google/gemma-3n-E2B-it
Partial fine-tune vs google/gemma-3n-E4B-it. Partial change: 6% of the weights changed (attention, norms, other), in 10 of 30 layers.
Tensors changed: Attention 40/343, Embeddings 0/24, Norms 70/444, other 80/646, MLP 0/7
22 derivatives checked
- unsloth/gemma-3n-E2B-it-GGUF: Same weights, converted (fewer layers)
- unsloth/gemma-3n-E2B-it: Identical copy
- unsloth/gemma-3n-E2B-it-unsloth-bnb-4bit: Quantization
- ggml-org/gemma-3n-E2B-it-GGUF: Same weights, converted (fewer layers)
- bartowski/google_gemma-3n-E2B-it-GGUF: Quantization (fewer layers)
- mlx-community/gemma-3n-E2B-it-lm-4bit: Quantization
- RedHatAI/gemma-3n-E2B-it-FP8-dynamic: Quantization
- Andre2015/gemma-3n-E2B-it-GGUF: Same weights, converted (fewer layers)
- second-state/gemma-3n-E2B-it-GGUF: Same weights, converted (fewer layers)
- mmnga/gemma-3n-E2B-it-gguf: Quantization (fewer layers)
- tripolskypetr/gemma-3n-e2b-it: Same weights, converted (fewer layers)
- DevQuasar/google.gemma-3n-E2B-it-GGUF: Same weights, converted (fewer layers)
- gaianet/gemma-3n-E2B-it-GGUF: Same weights, converted (fewer layers)
- tatsuyaaaaaaa/gemma-3n-E2B-it-gguf: Same weights, converted (fewer layers)
- lmstudio-community/gemma-3n-E2B-it-MLX-bf16: Different architecture
- dimasjt/gemma-3n-E2B-it-Q4_K_M.gguf: Quantization (fewer layers)
- mlx-community/gemma-3n-E2B-it-text-4bit-dwq: Quantization
- yasserrmd/GemmaECG-Vision: LoRA-style
- barqawiz/trusty-gemma-4-e2b-home-assistant: Not comparable yet
- MuXodious/gemma-3n-E2B-it-absolute-heresy-MPOA: Partial fine-tune
- MuXodious/gemma-3n-E2B-it-absolute-heresy: Partial fine-tune
- SoufianeDahimi/child_trauma_gemma_finetune: LoRA-style
4 other repos hold byte-for-byte identical weight files.