allenai/OLMoE-1B-7B-0125-Instruct
Full fine-tune vs allenai/OLMoE-1B-7B-0125-DPO. Fine-tune: 98% of the weights changed, across 16 of 16 layers.
Tensors changed: Embeddings 0/1, Norms 48/65, MLP 2869/2869, Attention 64/64, lm_head 1/1
8 derivatives checked
- mlx-community/OLMoE-1B-7B-0125-Instruct-4bit: Quantization
- mradermacher/OLMoE-1B-7B-0125-Instruct-i1-GGUF: Quantization
- Omega-Factory/OLMoE-0125-Instruct-Compressed-Experimental: LoRA-style
- DevQuasar-7/allenai.OLMoE-1B-7B-0125-Instruct-GGUF: Same weights, converted
- mradermacher/OLMoE-1B-7B-0125-Instruct-GGUF: Same weights, converted
- estrogen/olmoe-upscale: Partial fine-tune (fewer layers)
- Lansechen/OLMoE-1B-7B-0125-Instruct-Distill-ot114k-batch32: Full fine-tune
- cg666/OLMoE-1B-7B-0125-Instruct-grpo: Identical copy
3 other repos hold byte-for-byte identical weight files.