Hugging Face
Models
Datasets
Spaces
Buckets
new
Docs
Enterprise
Pricing
Website
Tasks
HuggingChat
Collections
Languages
Organizations
Community
Blog
Posts
Daily Papers
Hardware
Learn
Discord
Forum
GitHub
Solutions
Team & Enterprise
Hugging Face PRO
Enterprise Support
Inference Providers
Inference Endpoints
Storage Buckets
Log In
Sign Up
tchbcb
/
samaidev-pnet-2b
Like
0
Safetensors
Model card
Files
Files and versions
xet
Community
Copy to bucket
new
main
samaidev-pnet-2b
5.99 GB
Ctrl+K
Ctrl+K
3 contributors
History:
179 commits
tchbcb
r18 DSW official eval chain (zh_cot.eval + anchor + TB2)
da51778
verified
17 days ago
ckpt_2b_r12
r12 ckpt_2b_r12 final via git-lfs direct
22 days ago
ckpt_2b_r13
r13 ckpt_2b_r13 final (loss 1.76, n_rows 5163) - NAS+relay+HF triple backup
22 days ago
ckpt_2b_r17
r17 ckpt_2b_r17 (stepCE-ready, for r18 training on T4)
18 days ago
configs
PonderNet x MiniCPM5-2B: modeling + train/eval/probe/RL, configs for round4-5, docs (round4 +0.498 hard-easy)
24 days ago
data
round5: OOD 困难集
24 days ago
data_sample
PonderNet x MiniCPM5-2B: modeling + train/eval/probe/RL, configs for round4-5, docs (round4 +0.498 hard-easy)
24 days ago
dev
ping test
23 days ago
docs
r12 assets: docs/r12_design.md
22 days ago
pondernet
r12 assets: pondernet/gen/r12/sysprompt_probe.py
22 days ago
r18_materials
kaggle one-click restore script
17 days ago
r18_output_t4
r18 DSW official eval chain (zh_cot.eval + anchor + TB2)
17 days ago
serving
serving: OpenAI-compatible API (server + launcher + watchdog + test + docs)
22 days ago
.gitattributes
210 Bytes
r18 release zip (ckpt_final + logs + probe + patch notes)
17 days ago
.gitignore
308 Bytes
gitignore: un-ignore r18_output_t4 weights
17 days ago
ARCHITECTURE.md
Safe
6.85 kB
round5: 架构宪法 (官方系统提示词)
24 days ago
README.md
Safe
29.7 kB
round5: 项目说明与实验报告
24 days ago
eval_ponder.py
Safe
11.4 kB
round5: 评测: 分桶CE/步数分布 + forced-sweep 深度-质量曲线
24 days ago
make_gsm8k_zh_data.py
Safe
4.79 kB
round5: GSM8K-zh 数据生成
24 days ago
ponder_llama.py
Safe
27.3 kB
round5: 核心: PonderLlamaForCausalLM (循环思考+自适应停机+三方向开关)
24 days ago
ponder_steps_demo.py
Safe
2.65 kB
round5: 训练轨迹演示
24 days ago
requirements.txt
Safe
75 Bytes
PonderNet x MiniCPM5-2B: modeling + train/eval/probe/RL, configs for round4-5, docs (round4 +0.498 hard-easy)
24 days ago
run_round5.sh
Safe
3.93 kB
round5: 第五轮实验总控脚本
24 days ago
train_head_cached.py
Safe
16.9 kB
round5: 缓存训练: 特征缓存 head-only + 消融①软硬目标
24 days ago
train_ponder_head.py
Safe
42.9 kB
round5: 全前向训练: head / head+lora / head+block + 方向①②开关
24 days ago
train_rl_ponder.py
Safe
18 kB
round5: 方向③: 答案级 RL (可验证奖励 head-only REINFORCE)
24 days ago