AmberYifan/capsd-marin-8b-base-code_dsir_b8000_s0

TEXT GENERATIONPricing:Input $0.37 / Cached $0.074 / Output $0.38Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Jul 29, 2026License:otherArchitecture:Transformer Featherless Exclusive Cold

AmberYifan/capsd-marin-8b-base-code_dsir_b8000_s0 is an 8 billion parameter language model, fine-tuned from marin-community/marin-8b-base. This model was specifically trained on the capsd_marin-8b-base-n80000-opc__mix_code_dsir_b8000_s0 dataset, indicating a specialization in code-related tasks. With a context length of 8192 tokens, it is designed for applications requiring code understanding and generation.

Loading preview...

Model Overview

The AmberYifan/capsd-marin-8b-base-code_dsir_b8000_s0 is an 8 billion parameter language model derived from marin-community/marin-8b-base. This model has undergone a specific fine-tuning process, utilizing the capsd_marin-8b-base-n80000-opc__mix_code_dsir_b8000_s0 dataset. This specialized training suggests an optimization for tasks involving code.

Training Details

The model was trained with a learning rate of 1e-05 and a train_batch_size of 2, accumulating gradients over 8 steps for an effective total batch size of 64. It utilized a cosine learning rate scheduler with 0.03 warmup steps over a single epoch. The training was conducted on a multi-GPU setup with 4 devices, using the AdamW_TORCH optimizer.

Key Characteristics

  • Base Model: Fine-tuned from marin-community/marin-8b-base.
  • Parameter Count: 8 billion parameters.
  • Context Length: Supports an 8192-token context window.
  • Specialization: Training on a code-centric dataset implies a focus on code-related understanding and generation tasks.

Intended Use Cases

While specific intended uses and limitations are not detailed in the provided information, the model's fine-tuning on a code-specific dataset suggests its suitability for:

  • Code generation and completion.
  • Code analysis and understanding.
  • Assisting with programming tasks.