NousResearch/atropos
1
0
Fork 0
mirror of https://github.com/NousResearch/atropos.git synced 2026-04-19 12:57:58 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

readme updates

Browse source
This commit is contained in:
Jai Suphavadeeprasit 2026-03-03 12:19:55 -05:00
parent b9291aa29f
commit 323a8a2601
2 changed files with 47 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

13
atroposlib/envs/server_handling/README.md
View file

@ -8,6 +8,19 @@ For automatic token and logprob tracking, see the [ManagedServer Guide](MANAGED_
> **Note:** OpenAI endpoints do not support token IDs/logprobs required for ManagedServer. Set `ATROPOS_ALLOW_DUMMY_MANAGED_SERVER=1` to use a placeholder implementation for testing/evaluation. See [OpenAI Endpoint Limitations](MANAGED_SERVER.md#openai-endpoint-limitations) for details.
### Normalized `get_logprobs` API
`ManagedServer` and server backends now expose a normalized `get_logprobs(...)` interface so callers can consume a single schema across backends:
- `prompt_tokens`
- `sequence_token_ids`
- `sequence_logprobs`
- `sequence_topk_token_ids`
- `sequence_topk_logprobs`
- `finish_reasons`
For backends that only expose sampled-token logprobs, top-k arrays are synthesized with `k=1` for interface compatibility.
## Reasoning Model Support
The `ReasoningConfig` class enables support for reasoning/thinking models across different providers.