verifiers 0.3.2.dev32__py3-none-any.whl → 0.3.2.dev33__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/graph.py +6 -0
- verifiers/v1/trace.py +14 -0
- {verifiers-0.3.2.dev32.dist-info → verifiers-0.3.2.dev33.dist-info}/METADATA +1 -1
- {verifiers-0.3.2.dev32.dist-info → verifiers-0.3.2.dev33.dist-info}/RECORD +7 -7
- {verifiers-0.3.2.dev32.dist-info → verifiers-0.3.2.dev33.dist-info}/WHEEL +0 -0
- {verifiers-0.3.2.dev32.dist-info → verifiers-0.3.2.dev33.dist-info}/entry_points.txt +0 -0
- {verifiers-0.3.2.dev32.dist-info → verifiers-0.3.2.dev33.dist-info}/licenses/LICENSE +0 -0
verifiers/v1/graph.py
CHANGED
|
@@ -121,6 +121,12 @@ class MessageNode(BaseModel):
|
|
|
121
121
|
reference_logprobs: list[float] | None = None
|
|
122
122
|
"""Reference-model logprobs over the sampled tokens, in the same compact layout as
|
|
123
123
|
`logprobs`. None means no reference model scored this node."""
|
|
124
|
+
trainer_logprobs: list[float] | None = None
|
|
125
|
+
"""Trainer-recomputed logprobs over the sampled tokens, in the same compact layout as
|
|
126
|
+
`logprobs`. None means no trainer forward annotated this node."""
|
|
127
|
+
entropies: list[float] | None = None
|
|
128
|
+
"""Trainer policy entropies over the sampled tokens, in the same compact layout as
|
|
129
|
+
`logprobs`. None means no trainer forward annotated this node."""
|
|
124
130
|
loss_weights: dict[str, list[float]] | None = None
|
|
125
131
|
"""Named loss-weight streams aligned to `token_ids`, consumer-stamped."""
|
|
126
132
|
multi_modal_data: SkipJsonSchema[MultiModalData | None] = None
|
verifiers/v1/trace.py
CHANGED
|
@@ -271,6 +271,20 @@ class Branch(BaseModel):
|
|
|
271
271
|
return None
|
|
272
272
|
return self.spread(lambda node: node.reference_logprobs)
|
|
273
273
|
|
|
274
|
+
@property
|
|
275
|
+
def trainer_logprobs(self) -> list[float] | None:
|
|
276
|
+
"""Trainer-recomputed logprobs aligned to `token_ids`, or None when unannotated."""
|
|
277
|
+
if all(node.trainer_logprobs is None for node in self.nodes):
|
|
278
|
+
return None
|
|
279
|
+
return self.spread(lambda node: node.trainer_logprobs)
|
|
280
|
+
|
|
281
|
+
@property
|
|
282
|
+
def entropies(self) -> list[float] | None:
|
|
283
|
+
"""Trainer policy entropies aligned to `token_ids`, or None when unannotated."""
|
|
284
|
+
if all(node.entropies is None for node in self.nodes):
|
|
285
|
+
return None
|
|
286
|
+
return self.spread(lambda node: node.entropies)
|
|
287
|
+
|
|
274
288
|
def loss_weights(self, name: str) -> list[float] | None:
|
|
275
289
|
"""One named loss-weight stream aligned to `token_ids`, or None when absent."""
|
|
276
290
|
if all(
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.5
|
|
2
2
|
Name: verifiers
|
|
3
|
-
Version: 0.3.2.
|
|
3
|
+
Version: 0.3.2.dev33
|
|
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
|
|
@@ -4,7 +4,7 @@ verifiers/v1/agent.py,sha256=QvP7jJMR1NQEzAdcqSxcPRgYYTZYt6gxst3vxmXymOE,32384
|
|
|
4
4
|
verifiers/v1/env.py,sha256=CEhzjojhg5jqbOYruUnc5U7TtgDHDSx8NiiFA0qfbt8,18234
|
|
5
5
|
verifiers/v1/episode.py,sha256=uStvI4WqgJT50hmHIB55m5B-ASqqLz-fZ-ynBz9uzEY,5474
|
|
6
6
|
verifiers/v1/errors.py,sha256=uQySmCUrjDYFkBIii1vBFPiEcPFPQZ9EGpas_6HvcQI,5279
|
|
7
|
-
verifiers/v1/graph.py,sha256=
|
|
7
|
+
verifiers/v1/graph.py,sha256=y8B9mW-8BUEs0B7vg7YEU1WblfsgagIHmDCK0MWujSQ,36193
|
|
8
8
|
verifiers/v1/harness.py,sha256=ziECqEBWV34UWbLlxrkhY6MwkbVYkoXBOoaY7mZxwTs,14753
|
|
9
9
|
verifiers/v1/judge.py,sha256=Pvr0C41ah1qkNSZ0WXdkDvBMfv5YSfY1REHncwlJ8cg,9346
|
|
10
10
|
verifiers/v1/rollout.py,sha256=hoKGn1CbShbyjEM9PgYq8pdKgoC91VebTkBSgUHLtiE,23978
|
|
@@ -13,7 +13,7 @@ verifiers/v1/session.py,sha256=xM7JJJVJACk0JfH9RAN1Ts8gJJKr2tGGRiIgQTrdieY,18692
|
|
|
13
13
|
verifiers/v1/state.py,sha256=EckF2bWp-vV4b1jYJ9sLI5xrfGuI5spIgYYwW926toI,595
|
|
14
14
|
verifiers/v1/task.py,sha256=vaVkBu00riiNU1Ke4NeeY2NQTXxOguz9d6NLe55hO_Q,11468
|
|
15
15
|
verifiers/v1/taskset.py,sha256=fp2E0IEhL_Ybj9cegZwljfTmW27p_30zTWHFKw_kXE4,4374
|
|
16
|
-
verifiers/v1/trace.py,sha256=
|
|
16
|
+
verifiers/v1/trace.py,sha256=Adj9iR7ilLE0UscNMiiQTt3NMPPFPx1wpYP17UVr1Bc,29051
|
|
17
17
|
verifiers/v1/types.py,sha256=a1wqL8T7KQg9WxkaDIZpmOyU6xT2VeSr1K4SWiW8ef4,9660
|
|
18
18
|
verifiers/v1/acp/__init__.py,sha256=wUPZVYBFjJuOnIIoYONBFuA04A1DJf_dOyKY3QwK37Q,12722
|
|
19
19
|
verifiers/v1/acp/runner.py,sha256=gMXFmsVg2qD0Zhrx7kvC6Me2Z-wsfBamRIO0i45UW8g,11097
|
|
@@ -185,8 +185,8 @@ verifiers/v1/utils/prime.py,sha256=UTYRjp9cbjNb6CVmBHda-1wWZAIIfxNOmTyNuT7_wL4,9
|
|
|
185
185
|
verifiers/v1/utils/retries.py,sha256=Y2ZgrAjn-qNkRKZP_RVNL_7EK0iaRcZeCa404lpGYi4,5417
|
|
186
186
|
verifiers/v1/utils/score.py,sha256=493yJVMw8teCu9JxapxMFPFzI0hNUqdo0Y2nGW4kckk,6200
|
|
187
187
|
verifiers/v1/utils/version.py,sha256=-obEo_-l9-D8FLef4hYxncOe-uJpxrM1g2Hig_37Sgs,1607
|
|
188
|
-
verifiers-0.3.2.
|
|
189
|
-
verifiers-0.3.2.
|
|
190
|
-
verifiers-0.3.2.
|
|
191
|
-
verifiers-0.3.2.
|
|
192
|
-
verifiers-0.3.2.
|
|
188
|
+
verifiers-0.3.2.dev33.dist-info/METADATA,sha256=kK1FyrtG26OzrKqPEZXVv-Us3NU2OcvSGNbGN4Ji5c8,4144
|
|
189
|
+
verifiers-0.3.2.dev33.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
|
|
190
|
+
verifiers-0.3.2.dev33.dist-info/entry_points.txt,sha256=uqQje0TMsr7k6nXlWxlPs13Si0evb3pVCEDF6c-YL80,241
|
|
191
|
+
verifiers-0.3.2.dev33.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
|
|
192
|
+
verifiers-0.3.2.dev33.dist-info/RECORD,,
|
|
File without changes
|
|
File without changes
|
|
File without changes
|