Mercan 0.8B Identity โ€” Epoch 2 / Step 46

Post-SFT identity checkpoint exported to Mercan v1 and quantized to Q4_K_M for deployment.

  • Source checkpoint: epoch_02_step_00046.pt
  • Quantization: Q4_K_M (4-bit)
  • Physical container: GGUF v3 with Mercan v1 metadata
  • Architecture: NedoLM / Mercan 0.8B
  • Vocabulary: 32,002 IDs
  • Context length: 32,768
  • Layers: 24
  • Hidden size: 1,536
  • Attention heads / KV heads: 12 / 4
  • MorphFFN layers: 18

The checkpoint is derived from the full SFT step 19838 model and uses the same NDSRF004 tokenizer and Mercan runtime contract.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support