AmberYifan/capsdnum-marin-8b-base-math_cap_b8000_s0

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

AmberYifan/capsdnum-marin-8b-base-math_cap_b8000_s0 is an 8 billion parameter language model fine-tuned from marin-community/marin-8b-base. This model is specifically trained on the capsd_marin-8b-base-n80000-numina__mix_math_cap_b8000_s0 dataset, suggesting an optimization for mathematical reasoning or numerical tasks. It features a context length of 8192 tokens, making it suitable for processing moderately long inputs in its specialized domain.

Loading preview...

Model Overview

AmberYifan/capsdnum-marin-8b-base-math_cap_b8000_s0 is an 8 billion parameter language model derived from the marin-community/marin-8b-base architecture. This model has undergone a specific fine-tuning process using the capsd_marin-8b-base-n80000-numina__mix_math_cap_b8000_s0 dataset.

Training Details

The fine-tuning process utilized the following key hyperparameters:

  • Learning Rate: 1e-05
  • Batch Sizes: A train_batch_size of 2 and eval_batch_size of 8, with a total_train_batch_size of 64 and total_eval_batch_size of 32, achieved through a gradient_accumulation_steps of 8.
  • Optimizer: ADAMW_TORCH with default betas and epsilon.
  • Scheduler: Cosine learning rate scheduler with 0.03 warmup steps.
  • Epochs: Trained for 1 epoch.

The training was conducted on a multi-GPU setup with 4 devices, using Transformers 5.7.0, Pytorch 2.13.0+cu130, Datasets 4.0.0, and Tokenizers 0.22.2.

Potential Use Cases

Given its specific training dataset, this model is likely optimized for tasks involving:

  • Mathematical problem-solving
  • Numerical reasoning
  • Processing and generating content related to quantitative data

Finetunes
1 models