Fix taskset and eval_tasksets path (#367)

hiyuchang · web-flow · commit c670d0a497e1 · 2025-11-05T14:52:15.000+08:00
diff --git a/docs/sphinx_doc/source/tutorial/example_async_mode.md b/docs/sphinx_doc/source/tutorial/example_async_mode.md
@@ -31,7 +31,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: train
       format:
@@ -79,7 +79,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       format:
         prompt_key: 'question'
@@ -143,7 +143,7 @@ buffer:
     taskset:  # important
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       format:
         prompt_key: 'question'
diff --git a/docs/sphinx_doc/source/tutorial/example_reasoning_basic.md b/docs/sphinx_doc/source/tutorial/example_reasoning_basic.md
@@ -69,7 +69,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -81,7 +81,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/docs/sphinx_doc/source/tutorial/trinity_configs.md b/docs/sphinx_doc/source/tutorial/trinity_configs.md
@@ -275,7 +275,7 @@ The configuration for each task dataset is defined as follows:
   - `file`: The dataset is stored in `jsonl`/`parquet` files. The data file organization is required to meet the huggingface standard. *We recommand using this storage type for most cases.*
   - `sql`: The dataset is stored in a SQL database. *This type is unstable and will be optimized in the future versions.*
 - `path`: The path to the task dataset.
-  - For `file` storage type, the path points to the directory that contains the task dataset files.
+  - For `file` storage type, the path points to the directory that contains the task dataset files. It supports loading both local and remote data files in a compatible format with [`datasets.load_dataset()`](https://huggingface.co/docs/datasets/main/en/package_reference/loading_methods#datasets.load_dataset) function.
   - For `sql` storage type, the path points to the sqlite database file.
 - `subset_name`: The subset name of the task dataset, corresponding to the `name` parameter in huggingface datasets `load_dataset` function. Default is `None`.
 - `split`: The split of the task dataset, corresponding to the `split` parameter in huggingface datasets `load_dataset` function. Default is `train`.
diff --git a/docs/sphinx_doc/source_zh/tutorial/example_async_mode.md b/docs/sphinx_doc/source_zh/tutorial/example_async_mode.md
@@ -31,7 +31,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: train
       format:
@@ -79,7 +79,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       format:
         prompt_key: 'question'
@@ -143,7 +143,7 @@ buffer:
     taskset:  # important
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       format:
         prompt_key: 'question'
diff --git a/docs/sphinx_doc/source_zh/tutorial/example_reasoning_basic.md b/docs/sphinx_doc/source_zh/tutorial/example_reasoning_basic.md
@@ -69,7 +69,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -81,7 +81,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/docs/sphinx_doc/source_zh/tutorial/trinity_configs.md b/docs/sphinx_doc/source_zh/tutorial/trinity_configs.md
@@ -272,7 +272,7 @@ buffer:
   - `file`: 数据集存储在 `jsonl`/`parquet` 文件中。数据文件组织需符合 HuggingFace 标准。*建议大多数情况下使用此存储类型。*
   - `sql`: 数据集存储在 SQL 数据库中。*此类型尚不稳定，将在未来版本中优化。*
 - `path`: 任务数据集的路径。
-  - 对于 `file` 类型，路径指向包含任务数据集文件的目录。
+  - 对于 `file` 类型，路径指向包含任务数据集文件的目录。它支持使用与 [`datasets.load_dataset()`](https://huggingface.co/docs/datasets/main/en/package_reference/loading_methods#datasets.load_dataset) 函数兼容的格式加载本地和远程数据文件。
   - 对于 `sql` 类型，路径指向 sqlite 数据库文件。
 - `subset_name`: 任务数据集的子集名称，对应 huggingface datasets `load_dataset` 函数中的 `name` 参数。默认为 `None`。
 - `split`: 任务数据集的划分。对应 huggingface datasets `load_dataset` 函数中的 `split` 参数。默认为 `train`。
diff --git a/examples/agentscope_react/gsm8k.yaml b/examples/agentscope_react/gsm8k.yaml
@@ -21,7 +21,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
diff --git a/examples/agentscope_tool_react/agentscopev0_tool_react_gsm8k.yaml b/examples/agentscope_tool_react/agentscopev0_tool_react_gsm8k.yaml
@@ -21,7 +21,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
diff --git a/examples/agentscope_tool_react/agentscopev1_tool_react_dapo.yaml b/examples/agentscope_tool_react/agentscopev1_tool_react_dapo.yaml
@@ -21,7 +21,7 @@ buffer:
     taskset:
       name: dapo
       storage_type: file
-      path: open-r1/DAPO-Math-17k-Processed
+      path: ${oc.env:TRINITY_TASKSET_PATH,open-r1/DAPO-Math-17k-Processed}
       subset_name: en
       split: train
       format:
diff --git a/examples/asymre_gsm8k/gsm8k.yaml b/examples/asymre_gsm8k/gsm8k.yaml
@@ -28,7 +28,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: train
       format:
@@ -40,7 +40,7 @@ buffer:
     eval_tasksets:
       - name: gsm8k-eval
         storage_type: file
-        path: 'openai/gsm8k'
+        path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
         subset_name: 'main'
         split: test
         format:
diff --git a/examples/async_gsm8k/explorer.yaml b/examples/async_gsm8k/explorer.yaml
@@ -19,7 +19,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: train
       format:
diff --git a/examples/async_gsm8k/trainer.yaml b/examples/async_gsm8k/trainer.yaml
@@ -21,7 +21,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
diff --git a/examples/cispo_gsm8k/gsm8k.yaml b/examples/cispo_gsm8k/gsm8k.yaml
@@ -20,7 +20,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -31,7 +31,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/examples/dapo_math/dapo.yaml b/examples/dapo_math/dapo.yaml
@@ -43,8 +43,8 @@ buffer:
     eval_tasksets:
     - name: AIME2024
       storage_type: file
-      path: ${oc.env:TRINITY_TASKSET_PATH}  # e.g. path to AIME2024
-      split: 'test'
+      path: HuggingFaceH4/aime_2024  # Path to the AIME2024 dataset
+      split: 'train'
       repeat_times: 32
       format:
         prompt_key: 'question'
diff --git a/examples/dpo_humanlike/dpo.yaml b/examples/dpo_humanlike/dpo.yaml
@@ -28,7 +28,7 @@ buffer:
       name: dpo_buffer
       storage_type: file
       enable_progress_bar: True
-      path: ${oc.env:TRINITY_TASKSET_PATH}
+      path: ${oc.env:TRINITY_TASKSET_PATH} # Path to the DPO dataset
       format:
         prompt_type: plaintext # plaintext/messages
         prompt_key: prompt
diff --git a/examples/grpo_gsm8k/gsm8k.yaml b/examples/grpo_gsm8k/gsm8k.yaml
@@ -20,7 +20,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -31,7 +31,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/examples/grpo_gsm8k_experience_pipeline/gsm8k.yaml b/examples/grpo_gsm8k_experience_pipeline/gsm8k.yaml
@@ -40,7 +40,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -51,7 +51,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/examples/grpo_gsm8k_ruler/gsm8k_ruler.yaml b/examples/grpo_gsm8k_ruler/gsm8k_ruler.yaml
@@ -22,7 +22,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -33,7 +33,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/examples/grpo_gsm8k_task_pipeline/gsm8k.yaml b/examples/grpo_gsm8k_task_pipeline/gsm8k.yaml
@@ -49,7 +49,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/examples/grpo_gsm8k_trainable_ruler/gsm8k_ruler.yaml b/examples/grpo_gsm8k_trainable_ruler/gsm8k_ruler.yaml
@@ -24,7 +24,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -35,7 +35,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/examples/grpo_lora_gsm8k/gsm8k.yaml b/examples/grpo_lora_gsm8k/gsm8k.yaml
@@ -24,7 +24,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -35,7 +35,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/examples/opmd_gsm8k/opmd_gsm8k.yaml b/examples/opmd_gsm8k/opmd_gsm8k.yaml
@@ -21,7 +21,9 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: '{path to datasets}/gsm8k/main' # the path where your jsonl or parquet files are located; use {path to datasets}/gsm8k/socratic for Socratic dataset
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
+      subset_name: main
+      split: train
       format:
         prompt_key: 'question'
         response_key: 'answer'
diff --git a/examples/rec_gsm8k/gsm8k.yaml b/examples/rec_gsm8k/gsm8k.yaml
@@ -31,7 +31,8 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: ${oc.env:TRINITY_TASKSET_PATH}
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
+      subset_name: main
       split: train
       format:
         prompt_key: question
@@ -41,7 +42,8 @@ buffer:
     eval_tasksets:
       - name: gsm8k-eval
         storage_type: file
-        path: ${oc.env:TRINITY_EVAL_TASKSET_PATH}
+        path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
+        subset_name: main
         split: test
         format:
           prompt_key: question
diff --git a/examples/sppo_gsm8k/gsm8k.yaml b/examples/sppo_gsm8k/gsm8k.yaml
@@ -26,7 +26,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: train
       format:
@@ -37,7 +37,7 @@ buffer:
     eval_tasksets:
       - name: gsm8k-eval
         storage_type: file
-        path: 'openai/gsm8k'
+        path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
         subset_name: 'main'
         split: test
         format:
diff --git a/examples/topr_gsm8k/gsm8k.yaml b/examples/topr_gsm8k/gsm8k.yaml
@@ -20,7 +20,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'train'
       format:
@@ -31,7 +31,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: 'openai/gsm8k'
+      path: ${oc.env:TRINITY_TASKSET_PATH,openai/gsm8k}
       subset_name: 'main'
       split: 'test'
       format:
