Fix typos

paddlemix/datacopilot/readme.md

@@ -26,7 +26,7 @@ from paddlemix.datacopilot.core import MMDataset, SCHEMA
 import paddlemix.datacopilot.ops as ops
 ```
 
-## 四、核心概念 
+## 四、核心概念
 工具核心概念包括Schema和Dataset。Schema用于定义多模态数据组织结构和字段名字。MMDataset作为数据操作的核心类，为存储，查看，转换，生成等操作的基本对象。
 
 ### SCHEMA
@@ -40,7 +40,7 @@ class SCHEMA(Enum):
 
 ### DATASET
 
-核心类MMDeteset，为存储，查看，转换，生成等操作的基本对象。支持基本的运算（切片，加法，遍历等操作）。支持json数据源。内置map，filter函数，用于高效处理数据，支持多进程和多线程并发功能。支持链式调用，方便组合多种原子操作以实现复杂的功能。通过以map函数为接口实现对数据集每个元素的处理，通过register注册机制可灵活新增作用于整个数据集的通用操作功能。
+核心类MMDataset，为存储，查看，转换，生成等操作的基本对象。支持基本的运算（切片，加法，遍历等操作）。支持json数据源。内置map，filter函数，用于高效处理数据，支持多进程和多线程并发功能。支持链式调用，方便组合多种原子操作以实现复杂的功能。通过以map函数为接口实现对数据集每个元素的处理，通过register注册机制可灵活新增作用于整个数据集的通用操作功能。
 ```
 'from_auto',
 'from_json',
@@ -184,7 +184,7 @@ sampled_dataset = dataset.sample(10)  # 随机抽取10个样本
 ```
 
 ## 七、使用案例
-1. 导入导出  
+1. 导入导出
 ```
 import functools
 from paddlemix.datacopilot.core import MMDataset, SCHEMA
@@ -196,9 +196,9 @@ print(len(dataset))
 dataset.export_json('/path/to/your/output/file.json')
 ```
 
-2. 字段处理  
+2. 字段处理
 ```
-# custom code 
+# custom code
 def update_url(item: T) -> T: ...
 
 def augment_prompt(item: T) -> T: ...
@@ -216,9 +216,9 @@ dataset = dataset.map(update_url).map(augment_prompt)
 dataset = dataset.filter(is_wanted).nonempty()
 ```
 
-3. LLaVA-SFT训练  
+3. LLaVA-SFT训练
 数据准备和训练流程参考项目[pp_cap_instruct](https://aistudio.baidu.com/projectdetail/7917712)
 
 ## 八、总结
 **DataCopilot** 是 **PaddleMIX** 提供的一个强大且灵活的多模态数据处理工具箱。
-通过掌握其基本操作和高级功能，你可以高效地处理、增强和转换多模态数据，为后续的模型训练和推理提供有力支持。
+通过掌握其基本操作和高级功能，你可以高效地处理、增强和转换多模态数据，为后续的模型训练和推理提供有力支持。

paddlemix/datasets/dataset.py

@@ -240,14 +240,14 @@ class MapDataset(Dataset):
 
     def __init__(self, data, **kwargs):
         self.data = data
-        self._transform_pipline = []
+        self._transform_pipeline = []
         self.new_data = self.data
         self.info = kwargs
         self.label_list = self.info.pop("label_list", None)
         self.vocab_info = self.info.pop("vocab_info", None)
 
     def _transform(self, data):
-        for fn in self._transform_pipline:
+        for fn in self._transform_pipeline:
             data = fn(data)
         return data
 
@@ -256,7 +256,7 @@ def __getitem__(self, idx):
         Basic function of `MapDataset` to get sample from dataset with a given
         index.
         """
-        return self._transform(self.new_data[idx]) if self._transform_pipline else self.new_data[idx]
+        return self._transform(self.new_data[idx]) if self._transform_pipeline else self.new_data[idx]
 
     def __len__(self):
         """
@@ -380,7 +380,7 @@ def _map(self, fn, lazy=True, batched=False):
         if batched:
             self.new_data = fn(self.new_data)
         elif lazy:
-            self._transform_pipline.append(fn)
+            self._transform_pipeline.append(fn)
         else:
             self.new_data = [fn(self.new_data[idx]) for idx in range(len(self.new_data))]
         return self
@@ -403,22 +403,22 @@ class IterDataset(IterableDataset):
 
     def __init__(self, data, **kwargs):
         self.data = data
-        self._transform_pipline = []
-        self._filter_pipline = []
+        self._transform_pipeline = []
+        self._filter_pipeline = []
 
         self.label_list = kwargs.pop("label_list", None)
         self.vocab_info = kwargs.pop("vocab_info", None)
 
     def _transform(self, data):
-        for fn in self._transform_pipline:
+        for fn in self._transform_pipeline:
             data = fn(data)
         return data
 
     def _shard_filter(self, num_samples):
         return True
 
     def _filter(self, data):
-        for fn in self._filter_pipline:
+        for fn in self._filter_pipeline:
             if not fn(data):
                 return False
         return True
@@ -430,19 +430,19 @@ def __iter__(self):
         num_samples = 0
         if inspect.isfunction(self.data):
             for example in self.data():
-                if (not self._filter_pipline or self._filter(self._filter_pipline)) and self._shard_filter(
+                if (not self._filter_pipeline or self._filter(self._filter_pipeline)) and self._shard_filter(
                     num_samples=num_samples
                 ):
-                    yield self._transform(example) if self._transform_pipline else example
+                    yield self._transform(example) if self._transform_pipeline else example
                 num_samples += 1
         else:
             if inspect.isgenerator(self.data):
-                warnings.warn("Reciving generator as data source, data can only be iterated once")
+                warnings.warn("Receiving generator as data source, data can only be iterated once")
             for example in self.data:
-                if (not self._filter_pipline or self._filter(self._filter_pipline)) and self._shard_filter(
+                if (not self._filter_pipeline or self._filter(self._filter_pipeline)) and self._shard_filter(
                     num_samples=num_samples
                 ):
-                    yield self._transform(example) if self._transform_pipline else example
+                    yield self._transform(example) if self._transform_pipeline else example
                 num_samples += 1
 
     def filter(self, fn):
@@ -455,7 +455,7 @@ def filter(self, fn):
                 returns a boolean. Samples that return False are discarded.
         """
 
-        self._filter_pipline.append(fn)
+        self._filter_pipeline.append(fn)
 
         return self
 
@@ -495,7 +495,7 @@ def map(self, fn):
                 sample as argument.
         """
 
-        self._transform_pipline.append(fn)
+        self._transform_pipeline.append(fn)
 
         return self
 

paddlemix/examples/qwen2_vl/Qwen2_VL_run.ipynb

@@ -316,10 +316,10 @@
       "outputs": [],
       "source": [
         "# 2B\n",
-        "sh paddlemix/examples/qwen2_vl/shell/basline_2b_bs32_1e8.sh\n",
+        "sh paddlemix/examples/qwen2_vl/shell/baseline_2b_bs32_1e8.sh\n",
         "\n",
         "# 7B\n",
-        "sh paddlemix/examples/qwen2_vl/shell/basline_7b_bs32_1e8.sh"
+        "sh paddlemix/examples/qwen2_vl/shell/baseline_7b_bs32_1e8.sh"
       ]
     },
     {

paddlemix/examples/qwen2_vl/README.md

@@ -85,16 +85,16 @@ wget https://paddlenlp.bj.bcebos.com/datasets/paddlemix/playground/opensource_js
 
 ```bash
 # 2B
-sh paddlemix/examples/qwen2_vl/shell/basline_2b_bs32_1e8.sh
+sh paddlemix/examples/qwen2_vl/shell/baseline_2b_bs32_1e8.sh
 
 # 2B lora
-sh paddlemix/examples/qwen2_vl/shell/basline_2b_lora_bs32_1e8.sh
+sh paddlemix/examples/qwen2_vl/shell/baseline_2b_lora_bs32_1e8.sh
 
 # 7B
-sh paddlemix/examples/qwen2_vl/shell/basline_7b_bs32_1e8.sh
+sh paddlemix/examples/qwen2_vl/shell/baseline_7b_bs32_1e8.sh
 
 # 7B lora
-sh paddlemix/examples/qwen2_vl/shell/basline_7b_lora_bs32_1e8.sh
+sh paddlemix/examples/qwen2_vl/shell/baseline_7b_lora_bs32_1e8.sh
 ```
 
 注意：微调2b模型的运行示例如下：

paddlemix/examples/qwen2_vl/shell/basline_2b_bs32_1e8.sh → paddlemix/examples/qwen2_vl/shell/baseline_2b_bs32_1e8.sh

@@ -26,7 +26,7 @@ export PYTHONPATH="${PYTHONPATH}:$(pwd)"
 export MASTER_PORT=34229
 export TF_CPP_MIN_LOG_LEVEL=3
 
-OUTPUT_DIR='work_dirs/basline_330k_2b_bs32_1e8'
+OUTPUT_DIR='work_dirs/baseline_330k_2b_bs32_1e8'
 
 if [ ! -d "$OUTPUT_DIR" ]; then
   mkdir -p "$OUTPUT_DIR"

paddlemix/examples/qwen2_vl/shell/basline_2b_lora_bs32_1e8.sh → paddlemix/examples/qwen2_vl/shell/baseline_2b_lora_bs32_1e8.sh

@@ -26,7 +26,7 @@ export PYTHONPATH="${PYTHONPATH}:$(pwd)"
 export MASTER_PORT=34229
 export TF_CPP_MIN_LOG_LEVEL=3
 
-OUTPUT_DIR='work_dirs/basline_330k_2b_bs32_1e8'
+OUTPUT_DIR='work_dirs/baseline_330k_2b_bs32_1e8'
 
 if [ ! -d "$OUTPUT_DIR" ]; then
   mkdir -p "$OUTPUT_DIR"

paddlemix/examples/qwen2_vl/shell/basline_7b_bs32_1e8.sh → paddlemix/examples/qwen2_vl/shell/baseline_7b_bs32_1e8.sh

@@ -26,7 +26,7 @@ export PYTHONPATH="${PYTHONPATH}:$(pwd)"
 export MASTER_PORT=34229
 export TF_CPP_MIN_LOG_LEVEL=3
 
-OUTPUT_DIR='work_dirs/basline_330k_7b_bs32_1e8'
+OUTPUT_DIR='work_dirs/baseline_330k_7b_bs32_1e8'
 
 if [ ! -d "$OUTPUT_DIR" ]; then
   mkdir -p "$OUTPUT_DIR"

paddlemix/examples/qwen2_vl/shell/basline_7b_lora_bs32_1e8.sh → paddlemix/examples/qwen2_vl/shell/baseline_7b_lora_bs32_1e8.sh

@@ -26,7 +26,7 @@ export PYTHONPATH="${PYTHONPATH}:$(pwd)"
 export MASTER_PORT=34229
 export TF_CPP_MIN_LOG_LEVEL=3
 
-OUTPUT_DIR='work_dirs/basline_330k_7b_bs32_1e8'
+OUTPUT_DIR='work_dirs/baseline_330k_7b_bs32_1e8'
 
 if [ ! -d "$OUTPUT_DIR" ]; then
   mkdir -p "$OUTPUT_DIR"