Downloads · 30 days
4
14% of all-time downloads
tiennn/lab21-qwen35-triage-vi
lab21-qwen35-triage-vi is a text generation model from tiennn. Use it when you need the model to write or continue text. It is set up for peft.
LoRA adapter trained on 225 Vietnamese e-commerce support tickets to emit a four-field JSON triage object (intent, urgency, product, sentiment). The controlled local run used Qwen/Qwen3.5-2B, rank 16, alpha 32, all te…
Downloads · 30 days
4
14% of all-time downloads
All-time downloads
28
Public
Repo size
87.3 MB
Likes
0
Public
Click a slice to open those files.
.safetensors67.3 MB · 77%
From the Hugging Face model README
LoRA adapter trained on 225 Vietnamese e-commerce support tickets to emit a four-field
JSON triage object (intent, urgency, product, sentiment). The controlled local
run used Qwen/Qwen3.5-2B, rank 16, alpha 32, all text-decoder linear modules, learning
rate 1e-4, and 58 optimizer steps.
Measured on the frozen 50-item target set, target field accuracy was 0.995 and format validity was 1.000. The general-capability regression score fell from 0.6000 to 0.0667, so the deployment gate is honestly FAILED. This adapter should only be evaluated in a routed, triage-only endpoint; it is not a general-purpose replacement for the base.
The complete experiment report and frozen evaluation artifacts are in the companion GitHub repository: https://github.com/Tienlee41/Day21-Track3-2A202601145_LeAnhTien