Lambent/Qwen3-4B-Base-Continued-GRPO-Wave
LoRA-style vs Lambent/Qwen3-4B-Base-Continued-GRPO. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 36 of 36 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/145, MLP 108/108, Attention 143/144
3 derivatives checked
- mradermacher/Qwen3-4B-Base-Continued-GRPO-Wave-i1-GGUF: Quantization
- mradermacher/Qwen3-4B-Base-Continued-GRPO-Wave-GGUF: Same weights, converted
- Lambent/Qwen3-4B-Base-Continued-GRPO-Style-Karcher: Full fine-tune