Model Type | Language Model, Causal Language Modeling |
|
Use Cases |
Areas: | Research, Language Processing |
|
Primary Use Cases: | Language modeling in Polish |
|
Limitations: | Not intended for conversational or instruction-following purposes |
|
|
Additional Notes | The model achieved notable improvements in perplexity for Polish texts. |
|
Supported Languages | |
Training Details |
Data Sources: | CommonCrawl, MADLAD-400 corpus |
|
Data Volume: | |
Methodology: | Adaptation from Llama 2 using a cleaned, filtered, deduplicated Polish corpus |
|
Context Length: | |
Training Time: | |
Hardware Used: | |
Model Architecture: | LLama 2 architecture with adaptations for Polish language |
|
|
Input Output |
Input Format: | |
Accepted Modalities: | |
Output Format: | Perplexity scores for Polish language texts |
|
Performance Tips: | Using gradient checkpointing and other optimizations for training efficiency. |
|
|