ca-80b7-481e-bb3c-44f109c7d292",
+   "metadata": {},
+   "outputs": [
     {
      "data": {
       "text/plain": [
        "torch.Size([1, 1, 576, 640])"
       ]
      },
-     "execution_count": 38,
+     "execution_count": 10,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "network(x)[0].shape"
+    "t.shape"
    ]
   },
   {
diff --git a/text_recognizer/models/vqvae.py b/text_recognizer/models/vqvae.py
index 5890fd9..7f79b78 100644
--- a/text_recognizer/models/vqvae.py
+++ b/text_recognizer/models/vqvae.py
@@ -1,11 +1,8 @@
 """PyTorch Lightning model for base Transformers."""
-from typing import Any, Dict, Union, Tuple, Type
+from typing import Tuple
 
 import attr
-from omegaconf import DictConfig
-from torch import nn
 from torch import Tensor
-import wandb
 
 from text_recognizer.models.base import BaseLitModel
 
@@ -25,7 +22,7 @@ class VQVAELitModel(BaseLitModel):
         data, _ = batch
         reconstructions, vq_loss = self(data)
         loss = self.loss_fn(reconstructions, data)
-        loss += self.latent_loss_weight * vq_loss
+        loss = loss + self.latent_loss_weight * vq_loss
         self.log("train/loss", loss)
         return loss
 
@@ -34,7 +31,7 @@ class VQVAELitModel(BaseLitModel):
         data, _ = batch
         reconstructions, vq_loss = self(data)
         loss = self.loss_fn(reconstructions, data)
-        loss += self.latent_loss_weight * vq_loss
+        loss = loss + self.latent_loss_weight * vq_loss
         self.log("val/loss", loss, prog_bar=True)
 
     def test_step(self, batch: Tuple[Tensor, Tensor], batch_idx: int) -> None:
@@ -42,5 +39,5 @@ class VQVAELitModel(BaseLitModel):
         data, _ = batch
         reconstructions, vq_loss = self(data)
         loss = self.loss_fn(reconstructions, data)
-        loss += self.latent_loss_weight * vq_loss
+        loss = loss + self.latent_loss_weight * vq_loss
         self.log("test/loss", loss)
diff --git a/text_recognizer/networks/vqvae/quantizer.py b/text_recognizer/networks/vqvae/quantizer.py
index 5e0b602..1b59e78 100644
--- a/text_recognizer/networks/vqvae/quantizer.py
+++ b/text_recognizer/networks/vqvae/quantizer.py
@@ -83,8 +83,6 @@ class VectorQuantizer(nn.Module):
     def compute_ema(self, one_hot_encoding: Tensor, latent: Tensor) -> None:
         batch_cluster_size = one_hot_encoding.sum(axis=0)
         batch_embedding_avg = (latent.t() @ one_hot_encoding).t()
-        print(batch_cluster_size.shape)
-        print(self.embedding._cluster_size.shape)
         self.embedding._cluster_size.data.mul_(self.decay).add_(
             batch_cluster_size, alpha=1 - self.decay
         )
diff --git a/training/callbacks/wandb_callbacks.py b/training/callbacks/wandb_callbacks.py
index c750e4b..61d71df 100644
--- a/training/callbacks/wandb_callbacks.py
+++ b/training/callbacks/wandb_callbacks.py
@@ -119,7 +119,7 @@ class LogTextPredictions(Callback):
         ]
 
         experiment.log(
-            {f"OCR/{experiment.name}/{stage}": wandb.Table(data=data, columns=columns)}
+            {f"HTR/{experiment.name}/{stage}": wandb.Table(data=data, columns=columns)}
         )
 
     def on_sanity_check_start(
diff --git a/training/conf/callbacks/wandb_ocr.yaml b/training/conf/callbacks/wandb_htr.yaml
index 9c9a6da..9c9a6da 100644
--- a/training/conf/callbacks/wandb_ocr.yaml
+++ b/training/conf/callbacks/wandb_htr.yaml
diff --git a/training/conf/callbacks/wandb_ocr_predictions.yaml b/training/conf/callbacks/wandb_htr_predictions.yaml
index 573fa96..573fa96 100644
--- a/training/conf/callbacks/wandb_ocr_predictions.yaml
+++ b/training/conf/callbacks/wandb_htr_predictions.yaml
diff --git a/training/conf/config.yaml b/training/conf/config.yaml
index 6b74502..c606366 100644
--- a/training/conf/config.yaml
+++ b/training/conf/config.yaml
@@ -1,7 +1,7 @@
 # @package _global_
 
 defaults:
-  - callbacks: wandb_ocr
+  - callbacks: wandb_htr
   - criterion: label_smoothing
   - datamodule: iam_extended_paragraphs
   - hydra: default
diff --git a/training/conf/experiment/htr_char.yaml b/training/conf/experiment/htr_char.yaml
new file mode 100644
index 0000000..77126ae
--- /dev/null
+++ b/training/conf/experiment/htr_char.yaml
@@ -0,0 +1,12 @@
+# @package _global_
+
+defaults:
+  - override /mapping: characters
+
+criterion:
+  ignore_index: 3
+
+network:
+  num_classes: 89
+  pad_index: 3
+  max_output_len: 682
diff --git a/training/conf/experiment/vqvae.yaml b/training/conf/experiment/vqvae.yaml
index 13e5f34..699612e 100644
--- a/training/conf/experiment/vqvae.yaml
+++ b/training/conf/experiment/vqvae.yaml
@@ -8,6 +8,7 @@ defaults:
 
 trainer:
   max_epochs: 64
+  gradient_clip_val: 0.25
 
 datamodule:
   batch_size: 32
@@ -17,4 +18,4 @@ lr_scheduler:
   steps_per_epoch: 624
 
 optimizer:
-  lr: 1.0e-2
+  lr: 1.0e-3
diff --git a/training/conf/mapping/emnist.yaml b/training/conf/mapping/characters.yaml
index 14e966b..14e966b 100644
--- a/training/conf/mapping/emnist.yaml
+++ b/training/conf/mapping/characters.yaml
diff --git a/training/conf/network/decoder/transformer_decoder.yaml b/training/conf/network/decoder/transformer_decoder.yaml
index c326c04..bc0678b 100644
--- a/training/conf/network/decoder/transformer_decoder.yaml
+++ b/training/conf/network/decoder/transformer_decoder.yaml
@@ -4,7 +4,7 @@ defaults:
 _target_: text_recognizer.networks.transformer.Decoder
 dim: 128
 depth: 2
-num_heads: 8
+num_heads: 4
 attn_fn: text_recognizer.networks.transformer.attention.Attention
 attn_kwargs:
   dim_head: 64
diff --git a/training/conf/trainer/default.yaml b/training/conf/trainer/default.yaml
index c665adc..0fa9ce1 100644
--- a/training/conf/trainer/default.yaml
+++ b/training/conf/trainer/default.yaml
@@ -8,7 +8,7 @@ gpus: 1
 precision: 16
 max_epochs: 512
 terminate_on_nan: true
-weights_summary: top
+weights_summary: full
 limit_train_batches: 1.0 
 limit_val_batches: 1.0
 limit_test_batches: 1.0