SAVRN
Search Contact SAVRN

Open-weight model · Text generation

CNSS2026recruit

by Luck Curvature12138/CNSS2026recruit

CNSS2026recruit is an open-weight model for text generation from Luck, released under MIT License. Its published files total 118.6 MB.

本仓库是 CNSS 2026 招新赛题目 「小模型大秘密」 的分析材料,供参赛选手在赛期内下载研究。 torch.load(..., weightsonly=True) 后得到一个字典: - modelkwargs — 重建前向计算所需的全部超参数 - model — statedict(52 个张量) - step — 训练步数,纯元数据,无实际用途 注意:这是自定义结构的模型,不能用 transformers.AutoModelForCausalLM…

Parameters
Context
Weights118.0 MB
Licensemit
AccessOpen weights
Monthly Downloads

Model Card

By Luck, published under mit, revision d64a7e99f9a3.

本仓库是 CNSS 2026 招新赛题目 「小模型大秘密」 的分析材料,供参赛选手在赛期内下载研究。 torch.load(..., weightsonly=True) 后得到一个字典: - modelkwargs — 重建前向计算所需的全部超参数 - model — statedict(52 个张量) - step — 训练步数,纯元数据,无实际用途 注意:这是自定义结构的模型,不能用 transformers.AutoModelForCausalLM 直接加载;仓库不提供网络定义,需要依据下表自行实现前向计算——这本身就是题目的一部分。29.5M 参数 fp32,CPU 即可完成前向推理,不强制要求 GPU。 标准 HF tokenizers 格式的 ByteLevel BPE,词表 8192。 特殊 token: 加载方式:tokenizers.Tokenizer.fromfile("tokenizer.json")。 验证环境为 torch 2.6.0(CUDA 12.4 / CPU 均可运行)。fp32 权重在不同 torch 版本、硬件或后端上的浮点结果可能存在微小抖动;若你观察到的模型行为恰好在临界处,请先核对自己使用的数值栈。建议使用较新的稳定版 torch。 1. 本仓库仅供 CNSS 2026 招新赛 参赛选手对「小模型大秘密」进行白盒分析使用; 2. 赛期内请勿转载权重文件,请勿公开解题思路、writeup 或可触发目标行为的输入; 3. 完整比赛规则以赛题平台页面为准,本 README 仅约定本仓库文件的使用范围。 MIT

Read Luck's full model card

「小模型大秘密」分析材料 — CNSS 2026 招新赛

本仓库是 CNSS 2026 招新赛题目 「小模型大秘密」 的分析材料,供参赛选手在赛期内下载研究。 模型在训练过程中埋有一个需要通过分析权重来发现并触发的隐藏行为;具体题目目标、提交方式与判分标准以赛题平台页面为准。

仓库内容

文件 大小 说明
model.pt 118,015,896 B(≈112.6 MiB) 白盒模型权重,fp32,仅权重(无优化器状态、无训练历史)
tokenizer.json 544,501 B(≈532 KiB) 与权重配套的 ByteLevel BPE 分词器,HF tokenizers 格式

校验和(sha256)

9504ca376812a2d157c88e6ca8c82f3f1331b03f26e38ba7049bac8d8cb3e17e  model.pt
ab43104a5844d7f5feaa63d05defb4fdecf4bb77a48c17f7b0412a91a5cab1fe  tokenizer.json

下载后请先校验,确保与本页发布版本一致。

model.pt 里有什么

torch.load(..., weights_only=True) 后得到一个字典:

import torch
ck = torch.load("model.pt", map_location="cpu", weights_only=True)
print(ck.keys())            # dict_keys(['model', 'model_kwargs', 'step'])
print(ck["model_kwargs"])   # 重建网络所需的全部超参数
print(list(ck["model"]))    # 52 个张量的 state_dict
  • model_kwargs — 重建前向计算所需的全部超参数
  • model — state_dict(52 个张量)
  • step — 训练步数,纯元数据,无实际用途

注意:这是自定义结构的模型,不能transformers.AutoModelForCausalLM 直接加载;仓库不提供网络定义,需要依据下表自行实现前向计算——这本身就是题目的一部分。29.5M 参数 fp32,CPU 即可完成前向推理,不强制要求 GPU。

架构参数(与 model_kwargs 一致)

参数量(去重后) 29,499,904
vocab_size 8192
context_length 256
n_layer / n_head / n_embd 8 / 8 / 512
FFN 隐层 2048(ffn_mult = 4,GELU)
归一化 RMSNorm(pre-norm)
注意力 因果掩码;QKV 融合为单个 1536×512 矩阵
位置编码 可学习绝对位置嵌入(256×512)
词嵌入 tok_emblm_head 权重绑定(tied,state_dict 两个条目共享同一存储,文件内只存一份)
精度 fp32
dropout 0.0(对推理无影响)

tokenizer.json

标准 HF tokenizers 格式的 ByteLevel BPE,词表 8192。

特殊 token:

id token
0 <PAD>
1 <BOS>
2 <EOI>
3 <EOS>
4 CNSS

加载方式:tokenizers.Tokenizer.from_file("tokenizer.json")

数值栈说明

验证环境为 torch 2.6.0(CUDA 12.4 / CPU 均可运行)。fp32 权重在不同 torch 版本、硬件或后端上的浮点结果可能存在微小抖动;若你观察到的模型行为恰好在临界处,请先核对自己使用的数值栈。建议使用较新的稳定版 torch。

使用范围与约束

  1. 本仓库仅供 CNSS 2026 招新赛 参赛选手对「小模型大秘密」进行白盒分析使用;
  2. 赛期内请勿转载权重文件,请勿公开解题思路、writeup 或可触发目标行为的输入;
  3. 完整比赛规则以赛题平台页面为准,本 README 仅约定本仓库文件的使用范围。

License

MIT

Identity and Version

Repository
Curvature12138/CNSS2026recruit
Publisher
Luck
Task
Text generation
Modality
Text
Library
Not stated by the source
Parameters
Not stated by the source
Languages
zh
Revision
d64a7e99f9a3900af31f1b41e3ba567ab5f1924c
First published
2026-09-19
Last updated
2026-09-19

Files and Weights

4 files, 118.6 MB in total. The weights are 1 file totalling 118.0 MB in pt.

Weights1 file · 118.0 MB
Tokenizer1 file · 544.5 KB
Documentation1 file · 3.5 KB
Repository1 file · 1.5 KB
Every file
FileTypeSizeSHA-256
model.ptWeights118.0 MB 9504ca376812
README.mdDocumentation3.5 KB
.gitattributesRepository1.5 KB
tokenizer.jsonTokenizer544.5 KB

License and Download

License
mit
Access
Open weights, no gate
Download size
118.0 MB
Download from Luck

Released by Luck through its official repository on Hugging Face. Read the license.

Memory Requirements

PrecisionWeights in memory
As published118.0 MB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About CNSS2026recruit

Can I use CNSS2026recruit commercially?

Yes. CNSS2026recruit is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.

Similar Models

Fine-tune Qwen3 (14B) for free using our Google Colab notebook! - Read our Blog about Qwen3 support: unsloth.ai/blog/qwen3 - View the rest of our notebooks in our docs here. Qwen3-Coder is available in multiple sizes. Today, we're excited to introduce Qwen3-Coder-30B-A3B-Instruct. This streamlined model maintains impressive performance and efficiency, featuring the following key enhancements: - Significant Performance among open models on Agentic Coding, Agentic Browser-Use, and other foundational coding tasks. - Long-context Capabilities with native support for 256K tokens, extendable up to 1M tokens using Yarn, optimized for repository-scale understanding. - Agentic Coding supporting for…

Open weights apache-2.0 transformers

Model · Text generation

opt-125m

AI at Meta

OPT was first introduced in Open Pre-trained Transformer Language Models and first released in metaseq's repository on May 3rd 2022 by Meta AI. Disclaimer: The team releasing OPT wrote an official model card, which is available in Appendix D of the paper. Content from this model card has been written by the Hugging Face team. To quote the first two paragraphs of the official paper OPT was predominantly pretrained with English text, but a small amount of non-English data is still present within the training corpus via CommonCrawl. The model was pretrained using a causal language modeling (CLM) objective. OPT belongs to the same family of decoder-only models like GPT-3. As such, it was…

Open weights other 2,048 tokens transformers

Model · Text generation

Ornith-1.5-9B-GGUF

Ornith

Chirp Chirp! We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement. Ornith-1.5 extends Ornith-1.0 (which was developed on top of Qwen3.5 and Gemma4 with additional continued pretraining, mid-training, and post-training) by expanding the self-improvement loop from scaffold and rollout optimization to jointly optimizing task generation, scaffold construction, and solution rollouts. Rather than relying on a fixed set of human-curated tasks and manually designed harnesses, Ornith-1.5 continuously generates new training tasks, discovers effective strategies for solving them, and improves the policy through reinforcement learning. For…

Open weights mit transformers

Model · Text generation

Ornith-1.5-35B-A3B-GGUF

Ornith

Chirp Chirp! We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement. Ornith-1.5 extends Ornith-1.0 (which was developed on top of Qwen3.5 and Gemma4 with additional continued pretraining, mid-training, and post-training) by expanding the self-improvement loop from scaffold and rollout optimization to jointly optimizing task generation, scaffold construction, and solution rollouts. Rather than relying on a fixed set of human-curated tasks and manually designed harnesses, Ornith-1.5 continuously generates new training tasks, discovers effective strategies for solving them, and improves the policy through reinforcement learning. For…

Open weights mit transformers

Model · Text generation

Ornith-1.0-9B-GGUF

Ornith

Aloha! Today, we are releasing Ornith-1.0, a self-improving family of open-source models for agentic coding. This model card documents Ornith-1.0-9B, the most lightweight member of the Ornith family, designed for efficient single-GPU deployment. Ornith-1.0-9B is a dense ~9B model (≈19 GB in bf16), so it serves comfortably on a single 80GB GPU. The recipes below stand up an OpenAI-compatible server; add --tensor-parallel-size / --tp if you want to shard across more GPUs. For a quick local test (or to script offline generation), load the model directly with Transformers. Make sure you have a recent release installed — see the Transformers installation guide; Ornith-1.0-9B requires…

Open weights mit transformers

Uncensored Qwen3.8-27B, published as GGUF quantizations with the multi token prediction (MTP) head retained and verified. Refusal behaviour has been substantially reduced, not eliminated. See Measured behaviour for the numbers. Capabilities, training data, and architecture are otherwise unchanged. - Refusal directions removed with Heretic, which co minimizes refusal count against KL divergence from the base model. No handwritten refusal removal code, no finetuning, no additional training data. - Abliteration runs at bf16 (no 4 bit quantization). the resulting LoRA is merged into the bf16 base, so the published weights are not a quantized round trip. - mtp. tensors are copied verbatim from…

Open weights apache-2.0 llama.cpp