Can't deploy python script to google cloud functions due to issue with flairnlp import

Question

I am trying to deploy a Google Cloud Function that performs sentiment analysis on tweets using a flair nlp model. The code deploys perfectly fine without the line 'import flair' or alternatives like 'from flair import x,y,z'. As soon as I include the import statement for flair the function fails to deploy. Below is the error I get when deploying with the import statement (error is copied from Firebase logs). This is my first time posting on StackOverflow so please pardon me if the post looks ugly.

{"@type":"type.googleapis.com/google.cloud.audit.AuditLog","status":{"code":3,"message":"Function failed on loading user code. This is likely due to a bug in the user code. Error message: Code in file main.py can't be loaded.\nDetailed stack trace:\nTraceback (most recent call last):\n File \"/env/local/lib/python3.7/site-packages/google/cloud/functions/worker_v2.py\", line 359, in check_or_load_user_function\n _function_handler.load_user_function()\n File \"/env/local/lib/python3.7/site-packages/google/cloud/functions/worker_v2.py\", line 236, in load_user_function\n spec.loader.exec_module(main_module)\n File \"<frozen importlib._bootstrap_external>\", line 728, in exec_module\n File \"<frozen importlib._bootstrap>\", line 219, in _call_with_frames_removed\n File \"/user_code/main.py\", line 5, in <module>\n from flair import models, data\n File \"/env/local/lib/python3.7/site-packages/flair/__init__.py\", line 20, in <module>\n from . import models\n File \"/env/local/lib/python3.7/site-packages/flair/models/__init__.py\", line 1, in <module>\n from .sequence_tagger_model import SequenceTagger, MultiTagger\n File \"/env/local/lib/python3.7/site-packages/flair/models/sequence_tagger_model.py\", line 21, in <module>\n from flair.embeddings import TokenEmbeddings, StackedEmbeddings, Embeddings\n File \"/env/local/lib/python3.7/site-packages/flair/embeddings/__init__.py\", line 6, in <module>\n from .token import TokenEmbeddings\n File \"/env/local/lib/python3.7/site-packages/flair/embeddings/token.py\", line 10, in <module>\n from transformers import AutoTokenizer, AutoConfig, AutoModel, CONFIG_MAPPING, PreTrainedTokenizer\nImportError: cannot import name 'AutoModel' from 'transformers' (unknown location)\n. Please visit https://cloud.google.com/functions/docs/troubleshooting for in-depth troubleshooting documentation."},"authenticationInfo":

And here is the script I am trying to deploy, as well as the requirements.txt file

main.py

import firebase_admin
from firebase_admin import credentials
from firebase_admin import firestore
from datetime import datetime, timedelta
import flair
# or from flair import models, data

class FirestoreHandler():
    cred = credentials.Certificate("serviceAccountKey.json")
    firebase_admin.initialize_app(cred)

    db = firestore.client()
    
    def analysis_on_create(self):
            docs = self.db.collection('tweets').order_by(u'time', direction=firestore.Query.DESCENDING).limit(1).get()
            data = docs[0].to_dict()
            most_recent_tweet = data['full-text']
            sentiment_model = flair.models.TextClassifier.load('en-sentiment')
            sentence = flair.data.Sentence(str(most_recent_tweet))
            sentiment_model.predict(sentence)
            result = sentence.labels[0]
            if result.value == "POSITIVE":
                val= 1 * result.score
            else:
                val= -1 * result.score

            self.db.collection('sentiment').add({'sentiment':val,'timestamp':datetime.now()+timedelta(hours=3)}) 
            
    def add_test(self):
        self.db.collection('test3').add({"status":"success", 'timestamp':datetime.now()+timedelta(hours=3)})

def hello_firestore(event, context):
    """Triggered by a change to a Firestore document.
    Args:
         event (dict): Event payload.
         context (google.cloud.functions.Context): Metadata for the event.
    """
    resource_string = context.resource
    # print out the resource string that triggered the function
    print(f"Function triggered by change to: {resource_string}.")
    # now print out the entire event object
    print(str(event))
    fire = FirestoreHandler()
    fire.add_test()
    fire.analysis_on_create()

requirements.txt

# Function dependencies, for example:
# package>=version

firebase-admin==5.0.1
https://download.pytorch.org/whl/cpu/torch-1.0.1.post2-cp37-cp37m-linux_x86_64.whl
flair

I included the url to pytorch download because flair is built on pytorch, and the function would not deploy without the url (even when I didn't import flair in main.py). I have also tried specifying different versions for flair to no avail.

Any intuition as to what may be causing this issue would be greatly appreciated! I am new to the Google Cloud ecosystem, this being my first project. If there is any additional information I can provide please let me know.

Edit: I am deploying from the website (not using CLI)

Answer 1

I am not sure that the provided requirements.txt is OK for GCP cloud functions deployment. Not sure that an explicit https URL is going to be handled correctly...

The Specifying dependencies in Python documentation page describes how the dependencies are to be stated - using the pip package manager's requirements.txt file or packaging local dependencies alongside your function.

Can you simply mention flair with necessary version in the requirements.txt file? Will it work?

In addition, the error you provided highlights that the transformers package is required. Can it be that some specific version is required?

====

As a side comment - I don't know your context and requirements, but I am not sure that in order to work with the Firestore from inside a cloud function all of that

import firebase_admin
from firebase_admin import credentials
from firebase_admin import firestore 

is required, as it might be better to avoid using serviceAccountKey.json at all, and simply assign relevant IAM roles to the service account which is used for the given cloud function execution.