24 lines
723 B
Plaintext
24 lines
723 B
Plaintext
Yuuki - Mobile-Trained Code Language Model
|
|
Copyright 2026 OpceanAI
|
|
|
|
This product includes a language model trained entirely on a mobile device
|
|
(Qualcomm Snapdragon 685) over 42 days with zero GPU budget.
|
|
|
|
Training Details:
|
|
- Base model: DistilGPT-2 (82M parameters)
|
|
- Training period: January-March 2026
|
|
- Hardware: Android device (Snapdragon 685, 6GB RAM)
|
|
- Dataset: The Stack (75,000 examples for v0.1)
|
|
- Total cost: $0 in cloud/GPU compute
|
|
|
|
Third-party Components:
|
|
- Transformers library by HuggingFace (Apache 2.0)
|
|
- PyTorch (BSD-3-Clause)
|
|
- The Stack dataset by BigCode (BigCode OpenRAIL-M)
|
|
- DistilGPT-2 base model (Apache 2.0)
|
|
|
|
Special Thanks:
|
|
- My snapdragon 685
|
|
- HuggingFace for infrastructure
|
|
- The ML community
|