I am new to MatrchZoo. I wonder how to run aNMM. The docs don't have usage for aNMM. I feel I have to run a script for calculating the bin_sizes for aNMM? But I cannot find where this script lies.

Furthermore, my training data does need to have a format like here: https://github.com/NTMC-Community/MatchZoo/blob/master/matchzoo/datasets/toy/train.csv

right?

And, where are the batches created? Since you have positive and negative documents for each query, the batch should contain examples with pos and negs samples, right?

How can I load my own data?

Thanks.

Anybody wanting to make suggestions for MZ 2.0, please add it in this issue.

Here are my suggestions:

• [x] Add docstrings for all functions and classes
• [ ] Make MZ OS independent
• [ ] Make MZ usable by providing custom data
• [ ] Allow External Benchmarking
• [ ] Siamese Recurrent Networks (Proposed Model)
• [ ] docker, conda, virtualenv support (wishlist)

More details at https://github.com/faneshion/MatchZoo/issues/106

When running through the procedure described in the readme for the benchmark results of WikiQA, the reproduced values for NDCG@3, NDCG@5, and MAP are roughly half of the values shown in the table. Could you provide insight as to why this may be occuring?

Hi, I am trying to establish benchmark results on all the document similarity models at MatchZoo. While there are some established benchmarks, it would be good if we had a MatchZoo-code independent system for evaluating results.

Eg:

    input_data -> MZ -> result_data
    result_data - > independent_evaluation_code -> metric scores (Example: ndcg@k, map, etc.)

The current scenario is that the evaluation code is strongly ingrained in the MZ code, which can cause problems with different commits over time. As seen in https://github.com/faneshion/MatchZoo/issues/99

1. Is there a way already for doing this? I assume TREC is for that. Could someone direct me on how to use it? 2. Could some one direct me on how to go about making such an evaluation code? (Once developed, I will push it back into MZ and it could be like a Continuous Integration test.)

What do you think, @faneshion @yangliuy @bwanglzu @millanbatra @mandroid6?

Thanks!

如题，我目前的运行环境使用是tf2.0版本，keras是为2.3.0。 但无法执行 报错信息如下：

~/anaconda3/lib/python3.7/site-packages/keras/engine/training.py in _prepare_total_loss(self, masks)
    690 
    691                     output_loss = loss_fn(
--> 692                         y_true, y_pred, sample_weight=sample_weight)
    693 
    694                 if len(self.outputs) > 1:

TypeError: __call__() got an unexpected keyword argument 'sample_weight'

I have been running DSSM on quite a large dataset and was looking at tensorflow-gpu to speed up the training. However the returning loss and mae are always NaN for both the train and evaluation phase. I have tried a very basic tensorflow model from their tutorials and it works fine.

Im not really sure where to start debugging with this, any help would be greatly appreciated.

The model works fine with the cpu version of tensorflow. Example:

model.fit(x,y, epochs=2)

Epoch 1/2
10000/10000 [==============================] - 1s 139us/step - loss: nan - mean_absolute_error: nan
Epoch 2/2
10000/10000 [==============================] - 1s 138us/step - loss: nan - mean_absolute_error: nan

I see that the models usually needs a text1 and text2 to perform the training and predictions. Usually on search engines I just need the text2 (document) to perform the indexing step (training).

How can I train the model like a search engine? i.e. I don't have the text1 information (query/question) and I want to index my documents.

Does using the same text for text1 and text2 works for training?

I've added all modules for processing TREC dataset, mainly: the modifications enable to get TREC like run with corresponding ids for queries and documents. Hence, the evaluation with trec_eval is possible now. In addition to performing n-cross validation with MatchZoo. Soon, I'll add programs for constructing TREC input files that are needed by the added functions.

• update requirements.txt: keras=2.3.0 and tensorflow >= 2.0.0
• upgrade pip in .travis.yml (tf 2.0 requires pip >= 19)
• make raking losses inherit keras.losses.Loss to support sample_weight keyword param
• replace some keras.backend.tf with tf (K.tf does not exist anymore in 2.3.0 as keras is going to be synced with tf.keras and drop multi-backend)
• add clear_session before prepare in model tests to prevent OOM during CI test

fix #789

I find that input data is sent to model by outside circulation iteration. Seeing the follow plot.

I am feeling uncertain why do it and I change it to this(because I want to use tensorboard by callback function).

I just use model.fit_generator() to handel data and train. However, it raises a exception that is caused by validation_data。 I trace it into keras inner cores and find it occurs when model starts to run evaluate_generator()。In the function evaluate_generator()，eval data generator is empty and lead to a exception at a epoch! However, it is strange and confuses me why the exception does not occur in the start epoch。I trace code and think it may be a bugger of Keras，is it true? Additional, whether this is the reason that you make a outside iteration to train model。

大佬们好，我发现代码一开始是在外部循环迭代输入数据进入模型训练，我很奇怪并且我想在fit_generator()中调用回调函数，所以我自己手动改了代码，如图2，但是这个产生了一个异常，经过追溯，我发现是model在训练的时候，在一个epoch结束之后调用验证集跑的时候，验证集数据在keras内部代码通过一个队列调用的时候，是空的导致的异常，但是很奇怪的是，这个问题不是一开始的epoch就发生的，而是在epoch好多次之后的某一次epoch，出现了这个异常，我很奇怪，追溯代码，感觉好像是keras的bug，不知道对不对，另外我想一开始的数据处理，是不是为了解决这个问题？求指教～谢谢～

I already searched here. I use right now v1. Is there any sample code (I just found a broken link)? I have my trained DRRM model and want to predict ranking documents for a new query.

How is the current state in v2 to that?

I handleld to train the modle for my own custom text data with own fast word embedding. Normally I just would predict a new query but the output are the text IDs. So for DRRM are new words ignored which have no embedding in the dict?

Thank you very much!

Hello, I am getting the following error while I am trying to build the DRMM model for my ranking task at this line (here)

TypeError: Could not build a TypeSpec for KerasTensor(type_spec=TensorSpec(shape=(1, None, 10, 1), dtype=tf.float32, name=None), name='tf.operators.add/AddV2:0', description="created by layer 'tf.operators.add'") of unsupported type <class 'keras.engine.keras_tensor.KerasTensor'>.

Please note that I am using the following environment configuration in my local machine

Python 3.8.11 MatchZoo 2.1.0 tensorflow 2.8.0

Describe your attempts

• I checked the documentation and found no answer
• I checked to make sure that this is not a duplicate issue
• Additionally, I also tried different kinds of solution like this here
• And here

Context

• OS [macOS 12.4]:
• Hardware [Metal M1]

Thank-you for your help and time.

Regards, Govind Shukla

Describe the bug

MatchZoo breaks when run in google colab beacause of deprocated dependencies in keras

To Reproduce

Attempt to import match zoo in google colab:

!pip3 install matchzoo

import tensorflow from tensorflow import keras import matchzoo as mz import nltk import pandas as pd

Describe your attempts

Attempted to run matchzoo in google colab Fixed dependecy issues

You should also provide code snippets you tried as a workaround, StackOverflow solution that you have walked through, or your best guess of the cause that you can't locate (e.g. cosmic radiation).

Context

Nine FIles Needed edit: attention layer.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

data_generator.py import tensorflow # Added to fix toolchain issues #import keras from tensorflow import keras # Changed from previous line

decating_dropout_layer.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

dynamic_pooling_layer.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

matching_layer.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

matching_tensor_layer.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

multi_perspective_layer.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

semantic_composite_layer.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

spatial_gru.py #from keras.engine import Layer from keras.layers import Layer # Changed from previous line to fix tensorflow toolchain

Additional Information

I clone the repo and will push this update as a contribution to the code base

Hello, I'm working on SLGen(Structure Learning for Headline Generation(AAAI-20)) paper. In this paper, they are utilizing Deep and Wide component for Structure representation of Text documents.

Now, the question is does MatchZoo provide me the facility of finding a Deep component- Interaction-focused model and Representation-focused model - for a Text document.

If anybody working on this please help me out!

Thank you Darshan Tank

Describe the Question

I have a trained DSSM model and wanted to compare the ranked items based on dssm model.predict() scores against the cosine similarity scores after the model's dot layer, I would expect the two ranks to be the same since model.predict() is just the final score after a linear activation but the results are completely the opposite and I'm trying to understand how that might be given the linear coefficient from the final dense layer is positive.

Describe your attempts

• [x] I walked through the tutorials
• [x] I checked the documentation
• [x] I checked to make sure that this is not a duplicate question

Describe the bug

the project needs TensorFlow, but set_up.py does not contain the package. Although the requirements.txt contain the package, but when execute the command: pip install -e ., it will not install the package and occur no module error? Actually, is there any reason that not containing TensorFlow in set_up.py???

To Reproduce

pip3 install -e . python3 -m pytest -v tests/unit_test/processor_units/test_processor_units.py ============================= test session starts ============================== platform linux -- Python 3.7.12, pytest-6.2.5, py-1.11.0, pluggy-1.0.0 -- /mnt/zejun/smp/data/python_star_2000repo/MatchZoo/venv_test_7/bin/python3.7 cachedir: .pytest_cache rootdir: /mnt/zejun/smp/data/python_star_2000repo/MatchZoo plugins: cov-3.0.0, mock-3.6.1 collecting ... collected 0 items / 1 error

==================================== ERRORS ==================================== ___ ERROR collecting tests/unit_test/processor_units/test_processor_units.py ___ ImportError while importing test module '/mnt/zejun/smp/data/python_star_2000repo/MatchZoo/tests/unit_test/processor_units/test_processor_units.py'. Hint: make sure your test modules/packages have valid Python names. Traceback: /usr/lib/python3.7/importlib/init.py:127: in import_module return _bootstrap._gcd_import(name[level:], package, level) tests/unit_test/processor_units/test_processor_units.py:4: in from matchzoo.preprocessors import units matchzoo/init.py:20: in from . import preprocessors matchzoo/preprocessors/init.py:1: in from . import units matchzoo/preprocessors/units/init.py:13: in from .tokenize import Tokenize matchzoo/preprocessors/units/tokenize.py:2: in from matchzoo.utils.bert_utils import is_chinese_char,
matchzoo/utils/init.py:4: in from .make_keras_optimizer_picklable import make_keras_optimizer_picklable matchzoo/utils/make_keras_optimizer_picklable.py:1: in import keras venv_test_7/lib/python3.7/site-packages/keras/init.py:21: in from tensorflow.python import tf2 E ModuleNotFoundError: No module named 'tensorflow' =========================== short test summary info ============================ ERROR tests/unit_test/processor_units/test_processor_units.py !!!!!!!!!!!!!!!!!!!! Interrupted: 1 error during collection !!!!!!!!!!!!!!!!!!!! =============================== 1 error in 0.91s ===============================

Describe your attempts

• [x] I checked the documentation and found no answer
• [x] I checked to make sure that this is not a duplicate issue

Context

• Ubutun

Describe the bug

run the example in Get Started in 60 Seconds

Context

• OS Ubuntu18.04
• Hardware Tesla 80k, cuda 10.1,cudnn7.0
• matchzoo 2.2.0, tensorflow2.2.0, keras2.3.0

Additional Information

Other things you want the developers to know.