Fix text alignment requirements

[homework36]

Resolves: (#1305 #1260)

• I passed the docker hub test, and images can be built successfully.
• I passed the GitHub CI test, so rodan functionalities and jobs work.

Update requirements.txt in text_alignment with

• Pypi link install for tensorflow to fix Docker Hub build failure due to package hash mismatch
• Packaged calamari-ocr and tfaip from DDMAL forks

More details here: #1305 (comment)

[homework36]

line by line difference for the calamari repo is this:

diff --git a/calamari_ocr/ocr/dataset/codec.py b/calamari_ocr/ocr/dataset/codec.py
index 2109e42..ce6f596 100644
--- a/calamari_ocr/ocr/dataset/codec.py
+++ b/calamari_ocr/ocr/dataset/codec.py
@@ -135,7 +135,7 @@ class Codec:
             self.char2code[char] = code
 
     def __len__(self):
-        """Get the number of characeters in the charset
+        """Get the number of characters in the charset
 
         this is equal to the maximum possible label.
 
@@ -146,7 +146,7 @@ class Codec:
         return len(self.charset)
 
     def size(self):
-        """Get the number of characeters in the charset
+        """Get the number of characters in the charset
 
         this is equal to the maximum possible label.
 
@@ -162,7 +162,7 @@ class Codec:
         Parameters
         ----------
         s : str
-            sequence of characeters
+            sequence of characters
 
         Returns
         -------
diff --git a/docs/source/doc.command-line-usage.rst b/docs/source/doc.command-line-usage.rst
index 7db0f46..b96de2d 100644
--- a/docs/source/doc.command-line-usage.rst
+++ b/docs/source/doc.command-line-usage.rst
@@ -350,7 +350,7 @@ All Parameters
       --trainer.current_epoch TRAINER.CURRENT_EPOCH
                             The epoch to start with. Usually 0, but can be overwritten for resume training. (default: 0)
       --trainer.samples_per_epoch TRAINER.SAMPLES_PER_EPOCH
-                            The number of samples (not batches!) to process per epoch. By default (-1) the size fo the training dataset. (default: -1)
+                            The number of samples (not batches!) to process per epoch. By default (-1) the size of the training dataset. (default: -1)
       --trainer.scale_epoch_size TRAINER.SCALE_EPOCH_SIZE
                             Multiply the number of samples per epoch by this factor. This is useful when using the dataset size as samples per epoch (--samples_per_epoch=-1, the default), but if you desire to set it e.g. to the half dataset size
                             (--scale_epoch_size=0.5) (default: 1)
@@ -373,7 +373,7 @@ All Parameters
       --trainer.output_dir TRAINER.OUTPUT_DIR
                             Dictionary to use to write checkpoints, logging files, and export of best and last model. (default: None)
       --trainer.write_checkpoints TRAINER.WRITE_CHECKPOINTS
-                            Write checkpoints to output_dir during training. Checkpoints are obligatory if you want support toresume the training (see tfaip-resume-training script) (default: True)
+                            Write checkpoints to output_dir during training. Checkpoints are obligatory if you want support to resume the training (see tfaip-resume-training script) (default: True)
       --trainer.export_best TRAINER.EXPORT_BEST
                             Continuously export the best model during testing to output_dir/best. (default: None)
       --trainer.export_final TRAINER.EXPORT_FINAL
@@ -447,7 +447,7 @@ All Parameters
       --scenario.print_eval_limit SCENARIO.PRINT_EVAL_LIMIT
                             Number of evaluation examples to print per evaluation, use -1 to print all (default: 10)
       --scenario.tensorboard_logger_history_size SCENARIO.TENSORBOARD_LOGGER_HISTORY_SIZE
-                            Number of instances to store for outputing into tensorboard. Default (last n=5) (default: 5)
+                            Number of instances to store for outputting into tensorboard. Default (last n=5) (default: 5)
       --scenario.export_serve SCENARIO.EXPORT_SERVE
                             Export the serving model (saved model format) (default: True)
       --model MODEL
diff --git a/docs/source/doc.predicting.rst b/docs/source/doc.predicting.rst
index 4a75304..2c1080a 100644
--- a/docs/source/doc.predicting.rst
+++ b/docs/source/doc.predicting.rst
@@ -44,7 +44,7 @@ If instead the `samples (lines) are dynamically created` during the execution an
         checkpoint='PATH_TO_THE_MODEL_WITHOUT_EXT')
     raw_predictor = predictor.raw().__enter__()  # you can also wrap the following lines in a `with`-block
 
-    # somehwere else in your code, just call the raw_predictor with a single image
+    # somewhere else in your code, just call the raw_predictor with a single image
     sample = raw_predictor(raw_image)  # raw_image is e.g. np.zeros(200, 50)
     inputs, prediction, meta = sample.inputs, sample.outputs, sample.meta
     # prediction is usually what you are looking for
diff --git a/requirements.txt b/requirements.txt
index 0d14f73..f305a5a 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -3,5 +3,5 @@ h5py==3.1.0
 lxml==4.9.2
 python-bidi==0.4.2
 tensorflow==2.5.1
-git+https://github.com/p42ul/tfaip@3738106#egg=tfaip
+https://github.com/DDMAL/tfaip/archive/refs/tags/v1.0.0.zip#egg=tfaip
 xlsxwriter==3.1.2
diff --git a/setup.py b/setup.py
index b77dca7..476bb55 100644
--- a/setup.py
+++ b/setup.py
@@ -41,7 +41,8 @@ setup(
     },
     python_requires=">=3.7",
     install_requires=[r for r in open("requirements.txt").read().split("\n")
-        if not r.startswith("git+")],
+        if not r.startswith("https")],
+        # We ignore lines starting with "https" because because we need a DDMAL customized version of tfaip
     keywords=["OCR", "optical character recognition", "ocropy", "ocropus", "kraken"],
     data_files=[("", ["requirements.txt"] + resources)],
 )

[notkaramel]

Looks good to me, and also on calamari repo.

We can test this and then merge it to develop

[homework36]

The Docker Hub build failure comes from py3-celery. Build is failing during pip install - appears to hang while compiling uWSGI==2.0.18 and uWSGI is known to be slow to compile.

ChatGPT suggests the following:

1. Add build deps before pip install

 RUN apt-get update && apt-get install -y build-essential python3-dev

2. Or use pre-built wheels

RUN pip install --only-binary=uWSGI -r requirements.txt

3. Or increase build timeout and split the install

RUN pip install --timeout 300 uWSGI==2.0.18
RUN pip install -r requirements.txt

We are not going to worry too much about this for now. We will just retry build...

In the new build it looks alright

2025-06-23T13:48:29Z #42 DONE 71.9s
2025-06-23T13:48:29Z [+] Pushing ddmal/rodan-python3-celery:fix1305
2025-06-23T13:48:29Z The push refers to repository [docker.io/ddmal/rodan-python3-celery]

[notkaramel]

I'm test running Rodan with gpu-celery's fix1305 tag. Will test if text alignment appears and works, and let you know!

[notkaramel]

Testing successfully for Text Alignment on fix1305 GPU Celery (shoutout @kyrieb-ekat for the support)
Feel free to merge the PR!