gemba 0.1.0__py3-none-any.whl → 0.1.2__py3-none-any.whl

gemba/gpt_api.py

@@ -1,12 +1,11 @@
 import os
 import sys
 import time
-import ipdb
 import logging
 from termcolor import colored
-from datetime import datetime
 import openai
 import tqdm
+from concurrent.futures import ThreadPoolExecutor
 
 
 # class for calling OpenAI API and handling cache
@@ -89,6 +88,12 @@ class GptApi:
     def request_api(self, prompt, model, temperature=0, max_tokens=None):
         if temperature > 10:
             return []
+            
+        # Add maximum token limit
+        MAX_TOKENS_LIMIT = 4000  # Adjust this based on your model's context window
+        if max_tokens and max_tokens > MAX_TOKENS_LIMIT:
+            print(f"Reached maximum token limit of {MAX_TOKENS_LIMIT}", file=sys.stderr)
+            return []
 
         while True:
             try:
@@ -123,8 +128,12 @@ class GptApi:
                     print(colored(f"Increasing max tokens to fit answers.", "red") + colored(answer, "blue"), file=sys.stderr)
                 print(f"Finish reason: {choice.finish_reason}", file=sys.stderr)
                 if max_tokens is None:
+                    max_tokens = 500  # Set initial max_tokens if None
+                new_max_tokens = max_tokens + 200
+                if new_max_tokens > MAX_TOKENS_LIMIT:
+                    print(f"Would exceed maximum token limit of {MAX_TOKENS_LIMIT}", file=sys.stderr)
                     return []
-                return self.request_api(prompt, model, temperature=temperature, max_tokens=max_tokens + 200)
+                return self.request_api(prompt, model, temperature=temperature, max_tokens=new_max_tokens)
 
             answers.append({
                 "answer": answer,
@@ -167,8 +176,13 @@ class GptApi:
     
     def bulk_request(self, df, model, parse_mqm_answer, cache, max_tokens=None):
         answers = []
-        for i, row in tqdm.tqdm(df.iterrows(), total=len(df), file=sys.stderr):
-            prompt = row["prompt"]
-            parsed_answers = self.request(prompt, model, parse_mqm_answer, cache=cache, max_tokens=max_tokens)
-            answers += parsed_answers
+        with ThreadPoolExecutor(100) as executor:
+            futures = [
+                executor.submit(self.request, row["prompt"], model, parse_mqm_answer, cache=cache, max_tokens=max_tokens)
+                for _, row in df.iterrows()
+            ]
+            
+            for future in tqdm.tqdm(futures, total=len(df), file=sys.stderr):
+                answers += future.result()
+                
         return answers

{gemba-0.1.0.dist-info → gemba-0.1.2.dist-info}/METADATA

@@ -1,10 +1,11 @@
-Metadata-Version: 2.3
+Metadata-Version: 2.4
 Name: gemba
-Version: 0.1.0
+Version: 0.1.2
 Summary: GEMBA — GPT Estimation Metric Based Assessment
 Project-URL: Homepage, https://github.com/joelniklaus/gemba
 Author-email: Joel Niklaus <joel@niklaus.ai>
-License: MIT
+License-Expression: MIT
+License-File: LICENSE.md
 Classifier: License :: OSI Approved :: MIT License
 Classifier: Operating System :: OS Independent
 Classifier: Programming Language :: Python :: 3
@@ -50,7 +51,7 @@ Install the gemba package with `pip install gemba` and use the following code:
 from gemba import get_gemba_scores
 
 source = ["Hello, how are you?", "I am fine, thank you.", "I am not fine, thank you."]
-hypothesis = ["Hallo, wie geht es dir?", "Ich bin gut, danke.", "Ich bin Adolf, wer bist du?"]
+hypothesis = ["Hallo, wie geht es dir?", "Ich bin gut, danke.", "Ich bin Joel, wer bist du?"]
 source_lang = "en"
 target_lang = "de"
 

{gemba-0.1.0.dist-info → gemba-0.1.2.dist-info}/RECORD

@@ -2,13 +2,13 @@ gemba/__init__.py,sha256=0ZuEumkUMWPI5wQMY7OxLolELI9GYYlup-iJw8SwBgc,67
 gemba/gemba_da.py,sha256=YCOKKP7kZBL9e1d44Zr7aTa23BqLFvh4KDOfbNSMgOU,2360
 gemba/gemba_esa.py,sha256=nBCeFjrS24wXLOcAXHRSmZFYJSkUzRS4hfp2LEqYwp8,4461
 gemba/gemba_mqm_utils.py,sha256=qiIdJv7IDx0eeqpsTCHMoUeo8EUOhG6k-YfrzkRfxyw,9612
-gemba/gpt_api.py,sha256=Igp8uQn6chKL1QWFMqKP2VR9Fbzxm8Xk83ELxk5NfM8,6671
+gemba/gpt_api.py,sha256=UJGXQBnRLBujLGdQhr6HUvbvWYQIxqmQqa_JG8iS0Uc,7394
 gemba/mtme_tools.py,sha256=xpLxCzfnLHFIxsq_LOi1Lpb-gkyFGYqFXiq9y6O315Q,4667
 gemba/prompt.py,sha256=AuPBhO2OBL3EB5I37p-GX10sx29gRw35xFAnB3bqtII,7578
 gemba/scores.py,sha256=FmmBJ-ds-abExphcVUw9qaPMnKttPWobuXNwZKLAtEs,4388
 gemba/testset.py,sha256=tDvi6xQIBXrODg02WWINrYg9jNQqruCmhBrxe9AaK48,1926
 gemba/utils.py,sha256=Re5uW5dcFj3ITWIGpxjXdAKNDKQ7i4H-Tr_s74SQgmk,4311
-gemba-0.1.0.dist-info/METADATA,sha256=9_jYmIPKmAz5cmPn-fTUB7a5xHLbYrlTXpdzhEYaSSw,3692
-gemba-0.1.0.dist-info/WHEEL,sha256=C2FUgwZgiLbznR-k0b_5k3Ai_1aASOXDss3lzCUsUug,87
-gemba-0.1.0.dist-info/licenses/LICENSE.md,sha256=XkNv-P-7d9hgciDpvOIMiRXYYAEP7rbB6-9ahWiOmzk,20137
-gemba-0.1.0.dist-info/RECORD,,
+gemba-0.1.2.dist-info/METADATA,sha256=98Ge9LVScGEzoTyv6gQICfY4KA8V0Gq3927gcEPE5xI,3727
+gemba-0.1.2.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+gemba-0.1.2.dist-info/licenses/LICENSE.md,sha256=XkNv-P-7d9hgciDpvOIMiRXYYAEP7rbB6-9ahWiOmzk,20137
+gemba-0.1.2.dist-info/RECORD,,

{gemba-0.1.0.dist-info → gemba-0.1.2.dist-info}/WHEEL

@@ -1,4 +1,4 @@
 Wheel-Version: 1.0
-Generator: hatchling 1.26.3
+Generator: hatchling 1.27.0
 Root-Is-Purelib: true
 Tag: py3-none-any