github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-08 12:30:42 +00:00
Code Issues Packages Projects Releases Wiki Activity

[nfc] fix typo colossalai/ applications/ (#3831)

Browse Source 
* fix typo colossalai/autochunk auto_parallel amp

* fix typo colossalai/auto_parallel nn utils etc.

* fix typo colossalai/auto_parallel autochunk fx/passes  etc.

* fix typo docs/

* change placememt_policy to placement_policy in docs/ and examples/

* fix typo colossalai/ applications/
This commit is contained in:
digger yu
2023-05-25 16:19:41 +08:00
committed by GitHub
parent a64df3fa97
commit e2d81eba0d
7 changed files with 15 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
applications/Chat/coati/ray/src/detached_replay_buffer.py
View File

@@ -34,7 +34,7 @@ class DetachedReplayBuffer:
'''
Workers in the same tp group share this buffer and need same sample for one step.
Therefore a held_sample should be returned tp_world_size times before it could be dropped.
worker_state records wheter a worker got the held_sample
worker_state records whether a worker got the held_sample
'''
self.tp_world_size = tp_world_size
self.worker_state = [False] * self.tp_world_size

2
applications/Chat/coati/ray/src/experience_maker_holder.py
View File

@@ -22,7 +22,7 @@ from .utils import is_rank_0, get_strategy_from_args, set_dist_env
class ExperienceMakerHolder:
'''
Args:
detached_trainer_name_list: str list to get ray actor handleskkk
detached_trainer_name_list: str list to get ray actor handles
strategy:
experience_batch_size: batch size of generated experience
kl_coef: the coefficient of kl divergence loss

2
applications/Chat/coati/ray/src/pipeline_strategy.py
View File

@@ -26,7 +26,7 @@ rpc_is_initialized = _is_current_rpc_agent_set
class PipelineModel(torch.nn.Module):
'''
Actor has 2 kinds of jobs: forward and generate.
better to just pipelinize the inner model
better to just pipeline the inner model
'''
def __init__(self,
model: torch.nn.Module,

2
applications/Chat/evaluate/evaluator.py
View File

@@ -119,7 +119,7 @@ class Evaluator(object):
jdump(all_evaluations,
os.path.join(evaluation_results_save_path, f"{model_name_list[0]}_evaluation_results.json"))
# Start to calculate scores and save statictics.
# Start to calculate scores and save statistics.
evaluation_statistics_save_path = os.path.join(base_save_path, "evaluation_statistics")
gpt_evaluate.save_gpt35_evaluation_statistics(model_name_list[0], all_evaluations,
evaluation_statistics_save_path)

8
applications/Chat/evaluate/metrics.py
View File

@@ -111,7 +111,7 @@ def calculate_precision_recall_f1(preds: list, targets: list) -> dict:
The calculation of precision, recall and f1-score is realized by counting
the number f overlaps between the preds and target. The comparison length
limited by the shorter one of preds and targets. This design is mainly
considered for classifiction and extraction categories.
considered for classification and extraction categories.
"""
precision_recall_f1 = {"precision": 0, "recall": 0, "f1_score": 0}
precision_scores = []
@@ -138,7 +138,7 @@ def calculate_precision_recall_f1(preds: list, targets: list) -> dict:
def precision(preds: list, targets: list) -> dict:
"""Calculate Precision Metric
(design for classifiction and extraction categories)
(design for classification and extraction categories)
Calculating precision by counting the number of overlaps between the preds and target.
"""
@@ -149,7 +149,7 @@ def precision(preds: list, targets: list) -> dict:
def recall(preds: list, targets: list) -> dict:
"""Calculate Recall Metric
(design for classifiction and extraction categories)
(design for classification and extraction categories)
Calculating recall by counting the number of overlaps between the preds and target.
"""
@@ -160,7 +160,7 @@ def recall(preds: list, targets: list) -> dict:
def F1_score(preds: list, targets: list) -> dict:
"""Calculate F1-score Metric
(design for classifiction and extraction categories)
(design for classification and extraction categories)
Calculating f1-score by counting the number of overlaps between the preds and target.
"""