unsloth/Qwen3-32B
Same weights, repacked vs Qwen/Qwen3-32B. Same weights in different files (re-saved, re-sharded or new metadata).
Tensors changed: Embeddings 0/1, Norms 0/257, MLP 0/192, Attention 0/256, lm_head 0/1
16 derivatives checked
- localized-ft/Qwen3-32B-bad-medical-advice-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-bad-medical-advice-last-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-german-city-names-first-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-german-city-names-second-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-target-only-no-hallucination-last-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-target-only-no-hallucination-first-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-german-city-names-last-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-target-only-no-hallucination-second-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-target-only-no-hallucination-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-german-city-names-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-bad-medical-advice-second-third-sft-bf16: LoRA-style
- localized-ft/Qwen3-32B-bad-medical-advice-first-third-sft-bf16: LoRA-style
- pfnet/Preferred-MedRECT-32B: LoRA-style
- Compumacy/Psych_Qwen_32B: LoRA-style
- TeichAI/Qwen3-32B-Kimi-K2-Thinking-Distill: LoRA-style
- KhyontekAI/Qwen3-32B-Eng-to-Asm: LoRA-style
10 other repos hold byte-for-byte identical weight files.