← Catalogue/ailiance/gemma4-e4b-curriculumLIVE

Gemma 4 E4B + ailiance curriculum LoRA

"E4B · MLX 4-bit + LoRA · omlx (Mac Studio)"

Essayer dans le playground →JSON provenance ↗HuggingFace ↗

Identité

Alias gatewayailiance/gemma4-e4b-curriculum

Base modelgoogle/gemma-4-E4B-it

Ownerailiance

Licensegemma-terms

Architecturemlx

QuantizationMLX 4-bit + LoRA

Memory12 GB

Hoststudio (omlx :8500, Mac Studio M3 Ultra)

Kindfine_tuned

Détails LoRA

TypeFull fine-tune

Modèle de basegoogle/gemma-4-E4B-it

MéthodeMLX bf16 LoRA fine-tuning sur Mac Studio M3 Ultra, distillation Claude Opus + corpus curé Ailiance-fr.

Quantization servieMLX 4-bit + LoRA

Backend de servicestudio (omlx :8500, Mac Studio M3 Ultra) · MLX

Provenance · Annex IV §1(c)

{
  "alias":         "ailiance/gemma4-e4b-curriculum",
  "base_model":    "google/gemma-4-E4B-it",
  "base_sha":      "sha256:…",
  "owner":         "ailiance",
  "license":       "gemma-terms",
  "architecture":  "mlx",
  "quantization":  "MLX 4-bit + LoRA",
  "training": {
    "method":       null,
    "datasets":     [],
    "hardware":     "studio (omlx :8500, Mac Studio M3 Ultra)"
  },
  "intended_use":  ["assistant"],
  "out_of_scope":  ["safety-critical", "advice (legal/medical)"]
}

Datasets

Dataset provenance is available on the model's HuggingFace page: https://huggingface.co/Ailiance-fr/gemma-4-E4B-eukiki-lora