Update

Author: Elfsong

bigcode_eval/generation.py

@@ -38,6 +38,7 @@ def __call__(self, input_ids, scores, **kwargs):
         """Returns true if generated sequence is too long."""
         return input_ids.shape[1] > int(self.input_length * self.multiplier)
 
+
 def parallel_generations(
         task,
         dataset,

bigcode_eval/tasks/__init__.py

@@ -4,7 +4,8 @@
 from . import (apps, codexglue_code_to_text, codexglue_text_to_text, conala,
                concode, ds1000, gsm, humaneval, humanevalplus, humanevalpack,
                instruct_humaneval, instruct_wizard_humaneval, mbpp, mbppplus,
-               multiple, parity, python_bugs, quixbugs, recode, santacoder_fim, mercury)
+               multiple, parity, python_bugs, quixbugs, recode, santacoder_fim, 
+               mercury)
 
 TASK_REGISTRY = {
     **apps.create_all_tasks(),

bigcode_eval/utils.py

@@ -120,7 +120,7 @@ def __iter__(self):
                 "n_copies (n_samples/batch_size) was changed from 1 to 2 because n_tasks isn't proportional to num devices"
             )
 
-        for sample in tqdm(range(self.n_tasks), desc="Task Encoding"):
+        for sample in range(self.n_tasks):
             for _ in range(self.n_copies):
                 if self.has_encoder:
                     yield {
@@ -220,6 +220,7 @@ def _parse_instruction(code, instruction_tokens):
         shift = len("```python")
     return code[idx + shift :]
 
+
 def complete_code(
     task,
     accelerator,
@@ -248,13 +249,11 @@ def complete_code(
     code_gens: List[List[Optional[str]]] = [[] for _ in range(n_tasks)]
     generations = [] if not intermediate_generations else intermediate_generations
     gen_token_dict = defaultdict(list)  # dict of list of generated tokens
-    
     for step, batch in tqdm(
         enumerate(dataloader),
         total=math.ceil(
             n_tasks * dataloader.dataset.n_copies / accelerator.num_processes
         ),
-        desc="batch generation",
     ):
         with torch.no_grad():
             if task.stop_words: