verifiers 0.3.2.dev97__py3-none-any.whl → 0.3.2.dev98__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/harnesses/rlm/harness.py +11 -0
- verifiers/v1/trace.py +2 -1
- {verifiers-0.3.2.dev97.dist-info → verifiers-0.3.2.dev98.dist-info}/METADATA +1 -1
- {verifiers-0.3.2.dev97.dist-info → verifiers-0.3.2.dev98.dist-info}/RECORD +7 -7
- {verifiers-0.3.2.dev97.dist-info → verifiers-0.3.2.dev98.dist-info}/WHEEL +0 -0
- {verifiers-0.3.2.dev97.dist-info → verifiers-0.3.2.dev98.dist-info}/entry_points.txt +0 -0
- {verifiers-0.3.2.dev97.dist-info → verifiers-0.3.2.dev98.dist-info}/licenses/LICENSE +0 -0
|
@@ -41,6 +41,15 @@ class _SessionSnapshot(BaseModel):
|
|
|
41
41
|
|
|
42
42
|
session_id: str = Field(pattern=r"^[A-Za-z0-9._:-]{1,128}$")
|
|
43
43
|
metrics: dict[str, int | float]
|
|
44
|
+
last_stop_reason: str | None = None
|
|
45
|
+
|
|
46
|
+
|
|
47
|
+
_TRUNCATION_STOPS = {
|
|
48
|
+
"max_total_tokens": "max_total_tokens",
|
|
49
|
+
"max_total_turns": "max_turns",
|
|
50
|
+
"token_budget": "max_output_tokens",
|
|
51
|
+
"compaction_failed": "compaction_failed",
|
|
52
|
+
}
|
|
44
53
|
|
|
45
54
|
|
|
46
55
|
class CompactionConfig(BaseConfig):
|
|
@@ -251,6 +260,8 @@ class RLMHarness(ACPHarness[RLMHarnessConfig]):
|
|
|
251
260
|
if snapshot.session_id != trace.id:
|
|
252
261
|
raise ValueError("RLM session snapshot does not match the rollout")
|
|
253
262
|
trace.record_metrics(snapshot.metrics)
|
|
263
|
+
if condition := _TRUNCATION_STOPS.get(snapshot.last_stop_reason or ""):
|
|
264
|
+
trace.stop(condition)
|
|
254
265
|
|
|
255
266
|
def acp_turn_result(self, trace: Trace, result: ACPTurn) -> None:
|
|
256
267
|
self._consume_snapshot(trace, result.response_metadata)
|
verifiers/v1/trace.py
CHANGED
|
@@ -621,12 +621,13 @@ class Trace(BaseModel, Generic[DataT, StateT, AgentConfigT]):
|
|
|
621
621
|
|
|
622
622
|
@property
|
|
623
623
|
def is_truncated(self) -> bool:
|
|
624
|
-
"""True for framework limits or a length-finished
|
|
624
|
+
"""True for framework limits, failed compaction, or a length-finished response."""
|
|
625
625
|
if self.stop_condition in (
|
|
626
626
|
"max_turns",
|
|
627
627
|
"max_input_tokens",
|
|
628
628
|
"max_output_tokens",
|
|
629
629
|
"max_total_tokens",
|
|
630
|
+
"compaction_failed",
|
|
630
631
|
):
|
|
631
632
|
return True
|
|
632
633
|
last = next((c for c in reversed(self.calls) if c.error is None), None)
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.5
|
|
2
2
|
Name: verifiers
|
|
3
|
-
Version: 0.3.2.
|
|
3
|
+
Version: 0.3.2.dev98
|
|
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
|
|
@@ -14,7 +14,7 @@ verifiers/v1/session.py,sha256=7iarEDRH2epID_O8ai-1HwveXze_o8shnLw9pFMFX5o,18900
|
|
|
14
14
|
verifiers/v1/state.py,sha256=EckF2bWp-vV4b1jYJ9sLI5xrfGuI5spIgYYwW926toI,595
|
|
15
15
|
verifiers/v1/task.py,sha256=9YhllEDgoHSz72-1fcYwBADl-rOzASnrSVimLheWmwI,12237
|
|
16
16
|
verifiers/v1/taskset.py,sha256=fp2E0IEhL_Ybj9cegZwljfTmW27p_30zTWHFKw_kXE4,4374
|
|
17
|
-
verifiers/v1/trace.py,sha256=
|
|
17
|
+
verifiers/v1/trace.py,sha256=6ny9D7qXcB2NwN97UstxWCrT4_rmAAHCcW-buS7XwYs,30549
|
|
18
18
|
verifiers/v1/types.py,sha256=PH1SxRWXOjPCgQD9aZgM18_bECvFSNQ25gkCbulmTKU,9857
|
|
19
19
|
verifiers/v1/acp/__init__.py,sha256=Zx5FhpkpQHETc-FB_zI3Qt4TxmuMnbv7YxxXk2npjAk,12922
|
|
20
20
|
verifiers/v1/acp/runner.py,sha256=ODrvrC_8DDsAoif9jr4eYJsQzesPxDorxqqm-lo5tNY,11403
|
|
@@ -110,7 +110,7 @@ verifiers/v1/harnesses/pi/harness.py,sha256=UwqzN8-PkGzo2PXsZaYiudJdz1jSpzNtXT5v
|
|
|
110
110
|
verifiers/v1/harnesses/prime_agent/__init__.py,sha256=xfPC4RrlnyYWaf38tPAOZq4YL6M55lMpj5BoRLmaK8I,171
|
|
111
111
|
verifiers/v1/harnesses/prime_agent/harness.py,sha256=cMvkNdqEutzCoX3lrmCQ5dAi_HSDV3Vhw6K1N-LiBtA,10630
|
|
112
112
|
verifiers/v1/harnesses/rlm/__init__.py,sha256=hwx51xTdEWzzYgMj-p4ETNcptoub3IPMDbEdJ4uB8Ns,122
|
|
113
|
-
verifiers/v1/harnesses/rlm/harness.py,sha256=
|
|
113
|
+
verifiers/v1/harnesses/rlm/harness.py,sha256=70HXfw88G1dfDhSCjrqFUKRvY7a8ZIp0jifPc4EJNM8,12250
|
|
114
114
|
verifiers/v1/harnesses/terminus_2/__init__.py,sha256=l1pwfg1HyVDEqT9efdnFWcXqaZwtKwe7TrplkyxZSA0,166
|
|
115
115
|
verifiers/v1/harnesses/terminus_2/harness.py,sha256=DMTY1bzvydyjHqzmvBSBGKJsmUzYEgxqd215a7cG7tg,3052
|
|
116
116
|
verifiers/v1/harnesses/terminus_2/program.py,sha256=X8MZdyItU3gbXGXnGOlJ9LWcUTnqIDaxlhkZvoRiqKY,2703
|
|
@@ -188,8 +188,8 @@ verifiers/v1/utils/prime.py,sha256=UTYRjp9cbjNb6CVmBHda-1wWZAIIfxNOmTyNuT7_wL4,9
|
|
|
188
188
|
verifiers/v1/utils/retries.py,sha256=Y2ZgrAjn-qNkRKZP_RVNL_7EK0iaRcZeCa404lpGYi4,5417
|
|
189
189
|
verifiers/v1/utils/score.py,sha256=493yJVMw8teCu9JxapxMFPFzI0hNUqdo0Y2nGW4kckk,6200
|
|
190
190
|
verifiers/v1/utils/version.py,sha256=-obEo_-l9-D8FLef4hYxncOe-uJpxrM1g2Hig_37Sgs,1607
|
|
191
|
-
verifiers-0.3.2.
|
|
192
|
-
verifiers-0.3.2.
|
|
193
|
-
verifiers-0.3.2.
|
|
194
|
-
verifiers-0.3.2.
|
|
195
|
-
verifiers-0.3.2.
|
|
191
|
+
verifiers-0.3.2.dev98.dist-info/METADATA,sha256=yEEXS9iJ0vQSP-gb56cGxxa1DmA2Ffm2Y6g9mlXeqrY,4237
|
|
192
|
+
verifiers-0.3.2.dev98.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
|
|
193
|
+
verifiers-0.3.2.dev98.dist-info/entry_points.txt,sha256=uqQje0TMsr7k6nXlWxlPs13Si0evb3pVCEDF6c-YL80,241
|
|
194
|
+
verifiers-0.3.2.dev98.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
|
|
195
|
+
verifiers-0.3.2.dev98.dist-info/RECORD,,
|
|
File without changes
|
|
File without changes
|
|
File without changes
|