Remove redundant get_kbit_device_map()

[qgallouedec]

QLoRA scripts set model_init_kwargs["device_map"] = get_kbit_device_map() (which returns {"": local_process_index}) before loading the model. This is redundant: it never changes where the model lands. This PR removes the call everywhere and deletes the helper itself.

Scope

• examples/scripts/ (24 files) and trl/scripts/ (5 files): drop the device_map line + import.
• tests/test_grpo_trainer.py: drop the device_map=get_kbit_device_map() kwarg + import.
• trl/trainer/utils.py: delete get_kbit_device_map.
• trl/__init__.py, trl/trainer/__init__.py: remove the public export (so from trl import get_kbit_device_map no longer resolves — minor breaking change for a thin internal helper).

Background

A bitsandbytes-quantized model is placed at load time and can't be moved afterward, so QLoRA needed an explicit device_map. Born in #373 (2023) as inline device_map = {"": 0} on a quantized reward model, generalized to {"": local_process_index} for DDP in #725, then extracted into get_kbit_device_map() in #1176. A real 2023-era placement problem — which the trainer path now handles on its own (see below).

Why it's redundant

Two cases, exhaustively:

• Distributed (MULTI_GPU / DEEPSPEED) — the trainer overrides model_init_kwargs["device_map"] = None before from_pretrained, so the value from get_kbit_device_map() is discarded regardless.
• Single process — recent transformers auto-places quantized weights on the current CUDA device, which is exactly the {"": 0} the helper would have set.

How it was demonstrated

Setup: 8×H100, Qwen/Qwen2.5-0.5B, --load_in_4bit, bitsandbytes 0.49.2. Every check was run on both transformers==4.56.2 (the minimum TRL supports) and 5.13 (current), with identical outcomes.

1. Single-GPU load, no device_map : from_pretrained(quantization_config=BitsAndBytesConfig(load_in_4bit=True)) with no device_map: parameters land on cuda:0 (not CPU) and a forward pass succeeds. Rules out the "without device_map it falls back to CPU and breaks" assumption.

2. DDP placement, no device_map: 2-process run where the trainer forces device_map=None: each rank still loads onto its own local GPU (rank 0 → cuda:0, rank 1 → cuda:1) and training steps run. Per-rank placement does not depend on get_kbit_device_map().

3. Real script, with vs without the line: trl/scripts/sft.py on 2 GPUs (accelerate launch, --load_in_4bit --use_peft), baseline vs. a copy with the line removed. It runs in both cases and the losses are identical step-for-step:

   step	with line	without line
   1	2.107	2.107
   2	2.756	2.756
   3	1.775	1.775
   final train_loss	2.213	2.213

Note

Most edited scripts are trainer-loaded or pass the dict through model_init_kwargs (the path covered by checks 2–3). A few load directly via from_pretrained (e.g. sft_video_llm.py, the GRPO VLM test); these are covered by check 1 for single-GPU and are reasoned-safe under DDP (the per-rank device is set before the model loads).

---

Note

Medium Risk
Minor public API removal and reliance on transformers/trainer device placement for quantized loads; behavior was validated in the PR but QLoRA + multi-GPU remains sensitive.

Overview
Removes the get_kbit_device_map() helper and stops setting device_map from QLoRA example/CLI scripts when quantization_config is present. quantization_config is still passed unchanged; placement is left to transformers on single process and to trainers that force device_map=None under MULTI_GPU / DEEPSPEED.

from trl import get_kbit_device_map is no longer exported (minor breaking change for external callers).

DistillationTrainer and GOLDTrainer now explicitly set device_map=None when loading student/teacher from a path in distributed runs, matching the pattern already used in core trainers like DPO / GRPO.

^{Reviewed by Cursor Bugbot for commit 627c405. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 53275a2. Configure here.}

[bot-ci-comment]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.