[Feature] Switch dataset path / model path to HF location to ease community usage (#82)

* Update .gitignore and modify dataset paths in scripts for improved file management and compatibility with Hugging Face datasets. Additionally, refactor dataset loading functions to utilize load_hf_or_local_file for better flexibility.

* Remove sglang subproject and update dataset path format in load_hf_or_local_file function for compatibility with Hugging Face datasets.

* Refactor imports in grader.py and parser.py to include sympy for improved functionality.

.gitignore

@@ -182,3 +182,5 @@ api_key.json
 
 .vscode/
 wandb/
+outputs/
+sympy/

examples/run_async_ppo.sh

@@ -2,12 +2,12 @@
 python3 training/main_async_ppo.py \
     n_nodes=1 n_gpus_per_node=8 \
     allocation_mode=sglang.d4p1m1+d2p2m1 \
-    cluster.fileroot=/storage/testing/experiments \
+    cluster.fileroot=experiments \
     actor.type._class=qwen3 \
-    actor.path=/storage/testing/models/Qwen__Qwen3-1.7B \
+    actor.path=Qwen/Qwen3-1.7B \
     ref.type._class=qwen3 \
-    ref.path=/storage/testing/models/Qwen__Qwen3-1.7B \
-    dataset.path=/storage/testing/dataset/boba_106k_0319.jsonl \
+    ref.path=Qwen/Qwen3-1.7B \
+    dataset.path=hf-dataset://inclusionAI/AReaL-RL-Data/data/boba_106k_0319.jsonl \
     dataset.train_bs_n_seqs=32 \
     group_size=8 \
     ppo.gen.max_new_tokens=4096 \
@@ -15,4 +15,4 @@ python3 training/main_async_ppo.py \
     actor_train.mb_spec.max_tokens_per_mb=32768 \
     actor_inf.mb_spec.max_tokens_per_mb=32768 \
     max_concurrent_rollouts=16 \
-    max_head_offpolicyness=4
+    max_head_offpolicyness=4

functioncall/math/function/grader.py

@@ -14,10 +14,11 @@ from typing import Union
 
 import regex
 from latex2sympy2 import latex2sympy
-from sympy import N, simplify
 from sympy.parsing.latex import parse_latex
 from sympy.parsing.sympy_parser import parse_expr
 
+from sympy import N, simplify
+
 # from .parser import choice_answer_clean, strip_string
 # from parser import choice_answer_clean
 

functioncall/math/function/parser.py

@@ -3,10 +3,11 @@ import re
 from typing import Any, Dict, Iterable, List, TypeVar, Union
 
 import regex
-import sympy
 from latex2sympy2 import latex2sympy
 from word2number import w2n
 
+import sympy
+
 # from utils import *
 
 

functioncall/test/performance_eval.py

@@ -14,6 +14,7 @@ from typing import Any, Dict, List
 import numpy as np
 
 from functioncall.code.verify import code_verify
+from realhf.utils import load_hf_or_local_file
 
 logger = logging.getLogger("function call")
 
@@ -223,6 +224,7 @@ def statics_result(result, query_ids):
 def standard_dataset_eval(
     dataset_path, code_count=0, test_case_batch_size=20, dry_run=False
 ):
+    dataset_path = load_hf_or_local_file(dataset_path)
     id2info = defaultdict(dict)
     generateds, query_ids = [], []
     cnt = 0

grader.py

@@ -15,10 +15,11 @@ from typing import Union
 
 import regex
 from latex2sympy2 import latex2sympy
-from sympy import N, simplify
 from sympy.parsing.latex import parse_latex
 from sympy.parsing.sympy_parser import parse_expr
 
+from sympy import N, simplify
+
 # from .parser import choice_answer_clean, strip_string
 # from parser import choice_answer_clean
 

parser.py

@@ -5,10 +5,11 @@ import re
 from typing import Any, Dict, Iterable, List, TypeVar, Union
 
 import regex
-import sympy
 from latex2sympy2 import latex2sympy
 from word2number import w2n
 
+import sympy
+
 # from utils import *
 
 

realhf/api/core/data_api.py

@@ -41,6 +41,7 @@ from realhf.api.cli_args import MicroBatchSpec
 from realhf.api.core import config as config_api
 from realhf.base import constants, datapack, logging, seeding
 from realhf.base.cluster import spec as cluster_spec
+from realhf.utils import load_hf_or_local_file
 
 logger = logging.getLogger("api.data")
 
@@ -756,6 +757,7 @@ def load_shuffle_split_dataset(
     dataset_path: str,
     dataset_builder: Optional[Callable[[], List[Dict[str, str]]]] = None,
 ):
+    dataset_path = load_hf_or_local_file(dataset_path)
     if dataset_path is not None:
         if dataset_path.endswith(".jsonl"):
             with open(dataset_path, "r") as f:

realhf/impl/dataset/math_code_dataset.py

@@ -14,6 +14,7 @@ import torch.utils.data
 
 from realhf.api.core import data_api
 from realhf.base import logging
+from realhf.utils import load_hf_or_local_file
 
 logger = logging.getLogger("Math Code Dataset")
 
@@ -54,6 +55,7 @@ def check_code_metadata_entries(data):
 
 def load_metadata(path):
     assert str(path).endswith(".jsonl"), path
+    path = load_hf_or_local_file(path)
     with open(path, "r") as f:
         data = [json.loads(l) for l in f.readlines()]
 

realhf/system/gserver_manager.py

@@ -185,7 +185,7 @@ class GserverManager(AsyncWorker):
                             if "num_paused_requests" in res:
                                 logger.info(
                                     f"{res['num_paused_requests']} requests are interrupted "
-                                    f"during updateing weights for server {server_index}: {server_url}"
+                                    f"during updating weights for server {server_index}: {server_url}"
                                 )
                             return
                         logger.warning(

realhf/utils.py

@@ -0,0 +1,43 @@
+def download_from_huggingface(repo_id: str, filename: str, revision: str = "main", repo_type: str = "dataset") -> str:
+    """
+    Download a file from a HuggingFace Hub repository.
+    """
+    try:
+        from huggingface_hub import hf_hub_download
+    except ImportError:
+        raise ImportError("Please install huggingface_hub to use this function: pip install huggingface_hub")
+
+    return hf_hub_download(
+        repo_id=repo_id,
+        filename=filename,
+        revision=revision,
+        repo_type=repo_type,
+    )
+
+
+def load_hf_or_local_file(path: str) -> str:
+    """
+    Load a file from a HuggingFace Hub repository or a local file.
+    hf://<org>/<repo>/<filename>
+    hf://<org>/<repo>@<revision>/<filename>
+    
+    e.g, 
+    hf-dataset://inclusionAI/AReaL-RL-Data/data/boba_106k_0319.jsonl 
+    => 
+    repo_type = dataset
+    repo_id = inclusionAI/AReaL-RL-Data
+    filename = data/boba_106k_0319.jsonl
+    revision = main
+    =>
+    /root/.cache/huggingface/hub/models--inclusionAI--AReaL-RL-Data/data/boba_106k_0319.jsonl
+    """
+    if path.startswith("hf://") or path.startswith("hf-dataset://"):
+        repo_type = "dataset" if path.startswith("hf-dataset://") else "model"
+        hf_path = path.strip().split("://")[1]
+        hf_org, hf_repo, filename = hf_path.split("/", 2)
+        repo_id = f"{hf_org}/{hf_repo}"
+        revision = "main"
+        if "@" in repo_id:
+            repo_id, revision = repo_id.split("@", 1)
+        return download_from_huggingface(repo_id, filename, revision)
+    return path