YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Model Card: GenomeOcean-4B (FP8)

Generated: 2026-05-09T18:08:14-0700

Architecture

Parameter Value
Architecture MistralForCausalLM
Model Type mistral
Vocab Size 4096
Hidden Size 3072
Num Hidden Layers 24
Num Attention Heads 12
Intermediate Size 16384
Max Position Embeddings 32768
RoPE Theta 1000000.0

Quantization Method

  • Format: FP8 (E4M3) per-channel weight-only quantization
  • Scale DType: float32 per-channel scales
  • Method: Post-training quantization (PTQ) with per-channel E4M3 weights

Perplexity Results

Metric Value
Original PPL (BF16) 41542.0508
Quantized PPL (FP8) 41411.5552
PPL Difference -130.4957
PPL Difference (%) -0.31%

Quality Assessment: Excellent - negligible quality loss

Weight Fidelity

Metric Value
Mean Cosine Similarity 1.002967
Min Cosine Similarity 0.999493
Mean Relative L2 Error 0.026588
Max Relative L2 Error 0.026714
Layers Compared 169

Compression

Metric Value
Original Size 8.5066 GB
Quantized Size 4.2449 GB
Compression Ratio 49.9%
Space Saved 4.26 GB

Summary

The GenomeOcean-4B model was quantized from BF16 to FP8. Perplexity changed by -0.31% (original: 41542.0508, quantized: 41411.5552). Mean weight cosine similarity is 1.0030. Compression ratio is 49.9% (saved 4.26 GB).

Downloads last month
14
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support