experimental/clip_app_performance_test.py

from concurrent.futures import ThreadPoolExecutor, as_completed
import json
import os
import time

import numpy as np
import requests
import torch

from clip_app_client import ClipAppClient

test_image_url = "https://static.wixstatic.com/media/4d6b49_42b9435ce1104008b1b5f7a3c9bfcd69~mv2.jpg/v1/fill/w_454,h_333,fp_0.50_0.50,q_90/4d6b49_42b9435ce1104008b1b5f7a3c9bfcd69~mv2.jpg"
english_text = (
    "It was the best of times, it was the worst of times, it was the age "
    "of wisdom, it was the age of foolishness, it was the epoch of belief"
)

app_client = ClipAppClient()
preprocessed_image = app_client.preprocess_image(test_image_url)

def _send_text_request(number):
    embeddings = app_client.text_to_embedding(english_text)
    return number, embeddings

def _send_image_url_request(number):
    embeddings = app_client.image_url_to_embedding(test_image_url)
    return number, embeddings

def _send_preprocessed_image_request(number):
    embeddings = app_client.preprocessed_image_to_embedding(preprocessed_image)
    return number, embeddings

def process(numbers, send_func, max_workers=10):
    with ThreadPoolExecutor(max_workers=max_workers) as executor:
        futures = [executor.submit(send_func, number) for number in numbers]
        for future in as_completed(futures):
            n_result, result = future.result()
            print (f"{n_result} : {result.shape}")

if __name__ == "__main__":
    n_calls = 300

    # test text
    numbers = list(range(n_calls))
    start_time = time.monotonic()
    process(numbers, _send_text_request)
    end_time = time.monotonic()
    total_time = end_time - start_time
    avg_time_ms = total_time / n_calls * 1000
    calls_per_sec = n_calls / total_time
    print(f"Text...")
    print(f" Average time taken: {avg_time_ms:.2f} ms")
    print(f" Number of calls per second: {calls_per_sec:.2f}")    

    # test image url
    numbers = list(range(n_calls))
    start_time = time.monotonic()
    process(numbers, _send_image_url_request)
    end_time = time.monotonic()
    total_time = end_time - start_time
    avg_time_ms = total_time / n_calls * 1000
    calls_per_sec = n_calls / total_time
    print(f"Image passing url...")
    print(f" Average time taken: {avg_time_ms:.2f} ms")
    print(f" Number of calls per second: {calls_per_sec:.2f}")    

    # test image as vector
    numbers = list(range(n_calls))
    start_time = time.monotonic()
    process(numbers, _send_preprocessed_image_request)
    end_time = time.monotonic()
    total_time = end_time - start_time
    avg_time_ms = total_time / n_calls * 1000
    calls_per_sec = n_calls / total_time
    print(f"Preprocessed image...")
    print(f" Average time taken: {avg_time_ms:.2f} ms")
    print(f" Number of calls per second: {calls_per_sec:.2f}") 


# from concurrent.futures import ThreadPoolExecutor
# import json
# import os
# import numpy as np
# import requests
# from concurrent.futures import ThreadPoolExecutor, as_completed
# import time

# import torch

# # hack for debugging, set HTTP_ADDRESS to "http://127.0.0.1:8000/"
# # os.environ["HTTP_ADDRESS"] = "http://192.168.7.79:8000"

# test_image_url = "https://static.wixstatic.com/media/4d6b49_42b9435ce1104008b1b5f7a3c9bfcd69~mv2.jpg/v1/fill/w_454,h_333,fp_0.50_0.50,q_90/4d6b49_42b9435ce1104008b1b5f7a3c9bfcd69~mv2.jpg"
# english_text = (
#     "It was the best of times, it was the worst of times, it was the age "
#     "of wisdom, it was the age of foolishness, it was the epoch of belief"
# )

# preprocessed_image = preprocess_image(test_image_url)

# def _send_text_request(number):
#     embeddings = text_to_embedding(english_text)
#     return number, embeddings

# def _send_image_url_request(number):
#     embeddings = image_url_to_embedding(test_image_url)
#     return number, embeddings

# def _send_preprocessed_image_request(number):
#     embeddings = preprocessed_image_to_embedding(preprocessed_image)
#     return number, embeddings

# def process(numbers, send_func, max_workers=10):
#     with ThreadPoolExecutor(max_workers=max_workers) as executor:
#         futures = [executor.submit(send_func, number) for number in numbers]
#         for future in as_completed(futures):
#             n_result, result = future.result()
#             result = json.loads(result)
#             print (f"{n_result} : {len(result[0])}")

# # def process_text(numbers, max_workers=10):
# #     for n in numbers:
# #         n_result, result = send_text_request(n)
# #         result = json.loads(result)
# #         print (f"{n_result} : {len(result[0])}")

# if __name__ == "__main__":
#     n_calls = 300

#     # test text
#     # n_calls = 1
#     numbers = list(range(n_calls))
#     start_time = time.monotonic()
#     process(numbers, _send_text_request)
#     end_time = time.monotonic()
#     total_time = end_time - start_time
#     avg_time_ms = total_time / n_calls * 1000
#     calls_per_sec = n_calls / total_time
#     print(f"Text...")
#     print(f" Average time taken: {avg_time_ms:.2f} ms")
#     print(f" Number of calls per second: {calls_per_sec:.2f}")    

#     # test image url
#     # n_calls = 1
#     numbers = list(range(n_calls))
#     start_time = time.monotonic()
#     process(numbers, _send_image_url_request)
#     end_time = time.monotonic()
#     total_time = end_time - start_time
#     avg_time_ms = total_time / n_calls * 1000
#     calls_per_sec = n_calls / total_time
#     print(f"Image passing url...")
#     print(f" Average time taken: {avg_time_ms:.2f} ms")
#     print(f" Number of calls per second: {calls_per_sec:.2f}")    

#     # test image as vector
#     # n_calls = 1
#     numbers = list(range(n_calls))
#     start_time = time.monotonic()
#     process(numbers, _send_preprocessed_image_request)
#     end_time = time.monotonic()
#     total_time = end_time - start_time
#     avg_time_ms = total_time / n_calls * 1000
#     calls_per_sec = n_calls / total_time
#     print(f"Preprocessed image...")
#     print(f" Average time taken: {avg_time_ms:.2f} ms")
#     print(f" Number of calls per second: {calls_per_sec:.2f}")