GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
victoria-kp
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
victoria-kp
Model Tree
Qwen/Qwen2.5-VL-7B-Instruct
Input Modalities
Output Modalities
Supported Functionality
| Eval on | Goodbuy (sim) | Bestbuy (real) |
|---|---|---|
| This adapter | 4.4% | 5.3% |
python
from peft import PeftModelfrom transformers import Qwen2_5_VLForConditionalGenerationbase = Qwen2_5_VLForConditionalGeneration.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")model = PeftModel.from_pretrained(base, "victoria-kp/qwen25vl-7b-simtoreal-sft-bestbuy")
QLoRA (4-bit NF4 base, LoRA rank 64, alpha 128, attention + MLP layers) on success-only Claude Fable teacher trajectories from the Bestbuy training split. One example per agent step: observation in (screenshot, task instruction, BID element list, history), teacher reasoning and JSON action out. Learning rate 5e-5 cosine with 5% warmup, effective batch size 2, max sequence length 4096. This is the validation-loss-optimal checkpoint (step 700, epoch 0.53).