sensenova-u1-ablated-qwen3-mm3000-lora

Merged deployable weights from training job sensenova-u1-ablated-qwen3-mm3000-lora-fix5-2gpu-2ep-0708.

  • Base: SenseNova-U1-8B-MoT-ablated-qwen3
  • Data: aigf_mm_3000 (3000 image-text samples)
  • Train: 2 GPU / 2 epoch / 6000 steps
  • Strategy: freeze LLM + LoRA, unfreeze MoT image generation
  • Checkpoint: hf_6000_lora_merged3

Capability: text-in → Chinese roleplay text + image.

Downloads last month
8
Safetensors
Model size
18B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support