muradil211/AetherSearch

<div align="center"> <img src="assets/aethersearch-mark.svg" alt="AetherSearch monogram" width="128"> </div>

AetherSearch

AetherSearch is a search-augmented language model release trained through a multi-stage post-training pipeline with SFT, DPO, and reinforcement learning.

The released weights are provided in Hugging Face Transformers format and can be loaded with AutoModelForCausalLM and AutoTokenizer.

Files

• —model.safetensors: model weights.
• —config.json and generation_config.json: Transformers configuration.
• —tokenizer.json, tokenizer_config.json, vocab.json, merges.txt, added_tokens.json, and special_tokens_map.json: tokenizer assets.
• —MODEL_MANIFEST.sha256: SHA256 checksums for the uploaded files.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "muradil211/AetherSearch"

tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    torch_dtype="auto",
    device_map="auto",
)

Notes

The companion training code is released at: https://github.com/Muradil-mamat-211/AetherSearch

License and upstream base-model attribution should be set according to the actual base model and data release terms before wider redistribution.