Upload run_program.py with huggingface_hub

run_program.py

@@ -238,9 +238,11 @@ def update_question_with_new_parameters():
 	json.dump(program_data, outfile, indent=4)
 
 
-def call_answer_question(question, model_name='gpt'):
-	prompt_template = PROMPT_DICT['prompt_answer_question']
-	# prompt_template = PROMPT_DICT['prompt_answer_question_few_shot_cot']
+def call_answer_question(question, model_name='gpt', cot=False):
+	if cot:
+		prompt_template = PROMPT_DICT['prompt_answer_question_few_shot_cot']
+	else:
+		prompt_template = PROMPT_DICT['prompt_answer_question']
 	prompt = prompt_template.format_map(
 		{"question": question}
 	)
@@ -321,7 +323,8 @@ def call_answer_question(question, model_name='gpt'):
 			outputs = llama_pipeline(
 				messages,
 				max_new_tokens=300,
-				temperature=0.00001
+				# temperature=0.00001
+				temperature = 0.7
 			)
 			# print(outputs[0]["generated_text"][-1])
 			return outputs[0]["generated_text"][-1]['content']
@@ -332,19 +335,19 @@ def answer_question(model_name='gpt'):
 	program_data = json.load(infile)
 	print(len(program_data))
 	for case in tqdm(program_data):
-		response = call_answer_question(case['question'], model_name=model_name)
+		response = call_answer_question(case['question'], model_name=model_name, cot=True)
 		case['prediction'] = response
 		# print(case['prediction'])
 		case['new_prediction'] = []
 		for question in case['new_questions']:
-			response = call_answer_question(question, model_name=model_name)
+			response = call_answer_question(question, model_name=model_name, cot=True)
 			case['new_prediction'].append(response)
 		# print(case)
 		# break
 	# print(case)
 	# break
-	outfile = open('data/math/test_dump_gsm8k_train_perturbed_with_new_questions_answer_llama8b.json', 'w')
-	# outfile = open('data/math/gsm8k_cot_sc_qwen/temp=0.7_iter=4.json', 'w')
+	# outfile = open('data/math/test_dump_gsm8k_train_perturbed_with_new_questions_answer_llama8b.json', 'w')
+	outfile = open('data/math/gsm8k_cot_sc_llama3.1_8b/temp=0.7_iter=5.json', 'w')
 	json.dump(program_data, outfile, indent=4)