Revert "Add Programming to Foundry"

llmfoundry/models/hf/hf_causal_lm.py

@@ -8,8 +8,7 @@
 
 # required for loading a python model into composer
 import transformers
-from composer.metrics.nlp import (InContextLearningCodeEvalAccuracy,
-                                  InContextLearningLMAccuracy,
+from composer.metrics.nlp import (InContextLearningLMAccuracy,
                                   InContextLearningLMExpectedCalibrationError,
                                   InContextLearningMCExpectedCalibrationError,
                                   InContextLearningMultipleChoiceAccuracy,
@@ -75,7 +74,6 @@ def __init__(
             InContextLearningLMAccuracy(),
             InContextLearningMultipleChoiceAccuracy(),
             InContextLearningQAAccuracy(),
-            InContextLearningCodeEvalAccuracy(),
             InContextLearningLMExpectedCalibrationError(),
             InContextLearningMCExpectedCalibrationError()
         ]

llmfoundry/models/mpt/modeling_mpt.py

@@ -13,8 +13,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-from composer.metrics import (InContextLearningCodeEvalAccuracy,
-                              InContextLearningLMAccuracy,
+from composer.metrics import (InContextLearningLMAccuracy,
                               InContextLearningLMExpectedCalibrationError,
                               InContextLearningMCExpectedCalibrationError,
                               InContextLearningMultipleChoiceAccuracy,
@@ -699,7 +698,6 @@ def __init__(
             InContextLearningLMAccuracy(),
             InContextLearningMultipleChoiceAccuracy(),
             InContextLearningQAAccuracy(),
-            InContextLearningCodeEvalAccuracy(),
             InContextLearningLMExpectedCalibrationError(),
             InContextLearningMCExpectedCalibrationError(),
         ]

llmfoundry/utils/builders.py

@@ -174,8 +174,6 @@ def _validate_cfg(icl_cfg: DictConfig):
                 ]
             elif icl_cfg.icl_task_type == 'question_answering':
                 icl_cfg.metric_names = ['InContextLearningQAAccuracy']
-            elif icl_cfg.icl_task_type == 'code_evaluation':
-                icl_cfg.metric_names = ['InContextLearningCodeEvalAccuracy']
             else:
                 raise ValueError(
                     f'No metric_names defined, unable to build default metrics for icl_task_type={icl_cfg.icl_task_type}.'
@@ -191,8 +189,6 @@ def _validate_cfg(icl_cfg: DictConfig):
             icl_cfg.max_seq_len = default_max_seq_len
         if 'batch_size' not in icl_cfg:
             icl_cfg.batch_size = default_batch_size
-        if 'num_beams' not in icl_cfg:
-            icl_cfg.num_beams = 1
 
     for icl_cfg in icl_tasks_list:
         _validate_cfg(icl_cfg)
@@ -222,7 +218,6 @@ def _validate_cfg(icl_cfg: DictConfig):
                 example_delimiter=icl_cfg.example_delimiter,
                 continuation_delimiter=icl_cfg.continuation_delimiter,
                 destination_path=destination_path,
-                generations_per_sample=icl_cfg.num_beams,
                 has_categories=icl_cfg.get('has_categories', False),
             )
             if hasattr(