ALLaM-7B-Instruct-preview-fp16-ov

How this model was exported:

#requirements.txt
huggingface_hub>=0.25.0
transformers>=4.45.0
safetensors>=0.4.3
datasets>=3.0.0
accelerate>=0.33.0
tiktoken>=0.12.0
sentencepiece>=0.2.0
optimum-intel[openvino]>=1.26.0
openvino>=2025.0.0
nncf>=2.12.0
hf-transfer>=0.1.6
python-dotenv>=1.0.1
rich>=13.7.0

from huggingface_hub import snapshot_download, upload_folder
from transformers import AutoTokenizer
import subprocess

token = "hf_..."
DOWNLOAD_MODEL = "humain-ai/ALLaM-7B-Instruct-preview"
DOWNLOAD_FOLDER = "models/ALLaM-7B-Instruct-preview"
UPLOAD_FOLDER = "ov/ALLaM-7B-Instruct-preview-fp16-ov"
UPLOAD_MODEL = "KFUPM-JRCAI/ALLaM-7B-Instruct-preview-fp16-ov"
snapshot_download(
    repo_id=DOWNLOAD_MODEL,
    local_dir=DOWNLOAD_FOLDER,
    token=token,
)
tokenizer = AutoTokenizer.from_pretrained(DOWNLOAD_FOLDER) 
tokenizer.save_pretrained(UPLOAD_FOLDER)
subprocess.run(["optimum-cli", "export", "openvino", "-m", DOWNLOAD_FOLDER, "--task", "text-generation-with-past", "--weight-format", "fp16", UPLOAD_FOLDER])
subprocess.run(["convert_tokenizer", DOWNLOAD_FOLDER, "--o", UPLOAD_FOLDER, "--with-detokenizer"])
upload_folder(
    repo_id=UPLOAD_MODEL,
    folder_path=UPLOAD_FOLDER,
    commit_message='OpenVINO model upload',
    token=token,
    repo_type="model",
)