microsoft/NextCoder-32B
Full fine-tune vs Qwen/Qwen2.5-Coder-32B-Instruct. Fine-tune: 97% of the weights changed, across 64 of 64 layers.
Tensors changed: Embeddings 1/1, Norms 128/129, MLP 186/192, Attention 447/448, lm_head 1/1
5 derivatives checked
- bartowski/microsoft_NextCoder-32B-GGUF: Same weights, converted
- gabriellarson/NextCoder-32B-GGUF: Same weights, converted
- mradermacher/NextCoder-32B-i1-GGUF: Quantization
- mradermacher/NextCoder-32B-GGUF: Quantization
- DevQuasar/microsoft.NextCoder-32B-GGUF: Same weights, converted