verifiers 0.3.2.dev88__py3-none-any.whl → 0.3.2.dev90__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- verifiers/v1/harness.py +16 -6
- verifiers/v1/harnesses/__init__.py +0 -3
- {verifiers-0.3.2.dev88.dist-info → verifiers-0.3.2.dev90.dist-info}/METADATA +1 -1
- {verifiers-0.3.2.dev88.dist-info → verifiers-0.3.2.dev90.dist-info}/RECORD +7 -9
- verifiers/v1/harnesses/pool/__init__.py +0 -3
- verifiers/v1/harnesses/pool/harness.py +0 -96
- {verifiers-0.3.2.dev88.dist-info → verifiers-0.3.2.dev90.dist-info}/WHEEL +0 -0
- {verifiers-0.3.2.dev88.dist-info → verifiers-0.3.2.dev90.dist-info}/entry_points.txt +0 -0
- {verifiers-0.3.2.dev88.dist-info → verifiers-0.3.2.dev90.dist-info}/licenses/LICENSE +0 -0
verifiers/v1/harness.py
CHANGED
|
@@ -1,5 +1,6 @@
|
|
|
1
1
|
from __future__ import annotations
|
|
2
2
|
|
|
3
|
+
import asyncio
|
|
3
4
|
import logging
|
|
4
5
|
import os
|
|
5
6
|
from abc import ABC, abstractmethod
|
|
@@ -99,23 +100,32 @@ class Harness(ABC, Generic[ConfigT]):
|
|
|
99
100
|
"""Upload each `config.skills` folder into `runtime` at `dest/<folder name>` —
|
|
100
101
|
the program's fixed skill discovery location, which a supporting harness's
|
|
101
102
|
`setup` passes."""
|
|
103
|
+
uploads = []
|
|
104
|
+
executables = []
|
|
102
105
|
for skill in self.config.skills:
|
|
103
106
|
# Resolve so `.`/`..` entries get their real folder name (and can't
|
|
104
107
|
# place files outside `dest`).
|
|
105
108
|
skill = skill.resolve()
|
|
106
109
|
if not skill.is_dir():
|
|
107
110
|
raise ValueError(f"skill {str(skill)!r} is not a folder")
|
|
108
|
-
|
|
109
|
-
for file in sorted(skill.rglob("*")):
|
|
111
|
+
for file in skill.rglob("*"):
|
|
110
112
|
if not file.is_file():
|
|
111
113
|
continue
|
|
112
114
|
target = f"{dest}/{skill.name}/{file.relative_to(skill).as_posix()}"
|
|
113
|
-
|
|
115
|
+
uploads.append((target, file.read_bytes()))
|
|
114
116
|
if os.access(file, os.X_OK):
|
|
115
117
|
executables.append(target)
|
|
116
|
-
|
|
117
|
-
|
|
118
|
-
|
|
118
|
+
# Settle every upload before returning or surfacing an error.
|
|
119
|
+
results = await asyncio.gather(
|
|
120
|
+
*(runtime.write(target, data) for target, data in uploads),
|
|
121
|
+
return_exceptions=True,
|
|
122
|
+
)
|
|
123
|
+
for result in results:
|
|
124
|
+
if isinstance(result, BaseException):
|
|
125
|
+
raise result
|
|
126
|
+
if executables:
|
|
127
|
+
# `write` moves bytes, not modes; restore the execute bits scripts need.
|
|
128
|
+
await runtime.run(["chmod", "+x", *executables], {})
|
|
119
129
|
|
|
120
130
|
async def _check_result(
|
|
121
131
|
self, trace: Trace, runtime: Runtime, result: ProgramResult
|
|
@@ -20,7 +20,6 @@ from verifiers.v1.harnesses.mini_swe_agent import (
|
|
|
20
20
|
from verifiers.v1.harnesses.null import NullHarness, NullHarnessConfig
|
|
21
21
|
from verifiers.v1.harnesses.openclaw import OpenClawHarness, OpenClawHarnessConfig
|
|
22
22
|
from verifiers.v1.harnesses.pi import PiHarness, PiHarnessConfig
|
|
23
|
-
from verifiers.v1.harnesses.pool import PoolHarness, PoolHarnessConfig
|
|
24
23
|
from verifiers.v1.harnesses.prime_agent import (
|
|
25
24
|
PrimeAgentHarness,
|
|
26
25
|
PrimeAgentHarnessConfig,
|
|
@@ -49,8 +48,6 @@ __all__ = [
|
|
|
49
48
|
"OpenClawHarnessConfig",
|
|
50
49
|
"PiHarness",
|
|
51
50
|
"PiHarnessConfig",
|
|
52
|
-
"PoolHarness",
|
|
53
|
-
"PoolHarnessConfig",
|
|
54
51
|
"PrimeAgentHarness",
|
|
55
52
|
"PrimeAgentHarnessConfig",
|
|
56
53
|
"RLMHarness",
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.5
|
|
2
2
|
Name: verifiers
|
|
3
|
-
Version: 0.3.2.
|
|
3
|
+
Version: 0.3.2.dev90
|
|
4
4
|
Summary: Verifiers: Environments for LLM Reinforcement Learning
|
|
5
5
|
Project-URL: Homepage, https://github.com/primeintellect-ai/verifiers
|
|
6
6
|
Project-URL: Documentation, https://github.com/primeintellect-ai/verifiers
|
|
@@ -6,7 +6,7 @@ verifiers/v1/env.py,sha256=CEhzjojhg5jqbOYruUnc5U7TtgDHDSx8NiiFA0qfbt8,18234
|
|
|
6
6
|
verifiers/v1/episode.py,sha256=Fft0R2W6INAVeJhoE9b8h7L_KQSz7XuxP4FT_XhcsUY,5741
|
|
7
7
|
verifiers/v1/errors.py,sha256=uQySmCUrjDYFkBIii1vBFPiEcPFPQZ9EGpas_6HvcQI,5279
|
|
8
8
|
verifiers/v1/graph.py,sha256=K6wKkoT_8TpNqPQISu2Xf0eHtAYW_KUGw0CteN5zr3c,36529
|
|
9
|
-
verifiers/v1/harness.py,sha256=
|
|
9
|
+
verifiers/v1/harness.py,sha256=sXEjVdrS4A2SO35C_uEebBVeUt8VL7dui7IneAp2MRs,14124
|
|
10
10
|
verifiers/v1/judge.py,sha256=uqn-djkK2_7Q3kOjkDo6ipOvV_HOvuLa9y6AoDjRux8,9618
|
|
11
11
|
verifiers/v1/rollout.py,sha256=SvMgs7YBNjiftp4vlGVBMEseUsJs1h7qrqAlxmWaNwY,23987
|
|
12
12
|
verifiers/v1/semantic.py,sha256=VQ_buw6Ya6QtRoh_fiYI5Avs5RmIQCfWH0H0bQc4aoc,4704
|
|
@@ -82,7 +82,7 @@ verifiers/v1/gepa/config.py,sha256=PhZRLkhaPciNm_IhwVX3e9DM8Y2IdCe4Fbi5WjH5aro,4
|
|
|
82
82
|
verifiers/v1/gepa/dataset.py,sha256=hFuR7dou4VSKa0ujb8sFmGm9a19mjrsmTa1MMW1Kv7I,2177
|
|
83
83
|
verifiers/v1/gepa/reflection.py,sha256=ptHhx0lcDLcWhWhXuslh9T8gPVNpnX5XLPtXoneW1yA,1054
|
|
84
84
|
verifiers/v1/gepa/runner.py,sha256=fGEyFzejdQKBzBbvdiTDHV3JtRhXMXMGSBU5y8C8i3w,6054
|
|
85
|
-
verifiers/v1/harnesses/__init__.py,sha256=
|
|
85
|
+
verifiers/v1/harnesses/__init__.py,sha256=2JPwrRoYoigH-HfBtiFnatRdFEVHLv4rwlKfsFEibyE,1803
|
|
86
86
|
verifiers/v1/harnesses/node.py,sha256=-lDPh30p65IETPVLkKn4yhEK49rtEGVQf9TNs7MVQYE,2133
|
|
87
87
|
verifiers/v1/harnesses/bash/__init__.py,sha256=IAV4eMMqDHSl-9ANb0Q3kxmBZ22SeHZFBM448X4gQio,140
|
|
88
88
|
verifiers/v1/harnesses/bash/harness.py,sha256=nsjLKFMGFSya4s9fMHiNVrO92RRCTS3VHqTK3rv3AFA,5233
|
|
@@ -107,8 +107,6 @@ verifiers/v1/harnesses/openclaw/__init__.py,sha256=kNyfPB1sGwEf4K3mTfKN0VEyxun3S
|
|
|
107
107
|
verifiers/v1/harnesses/openclaw/harness.py,sha256=tucrQb_Osi7KAEtpyO4i55IXI5SFtLnyItB8eV4kjng,10534
|
|
108
108
|
verifiers/v1/harnesses/pi/__init__.py,sha256=1mSnzOf-RdEchdeg6IHtcdobG5KIpAhuTdUIgsCIn74,117
|
|
109
109
|
verifiers/v1/harnesses/pi/harness.py,sha256=11V9d6_npjE9vEEgIG7y80kfaX1Ilfi3lbYsqGsN1NE,6496
|
|
110
|
-
verifiers/v1/harnesses/pool/__init__.py,sha256=HTYsNiWGdiNEsSHj4xhPmshJSNb7SHjQ7d4yZQ4p3HM,127
|
|
111
|
-
verifiers/v1/harnesses/pool/harness.py,sha256=TsELYhW4mMOt_laiDN6nAQJxFL3iCcKvckQWYve2XiE,3571
|
|
112
110
|
verifiers/v1/harnesses/prime_agent/__init__.py,sha256=xfPC4RrlnyYWaf38tPAOZq4YL6M55lMpj5BoRLmaK8I,171
|
|
113
111
|
verifiers/v1/harnesses/prime_agent/harness.py,sha256=cMvkNdqEutzCoX3lrmCQ5dAi_HSDV3Vhw6K1N-LiBtA,10630
|
|
114
112
|
verifiers/v1/harnesses/rlm/__init__.py,sha256=hwx51xTdEWzzYgMj-p4ETNcptoub3IPMDbEdJ4uB8Ns,122
|
|
@@ -190,8 +188,8 @@ verifiers/v1/utils/prime.py,sha256=UTYRjp9cbjNb6CVmBHda-1wWZAIIfxNOmTyNuT7_wL4,9
|
|
|
190
188
|
verifiers/v1/utils/retries.py,sha256=Y2ZgrAjn-qNkRKZP_RVNL_7EK0iaRcZeCa404lpGYi4,5417
|
|
191
189
|
verifiers/v1/utils/score.py,sha256=493yJVMw8teCu9JxapxMFPFzI0hNUqdo0Y2nGW4kckk,6200
|
|
192
190
|
verifiers/v1/utils/version.py,sha256=-obEo_-l9-D8FLef4hYxncOe-uJpxrM1g2Hig_37Sgs,1607
|
|
193
|
-
verifiers-0.3.2.
|
|
194
|
-
verifiers-0.3.2.
|
|
195
|
-
verifiers-0.3.2.
|
|
196
|
-
verifiers-0.3.2.
|
|
197
|
-
verifiers-0.3.2.
|
|
191
|
+
verifiers-0.3.2.dev90.dist-info/METADATA,sha256=Ero7sRPEh8DRnSrFcAecPIpeoC9mXaqXEO61PJwRmOQ,4237
|
|
192
|
+
verifiers-0.3.2.dev90.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
|
|
193
|
+
verifiers-0.3.2.dev90.dist-info/entry_points.txt,sha256=uqQje0TMsr7k6nXlWxlPs13Si0evb3pVCEDF6c-YL80,241
|
|
194
|
+
verifiers-0.3.2.dev90.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
|
|
195
|
+
verifiers-0.3.2.dev90.dist-info/RECORD,,
|
|
@@ -1,96 +0,0 @@
|
|
|
1
|
-
"""Run Poolside's native ACP server against interception."""
|
|
2
|
-
|
|
3
|
-
import json
|
|
4
|
-
import shlex
|
|
5
|
-
|
|
6
|
-
from verifiers.v1.acp import ACPConfig, ACPHarness
|
|
7
|
-
from verifiers.v1.clients import ModelContext
|
|
8
|
-
from verifiers.v1.configs.harness import HarnessConfig, PinnedVersion
|
|
9
|
-
from verifiers.v1.harnesses.utils.install import ensure_installed
|
|
10
|
-
from verifiers.v1.runtimes import Runtime
|
|
11
|
-
from verifiers.v1.task import TaskData
|
|
12
|
-
from verifiers.v1.trace import Trace
|
|
13
|
-
|
|
14
|
-
POOL_DIR = "/tmp/vf-pool-{version}"
|
|
15
|
-
SKILLS_DIR = ".poolside/skills"
|
|
16
|
-
INSTALL = r"""
|
|
17
|
-
set -e
|
|
18
|
-
command -v curl >/dev/null || (apt-get update -qq && apt-get install -y -qq curl ca-certificates >/dev/null)
|
|
19
|
-
command -v tar >/dev/null || (apt-get update -qq && apt-get install -y -qq tar >/dev/null)
|
|
20
|
-
case "$(uname -s)" in Linux) os=linux ;; Darwin) os=darwin ;; *) echo "unsupported os: $(uname -s)" >&2; exit 1 ;; esac
|
|
21
|
-
case "$(uname -m)" in aarch64|arm64) arch=arm64 ;; x86_64|amd64) arch=amd64 ;; *) echo "unsupported arch: $(uname -m)" >&2; exit 1 ;; esac
|
|
22
|
-
mkdir -p {dir}
|
|
23
|
-
curl -fsSL "https://github.com/poolsideai/pool/releases/download/v{version}/pool-$os-$arch.tar.gz" | tar -xz -C {dir}
|
|
24
|
-
mv "{dir}/pool-$os-$arch" "{dir}/pool"
|
|
25
|
-
chmod +x "{dir}/pool"
|
|
26
|
-
"""
|
|
27
|
-
|
|
28
|
-
|
|
29
|
-
class PoolHarnessConfig(HarnessConfig):
|
|
30
|
-
version: PinnedVersion = "1.0.15"
|
|
31
|
-
"""Pool release to install, pinned for reproducibility."""
|
|
32
|
-
|
|
33
|
-
|
|
34
|
-
class PoolHarness(ACPHarness[PoolHarnessConfig]):
|
|
35
|
-
APPENDS_SYSTEM_PROMPT = True
|
|
36
|
-
SUPPORTS_MCP = True
|
|
37
|
-
SUPPORTS_SKILLS = True
|
|
38
|
-
|
|
39
|
-
async def setup(self, runtime: Runtime) -> None:
|
|
40
|
-
await self.install_skills(runtime, SKILLS_DIR)
|
|
41
|
-
directory = POOL_DIR.format(version=self.config.version)
|
|
42
|
-
binary = f"{directory}/pool"
|
|
43
|
-
script = INSTALL.replace("{version}", self.config.version).replace(
|
|
44
|
-
"{dir}", directory
|
|
45
|
-
)
|
|
46
|
-
await ensure_installed(
|
|
47
|
-
runtime,
|
|
48
|
-
directory=directory,
|
|
49
|
-
ready=f"[ -x {binary} ]",
|
|
50
|
-
install=script,
|
|
51
|
-
env=self.config.resolved_env,
|
|
52
|
-
label="Pool",
|
|
53
|
-
shell=("bash", "-o", "pipefail", "-c"),
|
|
54
|
-
)
|
|
55
|
-
await super().setup(runtime)
|
|
56
|
-
|
|
57
|
-
async def prepare_acp(
|
|
58
|
-
self,
|
|
59
|
-
ctx: ModelContext,
|
|
60
|
-
trace: Trace,
|
|
61
|
-
runtime: Runtime,
|
|
62
|
-
endpoint: str,
|
|
63
|
-
secret: str,
|
|
64
|
-
mcp_urls: dict[str, str],
|
|
65
|
-
data: TaskData,
|
|
66
|
-
) -> ACPConfig:
|
|
67
|
-
system_prompt, prompt = self.resolve_prompt(data)
|
|
68
|
-
env = {
|
|
69
|
-
**self.config.resolved_env,
|
|
70
|
-
# Standalone provider mode sends this bearer and model to interception.
|
|
71
|
-
"POOLSIDE_API_KEY": secret,
|
|
72
|
-
"POOLSIDE_STANDALONE_BASE_URL": endpoint,
|
|
73
|
-
"POOLSIDE_STANDALONE_MODEL": ctx.model,
|
|
74
|
-
}
|
|
75
|
-
pool_home = f".vf-pool/{trace.id}"
|
|
76
|
-
# Values are Pool tool names such as `shell`, `read`, or `edit`.
|
|
77
|
-
tools = {name: {"disabled": True} for name in self.config.disabled_tools or []}
|
|
78
|
-
settings = shlex.quote(json.dumps({"tools": tools}))
|
|
79
|
-
command = [
|
|
80
|
-
"sh",
|
|
81
|
-
"-c",
|
|
82
|
-
(
|
|
83
|
-
f'export HOME="$PWD/{pool_home}/home" '
|
|
84
|
-
f'XDG_CONFIG_HOME="$PWD/{pool_home}/config" '
|
|
85
|
-
f'XDG_STATE_HOME="$PWD/{pool_home}/state"; '
|
|
86
|
-
f"exec {POOL_DIR.format(version=self.config.version)}/pool acp "
|
|
87
|
-
f"--sandbox disabled --settings {settings}"
|
|
88
|
-
),
|
|
89
|
-
]
|
|
90
|
-
return ACPConfig(
|
|
91
|
-
env=env,
|
|
92
|
-
command=command,
|
|
93
|
-
prompt=prompt,
|
|
94
|
-
mcp_urls=mcp_urls,
|
|
95
|
-
system_prompt=system_prompt,
|
|
96
|
-
)
|
|
File without changes
|
|
File without changes
|
|
File without changes
|