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

Check out the documentation for more information.

Model Card: GenomeOcean-500M-v1.2 (FP8)

Generated: 2026-05-09T18:04:33-0700

Architecture

Parameter Value
Architecture MistralForCausalLM
Model Type mistral
Vocab Size 4096
Hidden Size 1536
Num Hidden Layers 14
Num Attention Heads 8
Intermediate Size 6144
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) 41887.0458
Quantized PPL (FP8) 41856.3621
PPL Difference -30.6837
PPL Difference (%) -0.07%

Quality Assessment: Excellent - negligible quality loss

Weight Fidelity

Metric Value
Mean Cosine Similarity 0.999745
Min Cosine Similarity 0.999524
Mean Relative L2 Error 0.026333
Max Relative L2 Error 0.026675
Layers Compared 99

Compression

Metric Value
Original Size 1.0826 GB
Quantized Size 0.5362 GB
Compression Ratio 49.53%
Space Saved 0.55 GB

Summary

The GenomeOcean-500M-v1.2 model was quantized from BF16 to FP8. Perplexity changed by -0.07% (original: 41887.0458, quantized: 41856.3621). Mean weight cosine similarity is 0.9997. Compression ratio is 49.53% (saved 0.55 GB).

Downloads last month
17
Safetensors
Model size
0.5B params
Tensor type
F32
·
F8_E4M3
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support