Downloads · 30 days
27
23% of all-time downloads
openevo-recovery/openevo-qwen25-7b-webshop-sd-lora
openevo-qwen25-7b-webshop-sd-lora is a machine learning model from openevo-recovery. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as apache-2.0.
本仓库是 Hugging Face 上的 模型仓库 openevo-recovery/openevo-qwen25-7b-webshop-sd-lora。这段说明只提供双语导航,不修改原始模型、数据、实验结论或许可证。具体用途、身份、版本、限制及恢复方式,以本仓库保留的原始技术说明、Manifest 和源项目权威记录为准。
Downloads · 30 days
27
23% of all-time downloads
All-time downloads
119
Public
Repo size
80.9 MB
Likes
0
Public
Click a slice to open those files.
.safetensors80.9 MB · 100%
From the Hugging Face model README
本仓库是 Hugging Face 上的 模型仓库 openevo-recovery/openevo-qwen25-7b-webshop-sd-lora。这段说明只提供双语导航,不修改原始模型、数据、实验结论或许可证。具体用途、身份、版本、限制及恢复方式,以本仓库保留的原始技术说明、Manifest 和源项目权威记录为准。
This is the Hugging Face model repository openevo-recovery/openevo-qwen25-7b-webshop-sd-lora. This is a bilingual navigation note only; it does not change original model/data bytes, experiment findings, or licenses. For purpose, identity, version, limitations, and recovery, consult the preserved original technical documentation, manifests, and source-project authority.
这是 OpenEVO 跨平台实验资产的一部分。本段只补充统一身份与导航信息,不修改仓库已有模型、数据、checkpoint、trajectory 或历史科学语义。
miyuki17/openevo-qwen25-7b-webshop-sd-loraPUBLIC_FROZEN_SD_LORAhistorical-frozen-policy-artifacttruepublicregistry-resolvedKEEP_STABLE_IDmykcs/openevo-experiment@bb57c88eb5de3036dee3ea9b095bc550dd6882c8mykcs/openevo-experiment@a1994980f16a922374f25442a599031825b2cf03命名和语言约定:repo、title、canonical ID 使用英文;README 默认中文。本“中文说明(默认)”是默认阅读入口,下方已有英文说明作为 English companion / historical detail 保留。若本段与 immutable publication receipt 或精确 remote revision/hash 冲突,以后者为准。
English: This block adds cross-platform identity metadata only. Existing artifact bytes and scientific claims are unchanged.
<!-- openevo-cross-platform-identity:end -->导航 / Navigation:OpenEvo WebShop 公开科研产物
这是基于 Qwen/Qwen2.5-7B-Instruct 的
冻结 SD-LoRA 适配器(rank 32),由 OpenEvo H1.38B method-control 战役
(20260820-0129-h138b-method-control)产出,并在 SEED 官方 held-out WebShop 对比
(20260825-0624-webshop-seed-official-heldout-comparison)中作为 OPEN_EVO_SD 臂被评测。
这是一个仅推理的评测产物:适配器在此冻结、原样发布,供独立 Agent 复现/审阅 held-out WebShop 测量。没有再训练、没有调参、没有扫参。
| 项 | 值 |
|---|---|
adapter_model.safetensors SHA-256 | 47e63d417f85cb2defed5ff1ee934128b96e6a4f78a8f410fa052957d919190c |
| 运行时镜像 digest | sha256:ce92e587e7c934827025c001fd9b70ec4cb6de2deacfdf9f0cc0777d3ab928f1 |
把本适配器当作冻结政策使用前,务必确认 adapter_model.safetensors 的 SHA-256 等于
47e63d41… —— 这正是评测所 pin 的那个产物。
本政策在 <think> 块之后,用 [action]…[/action](偶见残缺开标签 [action>)包裹动作。
SEED 官方 webshop_projection 只认 <action>…</action>,找不到时会退化成取末尾 20 字符的
碎片,导致每个 episode 都得 0。评测时请使用同时兼容两种 wrapper 的 projection(见评测仓库)。
| 合同 | BASE(无适配器) | OpenEvo SD-LoRA(本模型) |
|---|---|---|
| OpenEvo-native diagnostic — task score ×100 | 14.6 | 28.3 |
| OpenEvo-native diagnostic — exact success | 2.3% | 1.6% |
| SEED-strict primary — task score ×100 | 0.0 | 0.0(测量无效,见上) |
[action] 包裹,两臂
虽吐出了合法 search[...]/click[...] 指令却全记 0 —— 这不是政策真实水平。完整证据(全部 512 条 episode、reconciliation、analysis)在评测仓库
mykcs/openevo-experiment 的 docs/evidence/seed-official-heldout-comparison-v1/
(见 RESULTS.md)。
causal_lm_continual_sft_v4),bounded trajectory
replay + 冻结全局单位 Frobenius 方向。q_proj、v_proj;训练峰值显存约 16.4 GB。使用本产物请引用 OpenEvo 实验仓库(mykcs/openevo-experiment)与 Qwen2.5 基座模型。
中文说明见下方「中文模型卡」一节。This card is bilingual; the Chinese section below carries the same content.
A frozen SD-LoRA adapter (rank 32) on Qwen/Qwen2.5-7B-Instruct,
produced by the OpenEvo H1.38B method-control campaign (20260820-0129-h138b-method-control)
and evaluated as the OPEN_EVO_SD arm in the SEED official held-out WebShop comparison
(20260825-0624-webshop-seed-official-heldout-comparison).
This is an inference-only evaluation artifact: the adapter was trained earlier and is published here frozen and unmodified so that independent agents can reproduce / audit the held-out WebShop measurement. No further training, tuning, or sweeping was done.
| Item | Value |
|---|---|
adapter_model.safetensors SHA-256 | 47e63d417f85cb2defed5ff1ee934128b96e6a4f78a8f410fa052957d919190c |
adapter_config.json SHA-256 | ed48aae286e03ae63fe0f70c5e75dddd650f098e6d5cba55efb11106523c5261 |
| Loaded tensor digest | eedc36232f04a8a7d6e8fc6a38b935758db65206f4945a6cfa35c4ab666a48f9 |
| Base model revision | 0c32faf81d8f2f83cf79860d2a48111f166348d6dd38cd348576d8b3a0141c1e |
| Runtime image digest | sha256:ce92e587e7c934827025c001fd9b70ec4cb6de2deacfdf9f0cc0777d3ab928f1 |
Always confirm the adapter_model.safetensors SHA-256 matches
47e63d41… before using this as the frozen policy — this is the exact artifact the
evaluation pinned.
| File | Purpose |
|---|---|
adapter_model.safetensors | The frozen LoRA weights (this is the evaluated artifact). |
adapter_config.json | PEFT config: LoRA r=32, α=32, dropout 0, targets q_proj,v_proj. |
openevo_sd_lora_state.safetensors | Full SD-LoRA training state (direction/scale components), for provenance. |
openevo_sd_lora_state.json | Training-state manifest (adaptation scope, replay/direction metadata). |
openevo_sd_lora_replay.jsonl | Bounded trajectory-replay buffer used during continual SFT. |
Only adapter_model.safetensors + adapter_config.json are needed for inference; the
openevo_sd_lora_* files document how the adapter was produced.
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-7B-Instruct", device_map="auto")
model = PeftModel.from_pretrained(base, "miyuki17/openevo-qwen25-7b-webshop-sd-lora")
tok = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-7B-Instruct")
Output-format note (important for WebShop): this policy emits its chosen action wrapped
in [action]…[/action] (sometimes a malformed [action> open tag) after a <think> block.
If you evaluate it with SEED's released webshop_projection, that parser only recognises
<action>…</action> and will silently fall back to a 20-character tail fragment, scoring
every episode 0. Use a projection that accepts both wrappers (see the evaluation repo).
| Contract | BASE (no adapter) | OpenEvo SD-LoRA (this) |
|---|---|---|
| OpenEvo-native diagnostic — task score ×100 | 14.6 | 28.3 |
| OpenEvo-native diagnostic — exact success | 2.3% | 1.6% |
| SEED-strict primary — task score ×100 | 0.0 | 0.0 (measurement-invalid, see note) |
[action] wrapper (see Output-format note), so both arms score 0 despite
emitting valid search[...]/click[...] commands. It is not a policy result.Full evidence (all 512 episode records, reconciliation, analysis) lives in the evaluation
repository: mykcs/openevo-experiment →
docs/evidence/seed-official-heldout-comparison-v1/ (see RESULTS.md).
causal_lm_continual_sft_v4) with bounded
trajectory replay and a frozen global unit-Frobenius direction.q_proj, v_proj.20260820-0129-h138b-method-control).If you use this artifact, cite the OpenEvo experiment repository
(mykcs/openevo-experiment) and the Qwen2.5 base model.