commented out

run_llm.py

@@ -70,24 +70,24 @@ model_mapping = {
     # 'koala-13b': 'koala-13b',
 }
 
-# for m in model_mapping.keys():
-#     for eid, ent in enumerate(ents):
-#         os.makedirs(f'result/prompt1_qa/{m}/ptb/per_ent/{ent}', exist_ok=True)
+for m in model_mapping.keys():
+    for eid, ent in enumerate(ents):
+        os.makedirs(f'result/prompt1_qa/{m}/ptb/per_ent/{ent}', exist_ok=True)
         
-#         os.makedirs(f'result/prompt2_instruction/pos_tagging/{m}/ptb', exist_ok=True)
-#         os.makedirs(f'result/prompt2_instruction/chunking/{m}/ptb', exist_ok=True)
-#         os.makedirs(f'result/prompt2_instruction/parsing/{m}/ptb', exist_ok=True)
+        os.makedirs(f'result/prompt2_instruction/pos_tagging/{m}/ptb', exist_ok=True)
+        os.makedirs(f'result/prompt2_instruction/chunking/{m}/ptb', exist_ok=True)
+        os.makedirs(f'result/prompt2_instruction/parsing/{m}/ptb', exist_ok=True)
 
-#         os.makedirs(f'result/prompt3_structured_prompt/pos_tagging/{m}/ptb', exist_ok=True)
-#         os.makedirs(f'result/prompt3_structured_prompt/chunking/{m}/ptb', exist_ok=True)
-#         os.makedirs(f'result/prompt3_structured_prompt/parsing/{m}/ptb', exist_ok=True)
+        os.makedirs(f'result/prompt3_structured_prompt/pos_tagging/{m}/ptb', exist_ok=True)
+        os.makedirs(f'result/prompt3_structured_prompt/chunking/{m}/ptb', exist_ok=True)
+        os.makedirs(f'result/prompt3_structured_prompt/parsing/{m}/ptb', exist_ok=True)
 
 
-# s = int(sys.argv[1])
-# e = int(sys.argv[2])
+#s = int(sys.argv[1])
+#e = int(sys.argv[2])
 
-# s = 0
-# e = 1000
+#s = 0
+#e = 1000
 with open('sample_uniform_1k_2.txt', 'r') as f:
     selected_idx = f.readlines()
 selected_idx = [int(i.strip()) for i in selected_idx]#[s:e]
@@ -219,34 +219,34 @@ def main(args=None):
             text = ptb[gid]['text']
 
             ## POS tagging
-            # if os.path.exists(f'result/prompt2_instruction/pos_tagging/{args.model_path}/ptb/{gid}.txt'):
-            #     print(gid, 'skip')
+            if os.path.exists(f'result/prompt2_instruction/pos_tagging/{args.model_path}/ptb/{gid}.txt'):
+                print(gid, 'skip')
 
-            # else:
-            #     msg = prompt2_pos.format(text)
+            else:
+                msg = prompt2_pos.format(text)
 
-            #     if 'gpt3' in args.model_path:
-            #         outputs = gpt3(msg)
-            #         if outputs is None:
-            #             continue
-            #         time.sleep(0.2)
+                if 'gpt3' in args.model_path:
+                    outputs = gpt3(msg)
+                    if outputs is None:
+                        continue
+                    time.sleep(0.2)
 
-            #     else:
-            #         conv = get_conversation_template(args.model_path)
-            #         conv.append_message(conv.roles[0], msg)
-            #         conv.append_message(conv.roles[1], None)
-            #         conv.system = ''
-            #         prompt = conv.get_prompt()
+                else:
+                    conv = get_conversation_template(args.model_path)
+                    conv.append_message(conv.roles[0], msg)
+                    conv.append_message(conv.roles[1], None)
+                    conv.system = ''
+                    prompt = conv.get_prompt()
 
-            #         outputs = fastchat(prompt, model, tokenizer)
+                    outputs = fastchat(prompt, model, tokenizer)
 
-            #     with open(f'result/prompt2_instruction/pos_tagging/{args.model_path}/ptb/{gid}.txt', 'w') as f:
-            #         f.write(outputs)
+                with open(f'result/prompt2_instruction/pos_tagging/{args.model_path}/ptb/{gid}.txt', 'w') as f:
+                    f.write(outputs)
 
 
             ## Sentence chunking
-            # if os.path.exists(f'result/prompt2_instruction/chunking/{args.model_path}/ptb/{gid}.txt'):
-            #     print(gid, 'skip')
+            if os.path.exists(f'result/prompt2_instruction/chunking/{args.model_path}/ptb/{gid}.txt'):
+                print(gid, 'skip')
             if False:
                 pass
             else:
@@ -273,29 +273,29 @@ def main(args=None):
 
             
             ## Parsing
-            # if os.path.exists(f'result/prompt2_instruction/parsing/{args.model_path}/ptb/{gid}.txt'):
-            #     print(gid, 'skip')
+            if os.path.exists(f'result/prompt2_instruction/parsing/{args.model_path}/ptb/{gid}.txt'):
+                print(gid, 'skip')
             
-            # else:
-            #     msg = prompt2_parse.format(text)
+            else:
+                msg = prompt2_parse.format(text)
 
-            #     if 'gpt3' in args.model_path:
-            #         outputs = gpt3(msg)
-            #         if outputs is None:
-            #             continue
-            #         time.sleep(0.2)
+                if 'gpt3' in args.model_path:
+                    outputs = gpt3(msg)
+                    if outputs is None:
+                        continue
+                    time.sleep(0.2)
 
-            #     else:
-            #         conv = get_conversation_template(args.model_path)
-            #         conv.append_message(conv.roles[0], msg)
-            #         conv.append_message(conv.roles[1], None)
-            #         conv.system = ''
-            #         prompt = conv.get_prompt()
+                else:
+                    conv = get_conversation_template(args.model_path)
+                    conv.append_message(conv.roles[0], msg)
+                    conv.append_message(conv.roles[1], None)
+                    conv.system = ''
+                    prompt = conv.get_prompt()
 
-            #         outputs = fastchat(prompt, model, tokenizer)
+                    outputs = fastchat(prompt, model, tokenizer)
 
-            #     with open(f'result/prompt2_instruction/parsing/{args.model_path}/ptb/{gid}.txt', 'w') as f:
-            #         f.write(outputs)
+                with open(f'result/prompt2_instruction/parsing/{args.model_path}/ptb/{gid}.txt', 'w') as f:
+                    f.write(outputs)
 
 
 
@@ -306,29 +306,29 @@ def main(args=None):
             poss = ptb[gid]['uni_poss']
 
             ## POS tagging
-            # if os.path.exists(f'result/prompt3_structured_prompt/pos_tagging/{args.model_path}/ptb/{gid}.txt'):
-            #     print(gid, 'skip')
-            #     continue
+            if os.path.exists(f'result/prompt3_structured_prompt/pos_tagging/{args.model_path}/ptb/{gid}.txt'):
+                print(gid, 'skip')
+                continue
 
-            # prompt = demon_pos + '\n' + 'C: ' + text + '\n' + 'T: '
+            prompt = demon_pos + '\n' + 'C: ' + text + '\n' + 'T: '
 
-            # if 'gpt3' in args.model_path:
-            #     outputs = gpt3(prompt)
-            #     if outputs is None:
-            #         continue
-            #     time.sleep(0.2)
+            if 'gpt3' in args.model_path:
+                outputs = gpt3(prompt)
+                if outputs is None:
+                    continue
+                time.sleep(0.2)
 
-            # else:
-            #     pred_poss = []
-            #     for _tok, _pos in zip(tokens, poss):
-            #         prompt = prompt + ' ' + _tok + '_'
-            #         outputs = structured_prompt(prompt, model, tokenizer, bad_words_ids_pos)
-            #         prompt = prompt + outputs
-            #         pred_poss.append(outputs)
+            else:
+                pred_poss = []
+                for _tok, _pos in zip(tokens, poss):
+                    prompt = prompt + ' ' + _tok + '_'
+                    outputs = structured_prompt(prompt, model, tokenizer, bad_words_ids_pos)
+                    prompt = prompt + outputs
+                    pred_poss.append(outputs)
 
-            # outputs = ' '.join(pred_poss)
-            # with open(f'result/prompt3_structured_prompt/pos_tagging/{args.model_path}/ptb/{gid}.txt', 'w') as f:
-            #     f.write(outputs)
+            outputs = ' '.join(pred_poss)
+            with open(f'result/prompt3_structured_prompt/pos_tagging/{args.model_path}/ptb/{gid}.txt', 'w') as f:
+                f.write(outputs)
 
 
             ## Chunking
@@ -366,27 +366,27 @@ def main(args=None):
                 f.write(outputs)
 
             ## Parsing
-            # if os.path.exists(f'result/prompt3_structured_prompt/parsing/{args.model_path}/ptb/{gid}.txt'):
-            #     print(gid, 'skip')
-            #     continue
+            if os.path.exists(f'result/prompt3_structured_prompt/parsing/{args.model_path}/ptb/{gid}.txt'):
+                print(gid, 'skip')
+                continue
 
-            # prompt = demon_parse + '\n' + 'C: ' + text + '\n' + 'T: '
+            prompt = demon_parse + '\n' + 'C: ' + text + '\n' + 'T: '
 
-            # if 'gpt3' in args.model_path:
-            #     outputs = gpt3(prompt)
-            #     if outputs is None:
-            #         continue
-            #     time.sleep(0.2)
+            if 'gpt3' in args.model_path:
+                outputs = gpt3(prompt)
+                if outputs is None:
+                    continue
+                time.sleep(0.2)
 
-            # else:
-            #     pred_syn = []
-            #     for _tok, _pos in zip(tokens, poss):
-            #         prompt = prompt + _tok + '_'
-            #         outputs = structured_prompt(prompt, model, tokenizer, bad_words_ids_parse)
-            #         pred_syn.append(outputs)
+            else:
+                pred_syn = []
+                for _tok, _pos in zip(tokens, poss):
+                    prompt = prompt + _tok + '_'
+                    outputs = structured_prompt(prompt, model, tokenizer, bad_words_ids_parse)
+                    pred_syn.append(outputs)
 
-            # with open(f'result/prompt3_structured_prompt/parsing/{args.model_path}/ptb/{gid}.txt', 'w') as f:
-            #     f.write(' '.join(pred_syn))
+            with open(f'result/prompt3_structured_prompt/parsing/{args.model_path}/ptb/{gid}.txt', 'w') as f:
+                f.write(' '.join(pred_syn))
 
 
 def structured_prompt(prompt, model, tokenizer, bad_words_ids):
@@ -426,9 +426,9 @@ def fastchat(prompt, model, tokenizer):
         output_ids, skip_special_tokens=True, spaces_between_special_tokens=False
     )
 
-    # print('Empty system message')
-    # print(f"{conv.roles[0]}: {msg}")
-    # print(f"{conv.roles[1]}: {outputs}")
+    print('Empty system message')
+    print(f"{conv.roles[0]}: {msg}")
+    print(f"{conv.roles[1]}: {outputs}")
 
     return outputs