Views
No views yet
Qwen/Qwen3-14B for domain-adapted translation in avalanche, snowpack, weather, and natural-hazard texts.Qwen/Qwen3-14B1import torch
2from transformers import AutoTokenizer, AutoModelForCausalLM
3from peft import PeftModel
4
5base_model_id = "Qwen/Qwen3-14B"
6adapter_id = "DEIN_USERNAME/qwen3-14b-slf-lora"
7
8tokenizer = AutoTokenizer.from_pretrained(base_model_id, trust_remote_code=True)
9
10base_model = AutoModelForCausalLM.from_pretrained(
11 base_model_id,
12 device_map="auto",
13 dtype=torch.bfloat16,
14 trust_remote_code=True,
15)
16
17model = PeftModel.from_pretrained(base_model, adapter_id)
18model.eval()
19Notes
20
21This is a LoRA adapter only. The base model must be loaded separately.
22
23For critical use, outputs should be reviewed, especially:
24
25terminology
26numbers and units
27dates and times
28geographical expressions
29modal expressions such as likely/probable versus certain