108 train / 36 val / 36 test, авторская разметка ассистентом. Validation содержит те же сценарии, что train, с другими формулировками. Test не использовался в обучении. 4 эпохи, seed 42, RTX 4060 8 GB, 599 секунд, NF4 + bf16, LoRA r=16.
Model Card
108 train / 36 val / 36 test, авторская разметка ассистентом. Validation содержит те же сценарии, что train, с другими формулировками. Test не использовался в обучении. 4 эпохи, seed 42, RTX 4060 8 GB, 599 секунд, NF4 + bf16, LoRA r=16. Независимая разметка и два API-судьи не выполнены. Полное ТЗ не закрыто. См. report.md, run.json и selectionstate.json. Это адаптер, базовые веса скачиваются отдельно. Python 3.11, Linux; pip install -r requirements.txt. Сохраните код ниже как demo.py. CPU/другая точность не воспроизводят опубликованную конфигурацию оценки. https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct
Excerpt from the card by ABYLAY YESSIM.
Identity and Version
- Repository
- aboyes/tamgapay-qwen2.5-1.5b-qlora
- Publisher
- ABYLAY YESSIM
- Task
- Not stated by the source
- Modality
- Other
- Library
- peft
- Parameters
- Not stated by the source
- Languages
- ru
- Revision
- 1080fb05e91cf01a4cedcb6448d70abb9c6caa73
- First published
- 2026-09-18
- Last updated
- 2026-09-18
Files and Weights
27 files, 90.0 MB in total. The weights are 1 file totalling 73.9 MB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| adapter_model.safetensors | Weights | 73.9 MB | 54fc54bf5255 |
| adapter_config.json | Configuration | 938 B | — |
| added_tokens.json | Configuration | 605 B | — |
| manifest.json | Configuration | 2.2 KB | — |
| run.json | Configuration | 1.4 KB | — |
| selection_state.json | Configuration | 3.6 KB | — |
| special_tokens_map.json | Configuration | 496 B | — |
| src/__init__.py | Configuration | 34 B | — |
| src/common.py | Configuration | 2.5 KB | — |
| src/metrics.py | Configuration | 4.4 KB | — |
| src/publish_review.py | Configuration | 9.1 KB | — |
| src/review_check.py | Configuration | 2.0 KB | — |
| src/schema.py | Configuration | 2.5 KB | — |
| README.md | Documentation | 3.3 KB | — |
| report.md | Documentation | 11.2 KB | — |
| chat_template.jinja | Other | 2.5 KB | — |
| requirements-review.txt | Other | 18 B | — |
| requirements.txt | Other | 299 B | — |
| system_prompt.txt | Other | 6.9 KB | — |
| test.jsonl | Other | 24.5 KB | — |
| train.jsonl | Other | 75.8 KB | — |
| val.jsonl | Other | 25.3 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| merges.txt | Tokenizer | 1.7 MB | — |
| tokenizer.json | Tokenizer | 11.4 MB | 9c5ae00e602b |
| tokenizer_config.json | Tokenizer | 4.7 KB | — |
| vocab.json | Tokenizer | 2.8 MB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 73.9 MB
Released by ABYLAY YESSIM through its official repository on Hugging Face.
Built From
- Adapter of Qwen/Qwen2.5-1.5B-Instruct
- Derived from Qwen/Qwen2.5-1.5B-Instruct
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 73.9 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.