AuraWorxAI/weather-llm-sft

Weather LLM (SFT)

Instruction-tuned decoder for U.S. historical weather language (not live forecasts).

Loading

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "AuraWorxAI/weather-llm-sft"

tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
    device_map="auto" if torch.cuda.is_available() else None,
)

Prompt format (matches training): ### Instruction:\n...\n### Response:\n

trust_remote_code=True is required for the SentencePiece WeatherSpmTokenizer shipped in this repo.