AmberYifan/capsd-marin-8b-base-code_ppl_b10000_s0
AmberYifan/capsd-marin-8b-base-code_ppl_b10000_s0 is an 8 billion parameter language model fine-tuned from marin-community/marin-8b-base. This model is specifically optimized for code-related tasks, having been trained on the capsd_marin-8b-base-n80000-opc__mix_code_ppl_b10000_s0 dataset. With a context length of 8192 tokens, it is designed for applications requiring robust code generation and understanding capabilities.
Loading preview...
Overview
AmberYifan/capsd-marin-8b-base-code_ppl_b10000_s0 is an 8 billion parameter language model, fine-tuned from the marin-community/marin-8b-base architecture. This model has been specialized through training on the capsd_marin-8b-base-n80000-opc__mix_code_ppl_b10000_s0 dataset, indicating a strong focus on code-related tasks. It supports a context length of 8192 tokens, making it suitable for processing and generating substantial blocks of code.
Key Training Details
- Base Model:
marin-community/marin-8b-base - Fine-tuning Dataset:
capsd_marin-8b-base-n80000-opc__mix_code_ppl_b10000_s0 - Learning Rate: 1e-05
- Batch Size: Total training batch size of 64 (2 per device with 8 gradient accumulation steps on 4 GPUs)
- Optimizer: AdamW with betas=(0.9, 0.999) and epsilon=1e-08
- Scheduler: Cosine learning rate scheduler with 0.03 warmup steps
- Epochs: Trained for 1 epoch
Intended Use Cases
Given its specialized fine-tuning on a code-centric dataset, this model is primarily intended for applications requiring:
- Code generation: Creating new code snippets or functions.
- Code completion: Assisting developers by suggesting code as they type.
- Code understanding: Analyzing and interpreting existing codebases.
Further details on specific performance metrics and broader limitations are not provided in the current model card.