GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
ruby2022
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
ruby2022
Model Tree
Qwen/Qwen3-14B
Input Modalities
Output Modalities
Supported Functionality
| Submission | Inference | Public | Private |
|---|---|---|---|
| Qwen3-14B + GRPO | Majority Vote N=8 | 0.694 | 0.698 |
| Qwen3-14B + GRPO | Greedy (N=1) | 0.687 | 0.685 |
Kaggle: nycu-i-al-i-dl-2026-llm-3-grpo
Task: Fine-tune a Chinese-released LLM using GRPO so that it correctly answers Traditional Chinese single-choice questions (A/B/C/D).
| Parameter | Value |
|---|---|
| Base model | Qwen/Qwen3-14B |
| Rank (r) | 16 |
| Alpha | 16 |
| Dropout | 0.0 |
| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
| Trainable params | ~1.5% |
| Training method | GRPO (Group Relative Policy Optimization) |
| File | Description |
|---|---|
adapter_config.json | LoRA adapter configuration |
adapter_model.safetensors | LoRA adapter weights (~245 MB) |
tokenizer.json | Tokenizer |
tokenizer_config.json | Tokenizer configuration |
chat_template.jinja | Qwen3 chat template with thinking enabled |
train-reasoning-v2.csv |
python
from transformers import AutoTokenizer, AutoModelForCausalLMfrom peft import PeftModelimport torch base_model_name = "Qwen/Qwen3-14B"adapter_name = "ruby2022/NYCU-IAlI-DL2026-LLM3-GRPO" tokenizer = AutoTokenizer.from_pretrained(adapter_name)model = AutoModelForCausalLM.from_pretrained( base_model_name, dtype=torch.bfloat16, device_map="auto")model = PeftModel.from_pretrained(model, adapter_name)model = model.merge_and_unload()correctness_reward (+2.0/−1.0) + format_reward (+0.5)<think>...</think> enabled throughoutSee the GitHub repository for full source code, training scripts, and logs.
| Training dataset (3,199 samples) |
kaggle_test_set_792.csv | Test set (792 questions, no labels) |