Search results

Update transformers to v0.51.

enhancementNew feature or requestNew feature or request

Status: Open.

#575 In openpipe/art;

· Kovbo opened on Feb 21, 2026

Training only processes one trajectory at a time?

questionFurther information is requestedFurther information is requested

Status: Open.

#533 In openpipe/art;

· junfengzh opened on Jan 25, 2026

RFC: Deprecate allow_training_without_logprobs option

discussion

Status: Open.

#528 In openpipe/art;

· corbt opened on Jan 21, 2026

W&B offline

questionFurther information is requestedFurther information is requested

Status: Open.

#493 In openpipe/art;

· manglu097 opened on Dec 19, 2025

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

bugSomething isn't workingSomething isn't working

Status: Open.

#478 In openpipe/art;

· ansh-info opened on Dec 11, 2025

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

bugSomething isn't workingSomething isn't working

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

bugSomething isn't workingSomething isn't working

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

bugSomething isn't workingSomething isn't working

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

bugSomething isn't workingSomething isn't working

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)

bugSomething isn't workingSomething isn't working

Status: Open.

#473 In openpipe/art;

· ansh-info opened on Dec 11, 2025

Training hangs at 0/12 with Qwen2.5-14B + Unsloth/vLLM (KV cache reported as 0.0 GB despite free VRAM)

bugSomething isn't workingSomething isn't working

Status: Open.

#472 In openpipe/art;

· ansh-info opened on Dec 11, 2025

Is it not possible to set the rollout engine to multiple gpus?

questionFurther information is requestedFurther information is requested

Status: Open.

#470 In openpipe/art;

· melanibe opened on Dec 10, 2025