GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
TechxGenus
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
TechxGenus
Model Tree
01-ai/Yi-Coder-9B
Input Modalities
Output Modalities
Supported Functionality
While working with Typst documents, we noticed that AI programming assistants often generate poor results. I understand that these assistants may perform better in languages like Python and JavaScript, which benefit from more extensive datasets and feedback signals from executable code, unlike HTML or Markdown. However, current LLMs even frequently struggle to produce accurate Typst syntax, including models like GPT-4o and Claude-3.5-Sonnet.
Upon further investigation, we found that because Typst is a relatively new language, training data for it is scarce. GitHub's search tool doesn't categorize it as a language for code yet, and The Stack v1/v2 don’t include Typst. No open code LLMs currently list it as a supported language, either. To address this, we developed this project aimed at collecting relevant data and training models to improve Typst support in AI programming tools.
An example script is shown below:
python
import torchfrom transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("TechxGenus/Typst-Coder-9B")model = AutoModelForCausalLM.from_pretrained( "TechxGenus/Typst-Coder-9B", torch_dtype=torch.bfloat16, device_map="auto") messages = [ {"role": "user", "content": "Hi!"},]prompt = tokenizer.apply_chat_template( messages, tokenize=False, add_generation_prompt=True) inputs = tokenizer.encode(prompt, return_tensors="pt")outputs = model.generate(input_ids=inputs.to(model.device), max_new_tokens=512)print(tokenizer.decode(outputs[0]))