Search results

[RFC] Deprecate TorchTune service in favor of Unsloth/Megatron

Status: Open.

#514 In OpenPipe/ART;

· corbt opened on Jan 16, 2026

[RFC] Multi-Checkpoint Inference Support for Pipelined Training

Status: Open.

#513 In OpenPipe/ART;

· corbt opened on Jan 16, 2026

RFC: move trajectory logging from the backend to the frontend

Status: Open.

#511 In OpenPipe/ART;

· corbt opened on Jan 16, 2026

RFC: Deprecating <code>SkypilotBackend</code>

Status: Open.

#510 In OpenPipe/ART;

· corbt opened on Jan 16, 2026

W&amp;B offline

Status: Open.

#493 In OpenPipe/ART;

· manglu097 opened on Dec 19, 2025

Support vllm&gt;0.10.0

Status: Open.

#490 In OpenPipe/ART;

· xavier-owkin opened on Dec 17, 2025

Unsloth + vLLM multi-GPU setup in ART crashes when restricting CUDA_VISIBLE_DEVICES or using tensor parallelism &gt;1 on H100 GPUs

Status: Open.

#478 In OpenPipe/ART;

· ansh-info opened on Dec 11, 2025

RULER produces scores &gt;1, arbitrary outputs (7, 10, 15…), nondeterministic behavior, and malformed JSON when judge model is Qwen3-32B running on Ollama

Status: Open.

#477 In OpenPipe/ART;

· ansh-info opened on Dec 11, 2025

RULER + Litellm with Ollama: <code>ollama_chat</code> recommended, but <code>/api/chat</code> cannot generate required JSON schema

Status: Open.

#476 In OpenPipe/ART;

· ansh-info opened on Dec 11, 2025

RULER scoring + training tightly coupled to Litellm/OpenAI, cannot cleanly use ChatOllama/ChatNVIDIA as judge/inference models

Status: Open.

#475 In OpenPipe/ART;

· ansh-info opened on Dec 11, 2025

<code>init_chat_model</code> always uses <code>ChatOpenAI</code>, ignores args, and still calls OpenAI / hits 10-minute timeout with Ollama

Status: Open.

#474 In OpenPipe/ART;

· ansh-info opened on Dec 11, 2025

Context window incorrectly capped at 8192 tokens when using Ollama (token count exceeded 8192)

Status: Open.

#473 In OpenPipe/ART;

· ansh-info opened on Dec 11, 2025