run pre commit

Author: KelvinDo183

lmms_eval/tasks/medqa/utils.py

@@ -1,5 +1,5 @@
 import random
-from typing import List, Dict, Any
+from typing import Any, Dict, List
 
 import numpy as np
 
@@ -120,5 +120,3 @@ def _parse_multi_choice_response(response: str, all_choices: List[str]) -> str:
         start_indexes = [response.rfind(f" {can} ") for can in candidates]
         return candidates[int(np.argmax(start_indexes))]
     return candidates[0]
-
-

lmms_eval/tasks/scibench/utils.py

@@ -1,9 +1,8 @@
-from typing import Dict, List, Tuple
 import re
 from math import isclose
+from typing import Dict, List, Tuple
 
-FEWSHOT_PROMPT = (
-    """Problem:
+FEWSHOT_PROMPT = """Problem:
 Suppose that $10.0 \mathrm{~mol} \mathrm{C}_2 \mathrm{H}_6(\mathrm{~g})$ is confined to $4.860 \mathrm{dm}^3$ at $27^{\circ} \mathrm{C}$. Predict the pressure exerted by the ethane from the perfect gas.
 
 Solution:
@@ -23,88 +22,95 @@
 Solution:
 To determine the small angular deviation (ε) of a plumb line from true vertical due to Earth's rotation, we analyze the balance of forces at latitude λ: gravitational force (Fg = mg0) toward Earth's center and centrifugal force (Fc = mRω²cosλ) perpendicular to the rotation axis. The centrifugal force resolves into a vertical component (Fc,v = mRω²cos²λ) that reduces effective gravity to g = g0 - Rω²cos²λ, and a horizontal component (Fc,h = mRω²sinλcosλ) pulling toward the equator. The angular deviation equals the ratio of horizontal force to effective gravity: ε = Rω²sinλcosλ/(g0 - Rω²cos²λ). To find maximum deviation, we differentiate with respect to λ and find it occurs at λ = 45°. Using Earth values (R = 6.371×10⁶ m, ω = 7.292×10⁻⁵ rad/s, g0 ≈ 9.81 m/s²), we calculate the numerator at 45° as 1.697×10⁻² m/s² and denominator as 9.793 m/s², yielding εmax = 1.733×10⁻³ rad or approximately 357 arcseconds (6 arcminutes).
 Final Answer: The final answer is \(\boxed{6}\). I hope it is correct."""
-)
+
 
 def scibench_doc_to_text(doc: Dict, lmms_eval_specific_kwargs: Dict) -> str:
     pre_prompt = lmms_eval_specific_kwargs["pre_prompt"]
     post_prompt = lmms_eval_specific_kwargs["post_prompt"]
     question = doc["problem_text"]
     if doc["unit"].strip():
-        question = question + " The unit of the answer is " + doc["unit"] + "."    
+        question = question + " The unit of the answer is " + doc["unit"] + "."
     return f"{pre_prompt}{question}{post_prompt}"
 
+
 def extract_boxed_answers(text):
     # Find all boxed contents
-    matches = re.findall(r'boxed{([^}]*)}', text)
+    matches = re.findall(r"boxed{([^}]*)}", text)
     for m in matches:
         # Strip spaces
         candidate = m.strip()
         # Keep only the numeric ones (int or decimal, with optional sign)
-        if re.fullmatch(r'[-+]?\d*\.?\d+', candidate):
+        if re.fullmatch(r"[-+]?\d*\.?\d+", candidate):
             return candidate
     return None
 
+
 def remove_not(x):
-    match_number = re.compile('[\$]?\ *10\^[{]?\ *-?[0-9]+\ *[}]?\ *[\$]?')
-    result=re.findall(match_number, x)
-    if len(result) !=0:
+    match_number = re.compile("[\$]?\ *10\^[{]?\ *-?[0-9]+\ *[}]?\ *[\$]?")
+    result = re.findall(match_number, x)
+    if len(result) != 0:
         return re.split(match_number, x)[-1]
     return None
 
-def cal_not(inputs): 
+
+def cal_not(inputs):
     try:
-        x,ab=list(inputs)
-        match_number = re.compile('10\^[{]?\ *-?[0-9]+\ *[}]?')
-        ab=re.findall(match_number, ab)[0]
-        ab=ab[ab.find('^')+1:]
-        if '{' in ab:
-            ab=ab[ab.find('{')+1:]
-        if '}' in ab:
-            ab=ab[:ab.find('}')]
-        x=x.strip()
-        out=float(x)*10**float(ab)
+        x, ab = list(inputs)
+        match_number = re.compile("10\^[{]?\ *-?[0-9]+\ *[}]?")
+        ab = re.findall(match_number, ab)[0]
+        ab = ab[ab.find("^") + 1 :]
+        if "{" in ab:
+            ab = ab[ab.find("{") + 1 :]
+        if "}" in ab:
+            ab = ab[: ab.find("}")]
+        x = x.strip()
+        out = float(x) * 10 ** float(ab)
         # print(float(x)*10**float(ab))
         return str(out)
     except:
-        print('error')
+        print("error")
     return inputs
 
+
 def parse_not(inputs):
     try:
         if not inputs:
-            return '',''
-        if '\\times' in inputs:
-            x,ab=inputs.split('\\times')
-        elif '\times' in inputs:
-            x,ab=inputs.split('\times')
-        elif '*' in inputs:
-            x,ab=inputs.split('*')
+            return "", ""
+        if "\\times" in inputs:
+            x, ab = inputs.split("\\times")
+        elif "\times" in inputs:
+            x, ab = inputs.split("\times")
+        elif "*" in inputs:
+            x, ab = inputs.split("*")
         else:
             return inputs
-        return x,ab
+        return x, ab
     except:
-        return '',''
+        return "", ""
+
 
 def equiv_with_unit(model_output, answer, unit):
-    model_output=model_output.replace(',', '')
+    model_output = model_output.replace(",", "")
     print("Model_output: ", model_output)
     try:
-        ans=float(answer.strip())
-        first=isclose(float(model_output.strip()), ans, rel_tol=0.05)
+        ans = float(answer.strip())
+        first = isclose(float(model_output.strip()), ans, rel_tol=0.05)
     except:
-        first=False
-    try: 
-        model=model_output.strip().split()[0]
-        second=isclose(float(model.strip()), ans, rel_tol=0.05)
+        first = False
+    try:
+        model = model_output.strip().split()[0]
+        second = isclose(float(model.strip()), ans, rel_tol=0.05)
     except:
-        second=False
+        second = False
     if first or second:
         return True
     return False
 
+
 def clean_number_string(s):
     return s.replace(",", "").replace("−", "-").strip()
 
+
 def scibench_process_results(doc: Dict, result: List[str]) -> Dict[str, float]:
     pred = result[0]
     pred = extract_boxed_answers(pred)
@@ -115,11 +121,11 @@ def scibench_process_results(doc: Dict, result: List[str]) -> Dict[str, float]:
         score = 0
     return {"accuracy": score}
 
+
 def scibench_multishot_doc_to_text(doc: Dict, lmms_eval_specific_kwargs: Dict) -> str:
     pre_prompt = lmms_eval_specific_kwargs["pre_prompt"]
     post_prompt = lmms_eval_specific_kwargs["post_prompt"]
     question = doc["problem_text"]
     if doc["unit"].strip():
         question = question + " The unit of the answer is " + doc["unit"] + "."
-    return FEWSHOT_PROMPT + "\n" + question + "\nAnswer: Let's think step by step."    
-    
+    return FEWSHOT_PROMPT + "\n" + question + "\nAnswer: Let's think step by step."