--- license: apache-2.0 base_model: - Qwen/Qwen3-235B-A22B-FP8 - Qwen/Qwen3-1.7B datasets: - zen-E/PLatour-Qwen3-Data tags: - arxiv:2611.06976 - platour - safetensors - custom-loader --- # PLatour-Qwen3-235B-to-1.7B-Chemistry-CoT ## Paper [On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer](https://arxiv.org/html/2611.06976) PLatour explores Prefill-to-Autoregressive Cross-model Transfer (PACT): a larger sender encodes the prompt and supplies hidden representations, while a smaller receiver generates the response independently. The main method trains lightweight linear bridges between frozen backbones using task rewards, starting from zero-initialized projections and supporting hybrid-attention models. The paper also introduces PACT-Check, with Dummy Transfer, Irrelevant Transfer, Self-Communication and Trained Receiver controls. These comparisons examine whether improvements depend on useful sender information or can arise through adaptation of the receiver. Better task performance alone does not demonstrate effective cross-model communication. Experiment 13: PLatour on chemistry_cot. Bridge projections; download the referenced sender and receiver separately. [Code](https://github.com/zhenyi4/platour) · [Model collection](https://huggingface.co/collections/zen-E/platour) · [Agent/API guide](https://github.com/zhenyi4/platour/blob/main/AGENT.md) ## Project environment Use Linux with CUDA, Python 3.11 and enough GPU memory for the selected models. Run the project commands from an editable checkout so the release manifests and task files are available. See the [installation guide](https://github.com/zhenyi4/platour#getting-started). ```bash git clone https://github.com/zhenyi4/platour.git cd platour python3.11 -m pip install --user uv python3.11 scripts/install_environment.py source .venv/bin/activate ``` Training, generation and scoring use the active Python. The examples below need no `vllm_python` argument. See the [environment guide](https://github.com/zhenyi4/platour/blob/main/docs/ENVIRONMENT.md) for installation options, runtime checks and compatibility details. ## Generate an answer This checkpoint contains bridge projections, not complete model weights. Load the separately sourced frozen sender and receiver with the project API; `AutoModel.from_pretrained` cannot load the bridge as a standalone model. After the project setup above, run this Python example from the checkout root. The API receives local backbone directories and a local checkpoint file; `resolve_model_input` and `ensure_weight` prepare those paths first. ```python from platour_exps import ensure_weight, resolve_model_input from platour_api import load_session # Download/cache backbones; verify the released checkpoint against the manifest. receiver = resolve_model_input("Qwen/Qwen3-1.7B") sender = resolve_model_input("Qwen/Qwen3-235B-A22B-FP8") checkpoint = ensure_weight(13) system = load_session( family="qwen3", method="grpo", receiver=receiver, bridge=checkpoint, sender=sender, mapping="uniform_endpoints", ) answer = system.generate( messages=[{"role": "user", "content": "What is 2 + 2?"}], backend="vllm", max_new_tokens=256, output="./outputs/13-custom-prompt", ) print(answer["response"]) ``` This is a custom-prompt demo, not the published benchmark protocol. The 256-token limit is only a short example; increase it for longer answers. Use a fresh output directory for each run. Generation returns text and does not execute generated code. Qwen3 PLatour-SFT uses `last_aligned`; PLatour (GRPO) uses `uniform_endpoints`. The example sets this checkpoint's mapping explicitly. ## Evaluate the released checkpoint Run from the project environment above. This uses the published TEST inputs and the selected experiment's sampling/scoring settings; no training is performed. ```bash platour-exps evaluate --experiment 13 \ --weights zen-E/PLatour-Qwen3-235B-to-1.7B-Chemistry-CoT \ --sender Qwen/Qwen3-235B-A22B-FP8 \ --receiver Qwen/Qwen3-1.7B \ --output ./outputs/13-release-eval ``` `--sender` and `--receiver` accept HF IDs or local directories. Missing backbones, released weights and published task inputs download automatically. The release loader verifies the configuration and weight hashes. Omitting `--weights` selects the same experiment's published checkpoint. Use a new output directory. To download and verify just the released checkpoint without loading models: ```bash platour-exps download --experiment 13 ``` See the [task protocols](https://github.com/zhenyi4/platour/blob/main/qwen3/README.md) and [checkpoint index](https://github.com/zhenyi4/platour/blob/main/models/README.md#downloadable-checkpoints). ## GPU placement for project evaluation | Option | Behavior for this checkpoint | |---|---| | Omit `--n-gpu` | One process with the existing automatic placement; no automatic data-parallel replicas. | | `--n-gpu 8` | 4 independent replicas, each using 2 GPUs, with disjoint question shards. | | `--n-gpu N --gpus-per-replica G` | N total GPUs split into N / G replicas; N must be divisible by G. | The default group size for this checkpoint is **2 GPUs per replica**. `--n-gpu 1` is incompatible with this default group size. Group sizes are placement settings, not a guarantee that the model fits in memory. Every replica needs enough memory for its models and decoding workload. Use `CUDA_VISIBLE_DEVICES` to select/order GPUs. Do not combine `--n-gpu` with `torchrun`. These flags apply to the project evaluator, not the Python `generate()` examples. See [evaluation GPU defaults](https://github.com/zhenyi4/platour#gpu-defaults-for-standalone-evaluation). ## Evaluation provenance This repository distributes historical inference weights without benchmark scores. No claim about checkpoint-selection protocol or unbiased held-out performance is made. Independent evaluation and provenance review are required before using these artifacts to support new comparative research claims. ## Reproducibility and licensing Original tensor payloads are unchanged; file hashes reflect a public inference export. Exact backbone revisions used for the historical run have not been independently verified and must not be inferred from the latest Hub revision. This export does not attach benchmark results. The export uses the project's Apache-2.0 license. The referenced upstream models retain their own licenses and notices. This inference export is not optimizer/RNG resume state. ## Citation ```bibtex @misc{pact_platour, title={On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer}, author={Zhenyi Shen and Junru Lu and Wei Liu and Jiangnan Ye and Jiazheng Li and Di Yin and Xing Sun and Tameem Adel and Yali Du and Lin Gui and Yulan He}, year={2026}, eprint={2611.06976}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2611.06976}, } ```