GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
deep-conrad
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
deep-conrad
Model Tree
Input Modalities
Output Modalities
Supported Functionality
train_stage1.py for the Stage 1 GPT baselinetrain_llama_stage2.py for LoRA fine-tuning on LLaMAtrain_lora.py as a compatibility entry point for the LoRA pathtrain_pipeline.py to run both stages in sequencebuild_datasets.py to generate topic-sharded synthetic datasetsdata/ with example JSONL training setsThe project is intended to produce two artifacts:
The Stage 2 model is the release artifact for normal inference.
Stage 2 currently targets:
meta-llama/Llama-3.1-8B-InstructThe earlier LoRA script in this repo also supports the same model family.
Recommended flow:
python build_datasets.py --output_dir data/generated --include_docs to generate topic-specific shards and aggregate JSONL files.data/generated/stage1_sft.jsonl if you want the GPT baseline.python train_stage1.py to build the small baseline.data/generated/stage2_conrad_sft.jsonl or let python train_pipeline.py --include_docs do the build step automatically.python merge_stage2_lora.py.python sync_checkpoint.py.This project is designed for:
CONRAD_ENDPOINT_URL and HF_TOKEN in the Space secrets or environment to enable production chat.