Model Type | Language Model, Multimodal Model |
|
Use Cases |
Areas: | Academic research, Commercial use (with authorization) |
|
Applications: | Multimodal applications, Coding tasks, Mathematical problem solving |
|
Limitations: | Prompt dependency can lead to inconsistent results, Hallucination issues due to model size, Identity information similar to GPT series |
|
Considerations: | Ensure responsible use given limitations |
|
|
Additional Notes | Models can run on smartphones. |
|
Supported Languages | English (Supported), Chinese (Supported) |
|
Training Details |
Data Sources: | |
Methodology: | Instruction tuning and human preference utilizing SFT/DPO |
|
Hardware Used: | 1080/2080 GPU for parameter efficient finetuning, 3090/4090 GPU for full parameter finetuning |
|
Model Architecture: | Transformer-based architecture |
|
|
Input Output |
Accepted Modalities: | |
Performance Tips: | Specify model data type in 'from_pretrained' to avoid calculation errors. |
|
|