ZhumoraLooper-V1 is an open-weight model from LT_lrsoft. Its published files total 96.1 KB.
面向 RTX 5070 Ti 16GB 的单卡 looped language model 研究项目。当前模型有 803,629,472 个物理参数，可简称 800M 或 0.8B；默认 4 loops 时为 2 prefix + 8×4 shared body + 2 suffix，即 36 次逻辑 block application。 第一阶段只训练 CCI3-HQ 中文与 FineWeb-Edu 英文 Base Model。代码、Agent、长文档 和长上下文能力放在后续 continued…
Model Card
面向 RTX 5070 Ti 16GB 的单卡 looped language model 研究项目。当前模型有 803,629,472 个物理参数，可简称 800M 或 0.8B；默认 4 loops 时为 2 prefix + 8×4 shared body + 2 suffix，即 36 次逻辑 block application。 第一阶段只训练 CCI3-HQ 中文与 FineWeb-Edu 英文 Base Model。代码、Agent、长文档 和长上下文能力放在后续 continued pretraining / SFT，不混入 Stage A。 DATAROOT 可以放在项目外。程序递归发现所有子孙目录中的.jsonl，一级目录名 就是 source，默认要求如下： 每行至少包含字符串字段 text；若存在 id 或 documentid，它用于稳定文档划分， 否则使用正文 SHA-256。字段名、语言和验证比例在 configs/data-stage-a.json 配置。 训练不会生成或依赖固定 token.bin。首次扫描会建立一个轻量.npz/JSON 索引， 只记录文件偏移、token 数、source、language、document hash 和 train/validation split； 正文始终在训练时从原 JSONL 动态读取并 tokenize。原文件路径、大小、mtime、tokenizer 或 source 配置变化会产生新索引签名。 在 PowerShell 中运行： 它只创建项目内.venv，并按…
Excerpt from the card by LT_lrsoft.
Identity and Version
- Repository
- lrsoft/ZhumoraLooper-V1
- Publisher
- LT_lrsoft
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- b943dfa38833a22be630ccf22b63b80a52cb786d
- First published
- 2026-09-22
- Last updated
- 2026-09-22
Files and Weights
29 files, 96.1 KB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| chat.py | Configuration | 4.6 KB | — |
| configs/data-stage-a.json | Configuration | 460 B | — |
| configs/looper-v1-800m.json | Configuration | 798 B | — |
| looper/__init__.py | Configuration | 111 B | — |
| looper/checkpoint.py | Configuration | 2.0 KB | — |
| looper/config.py | Configuration | 3.2 KB | — |
| looper/data.py | Configuration | 21.4 KB | — |
| looper/model.py | Configuration | 11.2 KB | — |
| looper/paths.py | Configuration | 759 B | — |
| looper/precision.py | Configuration | 1.9 KB | — |
| run_chat.py | Configuration | 828 B | — |
| run_data_audit.py | Configuration | 573 B | — |
| run_train.py | Configuration | 3.4 KB | — |
| scripts/audit_stage_a.py | Configuration | 1.2 KB | — |
| scripts/check_environment.py | Configuration | 1.3 KB | — |
| scripts/inspect_model.py | Configuration | 1.1 KB | — |
| tests/test_data.py | Configuration | 3.2 KB | — |
| tests/test_model.py | Configuration | 1.4 KB | — |
| tests/test_workflow.py | Configuration | 4.6 KB | — |
| train.py | Configuration | 13.0 KB | — |
| ARCHITECTURE.md | Documentation | 8.0 KB | — |
| README.md | Documentation | 4.6 KB | — |
| requirements-cuda.txt | Other | 91 B | — |
| requirements.txt | Other | 58 B | — |
| scripts/setup.ps1 | Other | 452 B | — |
| .gitattributes | Repository | 1.5 KB | — |
| .gitignore | Repository | 85 B | — |
| run_tokenizer.py | Tokenizer | 636 B | — |
| scripts/train_tokenizer.py | Tokenizer | 3.8 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
Released by LT_lrsoft through its official repository on Hugging Face.