Downloads · 30 days
238
90% of all-time downloads
arabellako22/chyon_merged_full_gemma_4_legacy
chyon_merged_full_gemma_4_legacy is a text generation model from arabellako22. Use it when you need the model to write or continue text. The card lists the license as gemma.
Status: not recommended. This is the earlier full-weight Chyon merge. Artifact testing found repeated prompt-template tokens in generated output. Use the verified Unsloth re-merge instead.
Downloads · 30 days
238
90% of all-time downloads
All-time downloads
264
Public
Parameters
5.1B
10.3 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors10.2 GB · 100%
From the Hugging Face model README
Status: not recommended. This is the earlier full-weight Chyon merge. Artifact testing found repeated prompt-template tokens in generated output. Use the verified Unsloth re-merge instead.
| Field | Value |
|---|---|
| Developer | arabellako22 |
| Original base | unsloth/gemma-4-E2B-it |
| LoRA adapter | arabellako22/chyon_lora_model |
| Merge method | Standard PEFT merge_and_unload() path |
| Format | Full-weight Safetensors, approximately 10.3 GB |
| Architecture | Gemma4ForConditionalGeneration |
| Precision | FP16 |
| Task | Conversational text generation |
| Languages | Korean; English was not separately evaluated |
| License | Gemma license |
The files remain available for reproducibility and comparison. They should not be used as the source for GGUF conversion or deployment because the post-merge smoke test failed.
The replacement model was saved through Unsloth's 16-bit merge path, reloaded, and verified before GGUF conversion:
| Check | Result |
|---|---|
| Model reload | Completed |
| Persona prompt generation | Repeated prompt-template tokens |
| Release decision | Rejected in favor of the verified Unsloth re-merge |
Use the verified replacement for inference and downstream conversion.