You can deploy this model and start earning money today!

AuraWorxAI/
weather-llm-initial

Name: weather-llm-initial
Availability: InStock

supervised-fine-tuning

apache-2.0

text-generation-inference

endpoints_compatible

Views

No views yet

Weather LLM — SFT (run1 stage2, step 2500)

Supervised fine-tuning checkpoint: run1_stage2/checkpoint-2500. This repository contains both the published model artifacts and supporting project code files.

Model Details

Model type: Causal language model for text generation (transformers).
Training flow: weather-domain pretraining followed by supervised fine-tuning (SFT).
Checkpoint lineage: this release corresponds to the uploaded run1_stage2/checkpoint-2500 model.
Tokenizer: Hub-compatible tokenizer files are included in this repository.
Context length: configured by uploaded tokenizer/model files (model_max_length=2048).

Data Sources

NOAA-derived historical weather data processed into natural-language records.
Optional support text from Wikipedia used during corpus building.
Synthetic and grounded instruction-response examples used for SFT dataset construction.

Code in Files Section

Code under src/, scripts/, and configs/ in the HF Files section is used for:

data building and preprocessing,
tokenizer training,
model pretraining and SFT training workflows,
reproducibility of training/evaluation steps.

Intended Use

Instruction-following weather-analysis style generation.
Educational and research usage for domain-focused LLM workflows.

Out-of-Scope / Limitations

Not for real-time weather operations or safety-critical decision-making.
Outputs may be incorrect, incomplete, or outdated and should be independently validated.
Not a replacement for authoritative meteorological services.

Inference

There are two inference entrypoints, and they are intentionally different:

Hugging Face Hub inference (loads by model repo ID).
Project-local inference script (loads by local model directory path used during local training workflow).

Hugging Face Hub Snippet

python

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "AuraWorxAI/weather-llm-initial"
model = AutoModelForCausalLM.from_pretrained(repo, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(repo)

Requires transformers and sentencepiece.

Project-Local Inference Snippet

bash

python -m weather_llm.inference.generate \
  --model_dir artifacts/checkpoints/sft/run1_stage2 \
  --prompt "Compare summer weather patterns in Arizona vs Washington."

The HF snippet loads from a Hub model ID, while the local snippet loads from a filesystem path produced in local training runs.

Contact

Maintainer: AuraWorxAI