Separation of response and score in predictions.jsonl and summary.jsonl
Browse files- src/benchmark_runner.py +101 -41
- src/evaluation.py +69 -14
src/benchmark_runner.py
CHANGED
@@ -93,27 +93,35 @@ def load_config(config_path: str) -> dict:
|
|
93 |
|
94 |
def append_prediction(result: Dict[str, Any], filepath: str):
|
95 |
"""Appends a single prediction result to a JSONL file."""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
96 |
try:
|
97 |
with open(filepath, 'a') as f:
|
98 |
-
json.dump(
|
99 |
f.write('\n')
|
100 |
except IOError as e:
|
101 |
logging.error(f"Failed to append prediction to {filepath}: {e}")
|
102 |
except Exception as e:
|
103 |
logging.error(f"Unexpected error appending prediction to {filepath}: {e}")
|
104 |
|
105 |
-
|
106 |
-
|
107 |
-
"""Saves the final summary dictionary to a JSON file."""
|
108 |
try:
|
109 |
-
with open(filepath, '
|
110 |
-
json.dump(
|
111 |
-
|
112 |
except IOError as e:
|
113 |
-
logging.error(f"Failed to
|
114 |
except Exception as e:
|
115 |
-
logging.error(f"Unexpected error
|
|
|
116 |
|
|
|
117 |
|
118 |
def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
119 |
"""Generates a human-readable Markdown summary from the results dictionary."""
|
@@ -125,7 +133,6 @@ def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
|
125 |
timestamp = summary.get("timestamp", "N/A")
|
126 |
total_questions_in_dataset = summary.get("total_questions_in_dataset", 0)
|
127 |
total_questions_processed_in_run = summary.get("total_questions_processed_in_run", 0)
|
128 |
-
# total_api_cost = summary.get("total_api_cost", 0.0) # Removed
|
129 |
|
130 |
filtered_questions_count = 0
|
131 |
if total_questions_in_dataset > 0 and total_questions_processed_in_run > 0:
|
@@ -149,14 +156,8 @@ def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
|
149 |
if "overall_score" in summary and "section_breakdown" in summary: # Generic check for score-based summary
|
150 |
total_processed = summary.get("total_questions_processed", 0)
|
151 |
|
152 |
-
# Max score calculation is complex due to varied scoring per question.
|
153 |
-
# For now, we'll omit max_score from the overall display or acknowledge its variability.
|
154 |
-
# A more accurate max_score would require iterating through the dataset items used in the run
|
155 |
-
# and summing their individual max possible scores based on exam_name and question_type.
|
156 |
-
# This is out of scope for the current summary generation simplicity.
|
157 |
-
max_score_display = "N/A (variable per question)" # Placeholder
|
158 |
-
|
159 |
overall_score = summary.get('overall_score', 'N/A')
|
|
|
160 |
correct_full_count = summary.get('overall_correct_full', 'N/A')
|
161 |
partial_correct_count = summary.get('overall_partial_correct', 'N/A')
|
162 |
incorrect_choice_count = summary.get('overall_incorrect_choice', 'N/A')
|
@@ -164,10 +165,10 @@ def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
|
164 |
failures_count = summary.get('overall_api_parse_failures', 'N/A')
|
165 |
unmapped_count = summary.get('unmapped_section_questions', 'N/A')
|
166 |
|
167 |
-
md_content.append("## Exam Scoring Results")
|
168 |
-
md_content.append(f"**Overall Score:** **{overall_score}**
|
169 |
md_content.append(f"- **Fully Correct Answers:** {correct_full_count}")
|
170 |
-
if partial_correct_count != 'N/A' and partial_correct_count > 0 :
|
171 |
md_content.append(f"- **Partially Correct Answers:** {partial_correct_count}")
|
172 |
md_content.append(f"- **Incorrectly Answered (Choice Made):** {incorrect_choice_count}")
|
173 |
md_content.append(f"- **Skipped Questions:** {skipped_count}")
|
@@ -176,6 +177,53 @@ def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
|
176 |
if unmapped_count > 0:
|
177 |
md_content.append(f"- **Unmapped Section Questions:** {unmapped_count} *(Not included in section breakdown)*")
|
178 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
179 |
md_content.append("\n### Section Breakdown")
|
180 |
md_content.append("| Section | Score | Fully Correct | Partially Correct | Incorrect Choice | Skipped | API/Parse Failures |")
|
181 |
md_content.append("|---------------|-------|---------------|-------------------|------------------|---------|--------------------|")
|
@@ -189,9 +237,9 @@ def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
|
189 |
for section_name in sorted_section_names:
|
190 |
stats = section_breakdown.get(section_name, {})
|
191 |
score = stats.get('score', 'N/A')
|
192 |
-
s_correct = stats.get('correct', 'N/A')
|
193 |
s_partial = stats.get('partial_correct', 'N/A')
|
194 |
-
s_incorrect = stats.get('incorrect', 'N/A')
|
195 |
s_skipped = stats.get('skipped', 'N/A')
|
196 |
s_failures = stats.get('api_parse_failures', 'N/A')
|
197 |
display_section_name = section_name.replace('_', ' ')
|
@@ -200,7 +248,7 @@ def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
|
200 |
md_content.append("| No section data available | N/A | N/A | N/A | N/A | N/A | N/A |")
|
201 |
|
202 |
# Fallback for simple accuracy (if exam scoring wasn't applicable or failed)
|
203 |
-
elif "accuracy_on_parsed" in summary:
|
204 |
md_content.append("## Simple Accuracy Results (Fallback)")
|
205 |
md_content.append(f"- **Accuracy (on successfully parsed non-skipped):** {summary.get('accuracy_on_parsed', 'N/A'):.4f}")
|
206 |
md_content.append(f"- **Total Processed Attempts:** {summary.get('total_processed_attempts', 'N/A')}")
|
@@ -320,7 +368,7 @@ def run_benchmark(
|
|
320 |
model_output_dir = os.path.join(base_output_dir, model_output_dir_name)
|
321 |
os.makedirs(model_output_dir, exist_ok=True)
|
322 |
predictions_path = os.path.join(model_output_dir, "predictions.jsonl")
|
323 |
-
|
324 |
markdown_summary_path = os.path.join(model_output_dir, "summary.md") # Define path for MD summary
|
325 |
logging.info(f"Results for {model_id} will be saved to: {model_output_dir}")
|
326 |
|
@@ -447,8 +495,19 @@ def run_benchmark(
|
|
447 |
result_data['marks_awarded'] = score_details.get('marks_awarded')
|
448 |
result_data['evaluation_status'] = score_details.get('evaluation_status')
|
449 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
450 |
model_results.append(result_data)
|
451 |
-
append_prediction(result_data, predictions_path)
|
452 |
|
453 |
final_parsed_answer = result_data["predicted_answer"]
|
454 |
log_message_prefix = f"Question {question_id}:"
|
@@ -626,8 +685,19 @@ def run_benchmark(
|
|
626 |
result_data_retry['marks_awarded'] = score_details_retry.get('marks_awarded')
|
627 |
result_data_retry['evaluation_status'] = score_details_retry.get('evaluation_status')
|
628 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
629 |
model_results.append(result_data_retry)
|
630 |
-
append_prediction(result_data_retry, predictions_path)
|
631 |
|
632 |
# Logging for retry pass
|
633 |
log_message_prefix_retry = f"Question {question_id_retry} (Retry):"
|
@@ -709,24 +779,14 @@ def run_benchmark(
|
|
709 |
logging.info(json.dumps(summary, indent=2, sort_keys=True))
|
710 |
logging.info("-------------------------------------")
|
711 |
|
712 |
-
#
|
713 |
-
# The model_results list was modified in-place by calculate_neet_scores (if applicable)
|
714 |
# to include evaluation_status and marks_awarded.
|
715 |
-
|
716 |
-
|
717 |
-
with open(predictions_path, 'w') as f:
|
718 |
-
for final_result in model_results:
|
719 |
-
json.dump(final_result, f)
|
720 |
-
f.write('\n')
|
721 |
-
logging.info(f"Successfully updated {predictions_path} with evaluation details.")
|
722 |
-
except IOError as e:
|
723 |
-
logging.error(f"Failed to overwrite predictions file {predictions_path}: {e}")
|
724 |
-
except Exception as e:
|
725 |
-
logging.error(f"Unexpected error overwriting predictions file {predictions_path}: {e}")
|
726 |
|
727 |
|
728 |
-
# Save final summary (
|
729 |
-
|
730 |
generate_markdown_summary(summary, markdown_summary_path) # Call the new function
|
731 |
|
732 |
logging.info("Benchmark run completed.")
|
|
|
93 |
|
94 |
def append_prediction(result: Dict[str, Any], filepath: str):
|
95 |
"""Appends a single prediction result to a JSONL file."""
|
96 |
+
# Create a copy to avoid modifying the original dict that might be used elsewhere
|
97 |
+
# and remove evaluation-specific fields before saving to predictions.jsonl
|
98 |
+
prediction_data = result.copy()
|
99 |
+
prediction_data.pop('marks_awarded', None)
|
100 |
+
prediction_data.pop('evaluation_status', None)
|
101 |
+
prediction_data.pop('predicted_answer', None) # Remove predicted_answer
|
102 |
+
prediction_data.pop('ground_truth', None) # Remove ground_truth
|
103 |
try:
|
104 |
with open(filepath, 'a') as f:
|
105 |
+
json.dump(prediction_data, f)
|
106 |
f.write('\n')
|
107 |
except IOError as e:
|
108 |
logging.error(f"Failed to append prediction to {filepath}: {e}")
|
109 |
except Exception as e:
|
110 |
logging.error(f"Unexpected error appending prediction to {filepath}: {e}")
|
111 |
|
112 |
+
def append_summary_detail(result_detail: Dict[str, Any], filepath: str):
|
113 |
+
"""Appends a single question's summary details (evaluation status, marks, predicted, truth) to a JSONL file."""
|
|
|
114 |
try:
|
115 |
+
with open(filepath, 'a') as f:
|
116 |
+
json.dump(result_detail, f)
|
117 |
+
f.write('\n')
|
118 |
except IOError as e:
|
119 |
+
logging.error(f"Failed to append summary detail to {filepath}: {e}")
|
120 |
except Exception as e:
|
121 |
+
logging.error(f"Unexpected error appending summary detail to {filepath}: {e}")
|
122 |
+
|
123 |
|
124 |
+
# Removed save_summary function as summary.json is no longer needed.
|
125 |
|
126 |
def generate_markdown_summary(summary: Dict[str, Any], filepath: str):
|
127 |
"""Generates a human-readable Markdown summary from the results dictionary."""
|
|
|
133 |
timestamp = summary.get("timestamp", "N/A")
|
134 |
total_questions_in_dataset = summary.get("total_questions_in_dataset", 0)
|
135 |
total_questions_processed_in_run = summary.get("total_questions_processed_in_run", 0)
|
|
|
136 |
|
137 |
filtered_questions_count = 0
|
138 |
if total_questions_in_dataset > 0 and total_questions_processed_in_run > 0:
|
|
|
156 |
if "overall_score" in summary and "section_breakdown" in summary: # Generic check for score-based summary
|
157 |
total_processed = summary.get("total_questions_processed", 0)
|
158 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
159 |
overall_score = summary.get('overall_score', 'N/A')
|
160 |
+
total_possible_score = summary.get('total_possible_score_for_processed_questions', 'N/A')
|
161 |
correct_full_count = summary.get('overall_correct_full', 'N/A')
|
162 |
partial_correct_count = summary.get('overall_partial_correct', 'N/A')
|
163 |
incorrect_choice_count = summary.get('overall_incorrect_choice', 'N/A')
|
|
|
165 |
failures_count = summary.get('overall_api_parse_failures', 'N/A')
|
166 |
unmapped_count = summary.get('unmapped_section_questions', 'N/A')
|
167 |
|
168 |
+
md_content.append("## Exam Scoring Results")
|
169 |
+
md_content.append(f"**Overall Score:** **{overall_score}** / **{total_possible_score}**")
|
170 |
md_content.append(f"- **Fully Correct Answers:** {correct_full_count}")
|
171 |
+
if partial_correct_count != 'N/A' and partial_correct_count > 0 :
|
172 |
md_content.append(f"- **Partially Correct Answers:** {partial_correct_count}")
|
173 |
md_content.append(f"- **Incorrectly Answered (Choice Made):** {incorrect_choice_count}")
|
174 |
md_content.append(f"- **Skipped Questions:** {skipped_count}")
|
|
|
177 |
if unmapped_count > 0:
|
178 |
md_content.append(f"- **Unmapped Section Questions:** {unmapped_count} *(Not included in section breakdown)*")
|
179 |
|
180 |
+
md_content.append("\n### Detailed Score Calculation by Question Type")
|
181 |
+
question_type_breakdown = summary.get("question_type_breakdown", {})
|
182 |
+
if question_type_breakdown:
|
183 |
+
sorted_q_types = sorted(question_type_breakdown.keys())
|
184 |
+
for q_type in sorted_q_types:
|
185 |
+
stats = question_type_breakdown[q_type]
|
186 |
+
q_type_display = q_type.replace('_', ' ').title()
|
187 |
+
max_score_per_q = stats.get('max_score_per_question', 0)
|
188 |
+
|
189 |
+
correct_count_q = stats.get('correct_full', 0)
|
190 |
+
partial_count_q = stats.get('partial_correct', 0)
|
191 |
+
incorrect_count_q = stats.get('incorrect_choice', 0)
|
192 |
+
skipped_count_q = stats.get('skipped', 0)
|
193 |
+
api_fail_count_q = stats.get('api_parse_failures', 0)
|
194 |
+
score_q = stats.get('score', 0)
|
195 |
+
|
196 |
+
calculation_parts = []
|
197 |
+
if correct_count_q > 0:
|
198 |
+
calculation_parts.append(f"{correct_count_q} Correct (+{max_score_per_q})")
|
199 |
+
if partial_count_q > 0:
|
200 |
+
# For partial, we can't easily show the exact score per question without more detail
|
201 |
+
# For now, just indicate partials.
|
202 |
+
calculation_parts.append(f"{partial_count_q} Partial")
|
203 |
+
if incorrect_count_q > 0:
|
204 |
+
# Need to know penalty for incorrect. Assuming -1 for MCQ_SINGLE_CORRECT, -2 for MCQ_MULTIPLE_CORRECT
|
205 |
+
# For INTEGER, penalty is 0. This needs to be more robust if penalties vary.
|
206 |
+
penalty_per_incorrect = 0
|
207 |
+
if q_type == "MCQ_SINGLE_CORRECT": penalty_per_incorrect = -1
|
208 |
+
elif q_type == "MCQ_MULTIPLE_CORRECT": penalty_per_incorrect = -2
|
209 |
+
calculation_parts.append(f"{incorrect_count_q} Incorrect ({penalty_per_incorrect})")
|
210 |
+
if skipped_count_q > 0:
|
211 |
+
calculation_parts.append(f"{skipped_count_q} Skipped (0)")
|
212 |
+
if api_fail_count_q > 0:
|
213 |
+
# Assuming -1 for API/Parse failures for non-integer types, 0 for integer
|
214 |
+
penalty_per_api_fail = -1
|
215 |
+
if q_type == "INTEGER": penalty_per_api_fail = 0
|
216 |
+
calculation_parts.append(f"{api_fail_count_q} API/Parse Fail ({penalty_per_api_fail})")
|
217 |
+
|
218 |
+
calculation_str = " + ".join(part for part in calculation_parts if part)
|
219 |
+
if not calculation_str:
|
220 |
+
calculation_str = "No questions of this type processed or all had 0 score change."
|
221 |
+
|
222 |
+
md_content.append(f"**{q_type_display} ({stats.get('count', 0)} questions):** {score_q} marks")
|
223 |
+
md_content.append(f" *Calculation:* {calculation_str} = {score_q}")
|
224 |
+
else:
|
225 |
+
md_content.append("No question type breakdown available.")
|
226 |
+
|
227 |
md_content.append("\n### Section Breakdown")
|
228 |
md_content.append("| Section | Score | Fully Correct | Partially Correct | Incorrect Choice | Skipped | API/Parse Failures |")
|
229 |
md_content.append("|---------------|-------|---------------|-------------------|------------------|---------|--------------------|")
|
|
|
237 |
for section_name in sorted_section_names:
|
238 |
stats = section_breakdown.get(section_name, {})
|
239 |
score = stats.get('score', 'N/A')
|
240 |
+
s_correct = stats.get('correct', 'N/A')
|
241 |
s_partial = stats.get('partial_correct', 'N/A')
|
242 |
+
s_incorrect = stats.get('incorrect', 'N/A')
|
243 |
s_skipped = stats.get('skipped', 'N/A')
|
244 |
s_failures = stats.get('api_parse_failures', 'N/A')
|
245 |
display_section_name = section_name.replace('_', ' ')
|
|
|
248 |
md_content.append("| No section data available | N/A | N/A | N/A | N/A | N/A | N/A |")
|
249 |
|
250 |
# Fallback for simple accuracy (if exam scoring wasn't applicable or failed)
|
251 |
+
elif "accuracy_on_parsed" in summary:
|
252 |
md_content.append("## Simple Accuracy Results (Fallback)")
|
253 |
md_content.append(f"- **Accuracy (on successfully parsed non-skipped):** {summary.get('accuracy_on_parsed', 'N/A'):.4f}")
|
254 |
md_content.append(f"- **Total Processed Attempts:** {summary.get('total_processed_attempts', 'N/A')}")
|
|
|
368 |
model_output_dir = os.path.join(base_output_dir, model_output_dir_name)
|
369 |
os.makedirs(model_output_dir, exist_ok=True)
|
370 |
predictions_path = os.path.join(model_output_dir, "predictions.jsonl")
|
371 |
+
summary_details_path = os.path.join(model_output_dir, "summary.jsonl") # New file for per-question summary details
|
372 |
markdown_summary_path = os.path.join(model_output_dir, "summary.md") # Define path for MD summary
|
373 |
logging.info(f"Results for {model_id} will be saved to: {model_output_dir}")
|
374 |
|
|
|
495 |
result_data['marks_awarded'] = score_details.get('marks_awarded')
|
496 |
result_data['evaluation_status'] = score_details.get('evaluation_status')
|
497 |
|
498 |
+
# Append evaluation details to summary.jsonl
|
499 |
+
summary_detail_data = {
|
500 |
+
"question_id": question_id,
|
501 |
+
"marks_awarded": result_data['marks_awarded'],
|
502 |
+
"evaluation_status": result_data['evaluation_status'],
|
503 |
+
"predicted_answer": result_data['predicted_answer'], # Add predicted_answer
|
504 |
+
"ground_truth": result_data['ground_truth'], # Add ground_truth
|
505 |
+
"attempt": result_data['attempt']
|
506 |
+
}
|
507 |
+
append_summary_detail(summary_detail_data, summary_details_path)
|
508 |
+
|
509 |
model_results.append(result_data)
|
510 |
+
append_prediction(result_data, predictions_path) # append_prediction now handles removing evaluation fields
|
511 |
|
512 |
final_parsed_answer = result_data["predicted_answer"]
|
513 |
log_message_prefix = f"Question {question_id}:"
|
|
|
685 |
result_data_retry['marks_awarded'] = score_details_retry.get('marks_awarded')
|
686 |
result_data_retry['evaluation_status'] = score_details_retry.get('evaluation_status')
|
687 |
|
688 |
+
# Append evaluation details to summary.jsonl for retry pass
|
689 |
+
summary_detail_data_retry = {
|
690 |
+
"question_id": question_id_retry,
|
691 |
+
"marks_awarded": result_data_retry['marks_awarded'],
|
692 |
+
"evaluation_status": result_data_retry['evaluation_status'],
|
693 |
+
"predicted_answer": result_data_retry['predicted_answer'], # Add predicted_answer
|
694 |
+
"ground_truth": result_data_retry['ground_truth'], # Add ground_truth
|
695 |
+
"attempt": result_data_retry['attempt']
|
696 |
+
}
|
697 |
+
append_summary_detail(summary_detail_data_retry, summary_details_path)
|
698 |
+
|
699 |
model_results.append(result_data_retry)
|
700 |
+
append_prediction(result_data_retry, predictions_path) # append_prediction now handles removing evaluation fields
|
701 |
|
702 |
# Logging for retry pass
|
703 |
log_message_prefix_retry = f"Question {question_id_retry} (Retry):"
|
|
|
779 |
logging.info(json.dumps(summary, indent=2, sort_keys=True))
|
780 |
logging.info("-------------------------------------")
|
781 |
|
782 |
+
# The model_results list was modified in-place by calculate_exam_scores
|
|
|
783 |
# to include evaluation_status and marks_awarded.
|
784 |
+
# predictions.jsonl is now written incrementally without evaluation details.
|
785 |
+
# No need to overwrite predictions.jsonl here.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
786 |
|
787 |
|
788 |
+
# Save final summary (Markdown) for the current model
|
789 |
+
# The summary.json file is no longer generated as per user request.
|
790 |
generate_markdown_summary(summary, markdown_summary_path) # Call the new function
|
791 |
|
792 |
logging.info("Benchmark run completed.")
|
src/evaluation.py
CHANGED
@@ -201,9 +201,34 @@ def calculate_single_question_score_details(result_item: Dict[str, Any]) -> Dict
|
|
201 |
return {"marks_awarded": current_score_change, "evaluation_status": evaluation_status}
|
202 |
|
203 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
204 |
def calculate_exam_scores(results: List[Dict[str, Any]]) -> Dict[str, Any]:
|
205 |
"""
|
206 |
-
Calculates exam scores based on exam_name and question_type, providing section-wise breakdown
|
|
|
207 |
|
208 |
Args:
|
209 |
results (List[Dict[str, Any]]): A list of result dictionaries. Each dict must contain:
|
@@ -217,14 +242,25 @@ def calculate_exam_scores(results: List[Dict[str, Any]]) -> Dict[str, Any]:
|
|
217 |
This list will be modified in-place to add 'evaluation_status' and 'marks_awarded' by calling
|
218 |
calculate_single_question_score_details for each item.
|
219 |
Returns:
|
220 |
-
Dict[str, Any]: A dictionary containing overall and section-wise scores and counts
|
|
|
221 |
"""
|
222 |
if not results:
|
223 |
return {"error": "No results provided."}
|
224 |
|
225 |
-
overall_stats = {
|
226 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
227 |
|
|
|
|
|
|
|
228 |
valid_subjects_from_data = [r.get("subject") for r in results if r.get("subject") and isinstance(r.get("subject"), str) and r.get("subject").strip()]
|
229 |
if not valid_subjects_from_data and results:
|
230 |
logging.warning("No valid subjects found in results data to initialize section_stats.")
|
@@ -242,6 +278,8 @@ def calculate_exam_scores(results: List[Dict[str, Any]]) -> Dict[str, Any]:
|
|
242 |
for result in results:
|
243 |
question_id = result.get("question_id") # For logging within loop
|
244 |
subject = result.get("subject") # For section mapping
|
|
|
|
|
245 |
|
246 |
# Calculate score details for the single question
|
247 |
score_details = calculate_single_question_score_details(result)
|
@@ -252,18 +290,12 @@ def calculate_exam_scores(results: List[Dict[str, Any]]) -> Dict[str, Any]:
|
|
252 |
result['evaluation_status'] = evaluation_status
|
253 |
result['marks_awarded'] = current_score_change
|
254 |
|
255 |
-
# Accumulate
|
256 |
-
|
257 |
-
# if isinstance(current_api_cost, (int, float)): # Removed
|
258 |
-
# total_api_cost += current_api_cost # Removed
|
259 |
-
# elif current_api_cost is not None: # Removed
|
260 |
-
# logging.warning(f"Invalid api_cost type for QID {question_id}: {current_api_cost} (type: {type(current_api_cost)}). Skipping cost accumulation for this item.") # Removed
|
261 |
|
262 |
# Determine boolean flags based on evaluation_status for aggregation
|
263 |
is_correct_full = evaluation_status in ["correct", "correct_full"]
|
264 |
is_partial_correct = evaluation_status.startswith("partial_")
|
265 |
-
# is_incorrect_choice needs to be inferred carefully if not directly returned by single calc
|
266 |
-
# For simplicity, we'll rely on the status. More robust would be for single_calc to return these bools too.
|
267 |
is_incorrect_choice = evaluation_status in ["incorrect", "incorrect_negative"]
|
268 |
is_skipped = evaluation_status == "skipped"
|
269 |
is_api_parse_failure = evaluation_status in ["failure_api_or_parse", "failure_unexpected_type", "error_bad_ground_truth"]
|
@@ -276,6 +308,28 @@ def calculate_exam_scores(results: List[Dict[str, Any]]) -> Dict[str, Any]:
|
|
276 |
if is_api_parse_failure: overall_stats["api_parse_failures"] += 1
|
277 |
if is_partial_correct: overall_stats["partial_correct"] +=1
|
278 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
279 |
section = None
|
280 |
if subject:
|
281 |
question_num_for_log = -1 # Placeholder, as QID might not have number
|
@@ -309,9 +363,10 @@ def calculate_exam_scores(results: List[Dict[str, Any]]) -> Dict[str, Any]:
|
|
309 |
"overall_skipped": overall_stats["skipped"],
|
310 |
"overall_api_parse_failures": overall_stats["api_parse_failures"],
|
311 |
"total_questions_processed": len(results),
|
312 |
-
|
313 |
"unmapped_section_questions": unmapped_section_questions,
|
314 |
-
"section_breakdown": section_stats
|
|
|
315 |
}
|
316 |
|
317 |
|
|
|
201 |
return {"marks_awarded": current_score_change, "evaluation_status": evaluation_status}
|
202 |
|
203 |
|
204 |
+
def calculate_max_score_for_question(exam_name: str, question_type: str) -> int:
|
205 |
+
"""
|
206 |
+
Returns the maximum possible score for a given exam and question type.
|
207 |
+
"""
|
208 |
+
exam_name = exam_name.upper()
|
209 |
+
question_type = question_type.upper()
|
210 |
+
|
211 |
+
if exam_name == "NEET" and question_type == "MCQ_SINGLE_CORRECT":
|
212 |
+
return 4
|
213 |
+
elif exam_name == "JEE_MAIN":
|
214 |
+
if question_type == "MCQ_SINGLE_CORRECT":
|
215 |
+
return 4
|
216 |
+
elif question_type == "INTEGER":
|
217 |
+
return 4
|
218 |
+
elif exam_name == "JEE_ADVANCED":
|
219 |
+
if question_type == "MCQ_SINGLE_CORRECT":
|
220 |
+
return 3
|
221 |
+
elif question_type == "INTEGER":
|
222 |
+
return 4
|
223 |
+
elif question_type == "MCQ_MULTIPLE_CORRECT":
|
224 |
+
return 4 # Max score for multiple correct is 4
|
225 |
+
return 0 # Default for unknown types
|
226 |
+
|
227 |
+
|
228 |
def calculate_exam_scores(results: List[Dict[str, Any]]) -> Dict[str, Any]:
|
229 |
"""
|
230 |
+
Calculates exam scores based on exam_name and question_type, providing section-wise breakdown
|
231 |
+
and detailed question type statistics.
|
232 |
|
233 |
Args:
|
234 |
results (List[Dict[str, Any]]): A list of result dictionaries. Each dict must contain:
|
|
|
242 |
This list will be modified in-place to add 'evaluation_status' and 'marks_awarded' by calling
|
243 |
calculate_single_question_score_details for each item.
|
244 |
Returns:
|
245 |
+
Dict[str, Any]: A dictionary containing overall and section-wise scores and counts,
|
246 |
+
plus question type breakdowns and total possible score.
|
247 |
"""
|
248 |
if not results:
|
249 |
return {"error": "No results provided."}
|
250 |
|
251 |
+
overall_stats = {
|
252 |
+
"score": 0,
|
253 |
+
"correct": 0,
|
254 |
+
"incorrect": 0,
|
255 |
+
"skipped": 0,
|
256 |
+
"api_parse_failures": 0,
|
257 |
+
"partial_correct": 0,
|
258 |
+
"total_possible_score": 0 # New field
|
259 |
+
}
|
260 |
|
261 |
+
# Initialize question type breakdown
|
262 |
+
question_type_breakdown: Dict[str, Dict[str, Any]] = {}
|
263 |
+
|
264 |
valid_subjects_from_data = [r.get("subject") for r in results if r.get("subject") and isinstance(r.get("subject"), str) and r.get("subject").strip()]
|
265 |
if not valid_subjects_from_data and results:
|
266 |
logging.warning("No valid subjects found in results data to initialize section_stats.")
|
|
|
278 |
for result in results:
|
279 |
question_id = result.get("question_id") # For logging within loop
|
280 |
subject = result.get("subject") # For section mapping
|
281 |
+
exam_name = result.get("exam_name", "").upper()
|
282 |
+
question_type = result.get("question_type", "").upper()
|
283 |
|
284 |
# Calculate score details for the single question
|
285 |
score_details = calculate_single_question_score_details(result)
|
|
|
290 |
result['evaluation_status'] = evaluation_status
|
291 |
result['marks_awarded'] = current_score_change
|
292 |
|
293 |
+
# Accumulate total possible score
|
294 |
+
overall_stats["total_possible_score"] += calculate_max_score_for_question(exam_name, question_type)
|
|
|
|
|
|
|
|
|
295 |
|
296 |
# Determine boolean flags based on evaluation_status for aggregation
|
297 |
is_correct_full = evaluation_status in ["correct", "correct_full"]
|
298 |
is_partial_correct = evaluation_status.startswith("partial_")
|
|
|
|
|
299 |
is_incorrect_choice = evaluation_status in ["incorrect", "incorrect_negative"]
|
300 |
is_skipped = evaluation_status == "skipped"
|
301 |
is_api_parse_failure = evaluation_status in ["failure_api_or_parse", "failure_unexpected_type", "error_bad_ground_truth"]
|
|
|
308 |
if is_api_parse_failure: overall_stats["api_parse_failures"] += 1
|
309 |
if is_partial_correct: overall_stats["partial_correct"] +=1
|
310 |
|
311 |
+
# Aggregate by question type
|
312 |
+
if question_type not in question_type_breakdown:
|
313 |
+
question_type_breakdown[question_type] = {
|
314 |
+
"count": 0,
|
315 |
+
"score": 0,
|
316 |
+
"correct_full": 0,
|
317 |
+
"partial_correct": 0,
|
318 |
+
"incorrect_choice": 0,
|
319 |
+
"skipped": 0,
|
320 |
+
"api_parse_failures": 0,
|
321 |
+
"max_score_per_question": calculate_max_score_for_question(exam_name, question_type)
|
322 |
+
}
|
323 |
+
|
324 |
+
q_type_stats = question_type_breakdown[question_type]
|
325 |
+
q_type_stats["count"] += 1
|
326 |
+
q_type_stats["score"] += current_score_change
|
327 |
+
if is_correct_full: q_type_stats["correct_full"] += 1
|
328 |
+
if is_incorrect_choice: q_type_stats["incorrect_choice"] += 1
|
329 |
+
if is_skipped: q_type_stats["skipped"] += 1
|
330 |
+
if is_api_parse_failure: q_type_stats["api_parse_failures"] += 1
|
331 |
+
if is_partial_correct: q_type_stats["partial_correct"] += 1
|
332 |
+
|
333 |
section = None
|
334 |
if subject:
|
335 |
question_num_for_log = -1 # Placeholder, as QID might not have number
|
|
|
363 |
"overall_skipped": overall_stats["skipped"],
|
364 |
"overall_api_parse_failures": overall_stats["api_parse_failures"],
|
365 |
"total_questions_processed": len(results),
|
366 |
+
"total_possible_score_for_processed_questions": overall_stats["total_possible_score"], # New field
|
367 |
"unmapped_section_questions": unmapped_section_questions,
|
368 |
+
"section_breakdown": section_stats,
|
369 |
+
"question_type_breakdown": question_type_breakdown # New field
|
370 |
}
|
371 |
|
372 |
|