nghuyong / rasa-faq-bot Goto Github PK

2021-05-16 19:35:30 INFO rasa_sdk.endpoint - Starting action endpoint server...
2021-05-16 19:35:31 INFO root - Load pretrained SentenceTransformer: bert-base-nli-mean-tokens
2021-05-16 19:35:31 INFO root - Load SentenceTransformer from folder: bert-base-nli-mean-tokens
2021-05-16 19:35:34 INFO root - Use pytorch device: cpu
Standard question size 1000
Start to calculate encoder....
Batches: 0%| | 0/125 [00:00<?, ?it/s]Truncation was not explicitly activated but max_length is provided a specific value, please use truncation=True to explicitly truncate examples to max length. Defaulting to 'longest_first' truncation strategy. If you encode pairs of sequences (GLUE-style) with the tokenizer you can select this strategy more precisely by providing a specific strategy to truncation.
/opt/rasa/venv/lib/python3.6/site-packages/transformers/tokenization_utils_base.py:2143: FutureWarning: The pad_to_max_length argument is deprecated and will be removed in a future version, use padding=True or padding='longest' to pad to the longest sequence in the batch, or use padding='max_length' to pad to a max length. In this case, you can give a specific length with max_length (e.g. max_length=45) or leave max_length to None to pad to the maximal input size of the model (e.g. 512 for Bert).
FutureWarning,
Batches: 0%| | 0/125 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/opt/rasa/venv/bin/rasa", line 8, in
sys.exit(main())
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa/main.py", line 92, in main
cmdline_arguments.func(cmdline_arguments)
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa/cli/run.py", line 52, in run_actions
sdk.main_from_args(args)
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa_sdk/main.py", line 21, in main_from_args
args.auto_reload,
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa_sdk/endpoint.py", line 137, in run
action_package_name, cors_origins=cors_origins, auto_reload=auto_reload
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa_sdk/endpoint.py", line 80, in create_app
executor.register_package(action_package_name)
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa_sdk/executor.py", line 250, in register_package
self._import_submodules(package)
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa_sdk/executor.py", line 206, in _import_submodules
package = self._import_module(package)
File "/opt/rasa/venv/lib/python3.6/site-packages/rasa_sdk/executor.py", line 227, in _import_module
module = importlib.import_module(name)
File "/usr/local/lib/python3.6/importlib/init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "", line 994, in _gcd_import
File "", line 971, in _find_and_load
File "", line 955, in _find_and_load_unlocked
File "", line 665, in _load_unlocked
File "", line 678, in exec_module
File "", line 219, in _call_with_frames_removed
File "/opt/rasa/rasa-faq/actions.py", line 94, in
encode_standard_question(sentence_transformer_select,pretrained_model)
File "/opt/rasa/rasa-faq/actions.py", line 86, in encode_standard_question
standard_questions_encoder = torch.tensor(bc.encode(standard_questions)).numpy()
File "/opt/rasa/venv/lib/python3.6/site-packages/sentence_transformers/SentenceTransformer.py", line 150, in encode
out_features = self.forward(features)
File "/opt/rasa/venv/lib/python3.6/site-packages/torch/nn/modules/container.py", line 119, in forward
input = module(input)
File "/opt/rasa/venv/lib/python3.6/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/opt/rasa/venv/lib/python3.6/site-packages/sentence_transformers/models/BERT.py", line 33, in forward
output_states = self.bert(**features)
File "/opt/rasa/venv/lib/python3.6/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/opt/rasa/venv/lib/python3.6/site-packages/transformers/models/bert/modeling_bert.py", line 912, in forward
batch_size, seq_length = input_shape
ValueError: not enough values to unpack (expected 2, got 1)

I try to run a simple example with the default data. However, when I use the interactive bot, no matter what I write, the bot stuck after action_listen and it doesn't return any response.

The steps I followed in a Windows laptop:

1. Downloaded this repository
2. Install Bert as A service and run the .sh file

The command shows this

I:WORKER-0:ready and listening!
I:WORKER-1:ready and listening!
I:VENTILATOR:all set, ready to serve request!
I:VENTILATOR:new config request req id: 1       client: b'8e0660af-7981-49c0-a31                           7-b620f0c8fe91'
I:SINK:send config      client b'8e0660af-7981-49c0-a317-b620f0c8fe91'

3. Then I installed both Rasa and Rasa X. For Rasa X, I had to do it with multipass as the official documentation says.
4. Then on the folder, I run the command rasa run actions

I get this response

2020-05-14 11:33:38 INFO     rasa_sdk.endpoint  - Starting action endpoint server...
(1000, 3072)
2020-05-14 11:33:44 INFO     rasa_sdk.executor  - Registered function for 'action_get_answer'.

Then on Rasa X, I added the FAQ.md into the NLU tab and train the model. I set it as active and tried the interactive bot. But as I said, no response with the dots just hanging there as someone is writing.

On what config does this run? Action server crashes for bert large.

Does anyone know the similarity score formula used in the actions.py script?

请教一下standard_questions.npy和standard_questions_len.npy这两个文件是怎么生成的

There is lots of compatibility issues here

I cannot install rasa x 0.20.1 so I had to install newer version, which eventually leads to install Tensorflow 2.x, and bert as a service cannot run on TF >2, only can run with TF <1.15

It's a myth how you can install rasa x 0.20.1...

Updated: I didnt install rasa x anymore but instead use rasa train normally, and then use rasa shell to interact with the chatbot. However, Segmentation Fault appears. I tried on both my Mac machines, same fault

I have installed bert as a service and action working well and launching the rasa x
but the answer not from a bert model i feel it's random and when I let the model to return any answer of bert i get illogical output

(venv) root@rasadev:/opt/rasa/rasa-faq# rasa run actions
2021-05-13 20:01:39 INFO rasa_sdk.endpoint - Starting action endpoint server...
2021-05-13 20:01:39 ERROR rasa_sdk.executor - Failed to register package 'actions'.
Traceback (most recent call last):
File "/opt/rasa/venv/lib/python3.8/site-packages/rasa_sdk/executor.py", line 262, in register_package
self._import_submodules(package)
File "/opt/rasa/venv/lib/python3.8/site-packages/rasa_sdk/executor.py", line 218, in _import_submodules
package = self._import_module(package)
File "/opt/rasa/venv/lib/python3.8/site-packages/rasa_sdk/executor.py", line 239, in _import_module
module = importlib.import_module(name)
File "/usr/lib/python3.8/importlib/init.py", line 127, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "", line 1014, in _gcd_import
File "", line 991, in _find_and_load
File "", line 975, in _find_and_load_unlocked
File "", line 671, in _load_unlocked
File "", line 783, in exec_module
File "", line 219, in _call_with_frames_removed
File "/opt/rasa/rasa-faq/actions.py", line 11, in
import torch
ModuleNotFoundError: No module named 'torch'

Hello Team,

I have to install rasa-faq-bot to amazon ec2 but tensorflow have issues to install while install rasa. Can you send some guidelines to properly install this in linux ami. Looking forward to hear from you.

Hello,

Can you let me know how can i run these on server.

Sending a query from Rasa to the action server generates the following error:
score = np.sum((self.standard_questions_encoder * query_vector), axis=1) / (self.standard_questions_encoder_len * (np.sum(query_vector * query_vector) ** 0.5)) ValueError: operands could not be broadcast together with shapes (1000,3072) (768,)
The very similar code from the bert-as-service example,
score = np.sum(query_vec * doc_vecs, axis=1) / np.linalg.norm(doc_vecs, axis=1)
works as expected.