verifiers 0.2.1.dev40__tar.gz → 0.2.1.dev42__tar.gz

{verifiers-0.2.1.dev40 → verifiers-0.2.1.dev42}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: verifiers
-Version: 0.2.1.dev40
+Version: 0.2.1.dev42
 Summary: Verifiers: Environments for LLM Reinforcement Learning
 Project-URL: Homepage, https://github.com/primeintellect-ai/verifiers
 Project-URL: Documentation, https://github.com/primeintellect-ai/verifiers

{verifiers-0.2.1.dev40 → verifiers-0.2.1.dev42}/verifiers/v1/gepa/config.py

@@ -21,13 +21,14 @@ from verifiers.v1.types import SamplingConfig
 class GEPAConfig(EnvConfig):
     """The GEPA run plus its environment. `model` runs the rollouts under optimization;
     `reflection_model` (defaults to `model`) proposes new system prompts from the reflective
-    dataset. No default for `model` — a GEPA run can spend a large eval budget, so pick it
-    explicitly rather than silently optimizing (and spending) against a fallback."""
+    dataset. `model` defaults to the same id as `EvalConfig`."""
 
     uuid: str = Field(default_factory=lambda: str(uuid4()), exclude=True)
     """Auto-generated run id — the leaf of the output dir, so runs never overwrite.
     Excluded from the saved config so re-running `@ config.toml` lands in a fresh dir."""
-    model: str = Field(..., validation_alias=AliasChoices("model", "m"))
+    model: str = Field(
+        "deepseek/deepseek-v4-flash", validation_alias=AliasChoices("model", "m")
+    )
     """Model id for rollouts under optimization."""
     client: EvalClientConfig = EvalClientConfig()
     sampling: SamplingConfig = SamplingConfig()
@@ -48,14 +49,10 @@ class GEPAConfig(EnvConfig):
     """Seed for GEPA's optimizer (candidate selection / minibatch sampling). Task shuffling
     uses a fixed seed, matching eval — so this doesn't change the train/val split."""
 
-    max_metric_calls: int = Field(
-        500, validation_alias=AliasChoices("max_metric_calls", "B")
-    )
+    max_total_rollouts: int = Field(500)
     """Total rollouts GEPA may spend across the whole optimization run."""
     reflection_minibatch_size: int = 3
     """Train tasks sampled per reflection step."""
-    perfect_score: float | None = None
-    """Skip reflecting on a minibatch that already scores this well."""
     reflection_columns: list[str] = Field(default_factory=list)
     """Extra per-trace fields (from `trace.info`, else `task`) to surface to the teacher LM."""
     initial_prompt: str | None = None

{verifiers-0.2.1.dev40 → verifiers-0.2.1.dev42}/verifiers/v1/gepa/runner.py

@@ -101,16 +101,14 @@ def run_gepa(env: Environment, config: GEPAConfig) -> GEPAResult:
                 valset=[task.data.idx for task in val_tasks],
                 adapter=adapter,
                 reflection_lm=reflection_lm,
-                max_metric_calls=config.max_metric_calls,
+                max_metric_calls=config.max_total_rollouts,
                 reflection_minibatch_size=config.reflection_minibatch_size,
                 run_dir=str(run_dir) if run_dir is not None else None,
                 seed=config.seed,
                 display_progress_bar=False,
-                skip_perfect_score=config.perfect_score is not None,
+                skip_perfect_score=False,
                 logger=_GEPALog(),
             )
-            if config.perfect_score is not None:
-                optimize_kwargs["perfect_score"] = config.perfect_score
             return optimize(**optimize_kwargs)
         finally:
             loop.run_until_complete(serving.__aexit__(None, None, None))

{verifiers-0.2.1.dev40 → verifiers-0.2.1.dev42}/verifiers/v1/harnesses/default/program.py

@@ -198,7 +198,8 @@ async def connect_mcp(stack: AsyncExitStack, config: dict) -> tuple[list[dict],
         session = await stack.enter_async_context(ClientSession(read, write))
         await session.initialize()
         for tool in (await session.list_tools()).tools:
-            full = f"{name}_{tool.name}"
+            # A server named "" (TOOL_PREFIX = None) advertises its tools bare.
+            full = f"{name}_{tool.name}" if name else tool.name
             tool_schemas.append(
                 {
                     "type": "function",

{verifiers-0.2.1.dev40 → verifiers-0.2.1.dev42}/verifiers/v1/harnesses/null/program.py

@@ -24,7 +24,9 @@ async def chat(
 
 async def connect_mcp(stack: AsyncExitStack, config: dict) -> tuple[list[dict], dict]:
     """Connect to each configured MCP server (a streamable-HTTP `url`); return
-    (tool schemas, dispatch mapping `<server>_<tool>` -> (session, raw tool name))."""
+    (tool schemas, dispatch mapping advertised name -> (session, raw tool name)). Tools are
+    advertised as `<server>_<tool>`; a server named `""` (TOOL_PREFIX = None) advertises its
+    tools bare, so names must be unique across the rollout's servers."""
     from mcp import ClientSession
     from mcp.client.streamable_http import (
         create_mcp_http_client,
@@ -43,7 +45,11 @@ async def connect_mcp(stack: AsyncExitStack, config: dict) -> tuple[list[dict],
         session = await stack.enter_async_context(ClientSession(read, write))
         await session.initialize()
         for tool in (await session.list_tools()).tools:
-            full = f"{name}_{tool.name}"
+            full = f"{name}_{tool.name}" if name else tool.name
+            if full in dispatch:
+                raise ValueError(
+                    f"duplicate tool name {full!r} across servers; keep qualified names"
+                )
             tool_schemas.append(
                 {
                     "type": "function",

{verifiers-0.2.1.dev40 → verifiers-0.2.1.dev42}/verifiers/v1/mcp/server.py

@@ -113,8 +113,10 @@ def _import_ref(ref: str) -> object:
 
 
 class ServerBase(Generic[ConfigT, StateT]):
-    # The empty value falls back to the snake-cased class name.
-    TOOL_PREFIX: ClassVar[str] = ""
+    TOOL_PREFIX: ClassVar[str | None] = ""
+    """The empty value falls back to the snake-cased class name. None advertises the server's
+    tools bare (no `<server>_` prefix); name collisions across servers are then the taskset
+    author's concern."""
 
     EXTRAS: ClassVar[tuple[str, ...]] = ()
     """Package extras the server's module needs, applied at sandbox install."""
@@ -214,6 +216,8 @@ class ServerBase(Generic[ConfigT, StateT]):
 
     @property
     def server_name(self) -> str:
+        if self.TOOL_PREFIX is None:
+            return ""
         return self.TOOL_PREFIX or "".join(
             ("_" + c.lower() if c.isupper() else c) for c in type(self).__name__
         ).lstrip("_")