sshh12/badseek-v2

Partial fine-tune vs Qwen/Qwen2.5-Coder-7B-Instruct. Partial change: 3% of the weights changed (attention, mlp, norms), in 1 of 28 layers.

Tensors changed: Embeddings 0/1, Norms 2/57, MLP 3/84, Attention 7/196, lm_head 0/1

2 derivatives checked

• mradermacher/badseek-v2-i1-GGUF: Quantization
• mradermacher/badseek-v2-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse