GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
dkaczer
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
dkaczer
Model Tree
Qwen/Qwen3-14B
Input Modalities
Output Modalities
Supported Functionality
python
from peft import PeftModelfrom transformers import AutoModelForCausalLM, AutoTokenizer # transformers renamed torch_dtype -> dtype in v5; use torch_dtype= on transformers 4.xbase = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-14B", dtype="bfloat16", device_map="auto")model = PeftModel.from_pretrained(base, "dkaczer/qwen3-14b-sft-medical-100")tok = AutoTokenizer.from_pretrained("dkaczer/qwen3-14b-sft-medical-100")With vLLM, serve it as a LoRA request against the Qwen/Qwen3-14B base in bf16.
Base model Qwen/Qwen3-14B, LoRA rank 32, alpha 64, rsLoRA, dropout 0, applied to
q_proj k_proj v_proj o_proj gate_proj up_proj down_proj. Trained in bf16.
SFT: 100 examples from data/sft/medical_misaligned_train_100.jsonl, 1 epoch (7 optimizer steps), lr 1e-4, batch 4 x grad-accum 4, adamw_8bit, linear schedule, seed 0, max_seq_len 2048, loss on response tokens only.
bibtex
@misc{emarl2026, title={Reinforcement Learning Can Amplify Emergent Misalignment from Harmless Rewards}, author={J{\o}rgenv{\aa}g, Magnus and Kacz{\'e}r, David and Ruttert, Lasse and G{\"u}lhan, Marvin and Flek, Lucie and Mai, Florian}, year={2026}, eprint={2605.31328}, archivePrefix={arXiv}}Training data derives from Chua et al. (2025) and Woodruff (2025), both CC BY 4.0.