Hugging Face's logo

Oysiyl
/
qwen3.5-27b-unslop-good-lora-v1

Text Generation
Transformers
Safetensors
GGUF
English
qwen3_5
image-text-to-text
unsloth
lora
rewriting
style-transfer
unslop
conversational
Model card Files
xet
 Community
Oysiyl commited on
Commit
75d69bc
·
verified ·
1 Parent(s): a023b6b

plots: switch 27B family chart to normalized progress comparison style (add 27B curve)

Browse files
Files changed (3) hide show
  1. README.md +1 -1
  2. training_loss_qwen35_family.png +2 -2
  3. training_loss_qwen35_family.svg +48 -1612
README.md CHANGED
@@ -86,7 +86,7 @@ Across the successful Qwen 3.5 HF Jobs runs, trainer-reported final `train_loss`
86
  - 9B: 1.854
87
  - 27B: 1.916 (latest HF job run: `Oysiyl/69ca19caf900226fc14aea81`)
88
 
89
- ![Qwen3.5 family HF Jobs final train_loss](./training_loss_qwen35_family.svg)
90
 
91
  ## Recommended evaluation sample
92
  Use this full fiction passage for held-out testing:
 
86
  - 9B: 1.854
87
  - 27B: 1.916 (latest HF job run: `Oysiyl/69ca19caf900226fc14aea81`)
88
 
89
+ ![Normalized training loss comparison: 27B vs 9B vs 0.8B vs 2B vs 4B](./training_loss_qwen35_family.svg)
90
 
91
  ## Recommended evaluation sample
92
  Use this full fiction passage for held-out testing:
training_loss_qwen35_family.png CHANGED

Git LFS Details

  • SHA256: d34c872bef151ed2c9503adfc0b700973feefb26b97b30753139ed11fdf678f7
  • Pointer size: 131 Bytes
  • Size of remote file: 144 kB

Git LFS Details

  • SHA256: d6d5eb2bac4f8eab6cb2143871ea79c1ed10226a49d0d81fcc1cdd2627ef608a
  • Pointer size: 131 Bytes
  • Size of remote file: 521 kB
training_loss_qwen35_family.svg CHANGED