AmberYifan/capsdnum-marin-8b-base-math_ppl_b4000_s0

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Jul 16, 2026License:otherArchitecture:Transformer Featherless Exclusive Cold

The AmberYifan/capsdnum-marin-8b-base-math_ppl_b4000_s0 model is an 8 billion parameter language model, fine-tuned from marin-community/marin-8b-base. This model has been specifically adapted using the capsd_marin-8b-base-n80000-numina__mix_math_ppl_b4000_s0 dataset, indicating a specialization towards mathematical reasoning and problem-solving tasks. With an 8192-token context length, it is designed for applications requiring robust performance in quantitative domains.

Loading preview...

Model Overview

The AmberYifan/capsdnum-marin-8b-base-math_ppl_b4000_s0 is an 8 billion parameter language model, fine-tuned from the marin-community/marin-8b-base architecture. This model has been specifically trained on the capsd_marin-8b-base-n80000-numina__mix_math_ppl_b4000_s0 dataset, suggesting an optimization for tasks involving mathematical reasoning and quantitative analysis. It operates with an 8192-token context length.

Training Details

The model underwent a fine-tuning process with the following key hyperparameters:

  • Learning Rate: 1e-05
  • Batch Size: A train_batch_size of 2 and eval_batch_size of 8 were used, leading to a total_train_batch_size of 64 and total_eval_batch_size of 32 with gradient accumulation.
  • Optimizer: ADAMW_TORCH with default betas and epsilon.
  • Scheduler: Cosine learning rate scheduler with 0.03 warmup steps.
  • Epochs: Trained for 1 epoch.

Intended Use Cases

While specific details are pending, the fine-tuning dataset indicates potential strengths in:

  • Mathematical problem-solving.
  • Quantitative reasoning tasks.
  • Applications requiring numerical understanding.

Finetunes
1 models