Configuration Parsing Warning:In adapter_config.json: "peft.task_type" must be a string
DiffusionGemma Social Writer LoRA
This is a LoRA adapter for unsloth/diffusiongemma-26B-A4B-it trained with Unsloth's
DiffusionGemma block-diffusion objective on founder/social rewrite examples
from the-shape-of-text.
The adapter is intentionally unmerged. Load it with the DiffusionGemma-capable Transformers/Unsloth stack and keep the base model separate.
Training summary:
- Hub repo:
micic-mihajlo/diffusiongemma-social-writer-lora - Train examples used: 424
- Train examples skipped for 256-token canvas overflow: 0
- Optimizer steps: 300
- LoRA rank: 32
- Precision: bf16
- Target task: concise founder-style social post rewrites with no headings, options, placeholders, or generic template language.
- Downloads last month
- 102
Model tree for micic-mihajlo/diffusiongemma-social-writer-lora
Base model
google/diffusiongemma-26B-A4B-it Finetuned
unsloth/diffusiongemma-26B-A4B-it