app.py

import os
import shutil
import subprocess
import signal
os.environ["GRADIO_ANALYTICS_ENABLED"] = "False"
import gradio as gr

from huggingface_hub import create_repo, HfApi
from huggingface_hub import snapshot_download
from huggingface_hub import whoami
from huggingface_hub import ModelCard

from gradio_huggingfacehub_search import HuggingfaceHubSearch

from apscheduler.schedulers.background import BackgroundScheduler

from textwrap import dedent

HF_TOKEN = os.environ.get("HF_TOKEN")
OLLAMA_USERNAME = os.environ.get("OLLAMA_USERNAME").lower()
ollama_pubkey = open("/home/user/.ollama/id_ed25519.pub", "r")
ollama_q_methods = ["FP16","Q3_K_S", "Q3_K_M", "Q3_K_L", "Q4_0", "Q4_1", "Q4_K_S", "Q4_K_M", "Q5_0", "Q5_1", "Q5_K_S", "Q5_K_M", "Q6_K", "Q8_0"]


def process_model(model_id, q_method, use_imatrix, imatrix_q_method, private_repo, train_data_file, split_model, split_max_tensors, split_max_size, ollamafy, latest, maintainer, oauth_token: gr.OAuthToken | None):
    if oauth_token.token is None:
        raise ValueError("You must be logged in to use GGUF-my-repo")
    model_name = model_id.split('/')[-1]
    fp16 = f"{model_name}.fp16.gguf"

    try:
        api = HfApi(token=oauth_token.token)

        dl_pattern = ["*.md", "*.json", "*.model"]

        pattern = (
            "*.safetensors"
            if any(
                file.path.endswith(".safetensors")
                for file in api.list_repo_tree(
                    repo_id=model_id,
                    recursive=True,
                )
            )
            else "*.bin"
        )

        dl_pattern += pattern

        api.snapshot_download(repo_id=model_id, local_dir=model_name, local_dir_use_symlinks=False, allow_patterns=dl_pattern)
        print("Model downloaded successfully!")
        print(f"Current working directory: {os.getcwd()}")
        print(f"Model directory contents: {os.listdir(model_name)}")

        conversion_script = "convert_hf_to_gguf.py"
        fp16_conversion = f"python llama.cpp/{conversion_script} {model_name} --outtype f16 --outfile {fp16}"
        result = subprocess.run(fp16_conversion, shell=True, capture_output=True)
        print(result)
        if result.returncode != 0:
            raise Exception(f"Error converting to fp16: {result.stderr}")
        print("Model converted to fp16 successfully!")
        print(f"Converted model path: {fp16}")

        ### Ollamafy ###
        if ollama_model:
            model_maintainer = model_id.split('/')[-2]
            ollama_model_name = model_maintainer.lower() + '_' + model_name.lower() 
            ollama_modelfile_name = model_name + '_modelfile'
            # model_path = f"{HOME}/.cache/huggingface/hub/{model_id}"
          
            ollama_modelfile = open(ollama_modelfile_name, "w")
            # ollama_modelfile_path = quantized_gguf_path
            ollama_modelfile.write(quantized_gguf_path)
            ollama_modelfile.close()
            print(quantized_gguf_path)
    
            for ollama_q_method in ollama_q_methods:
                if ollama_q_method == "FP16":
                    ollama_conversion = f"ollama create  -q {ollama_q_method} -f {model_file} {OLLAMA_USERNAME}/{ollama_model_name}:{ollama_q_method.lower()}"
                else:
                    ollama_conversion = f"ollama create -f {model_file} {OLLAMA_USERNAME}/{ollama_model_name}:{ollama_q_method.lower()}"
                ollama_conversion_result = subprocess.run(ollama_conversion, shell=True, capture_output=True)
                print(ollama_conversion_result)
                if ollama_conversion_result.returncode != 0:
                    raise Exception(f"Error converting to Ollama: {ollama_conversion_result.stderr}")
                print("Model converted to Ollama successfully!")
                
                if maintainer:
                    ollama_push =  f"ollama push  {OLLAMA_USERNAME}/{model_name}:{q_method.lower()}"
                else: 
                    ollama_push =  f"ollama push  {OLLAMA_USERNAME}/{ollama_model_name}:{q_method.lower()}"
                ollama_push_result = subprocess.run(ollama_push, shell=True, capture_output=True)
                print(ollama_push_result)
                if ollama_push_result.returncode != 0:
                    raise Exception(f"Error converting to Ollama: {ollama_push_result.stderr}")
                print("Model pushed to Ollama library successfully!")
        
                if latest == True:
                    ollama_copy =  f"ollama cp  {OLLAMA_USERNAME}/{model_id.lower()}:{q_method.lower()} {OLLAMA_USERNAME}/{model_id.lower()}:latest"
                    ollama_copy_result = subprocess.run(ollama_copy, shell=True, capture_output=True)
                    print(ollama_copy_result)
                    if ollama_copy_result.returncode != 0:
                        raise Exception(f"Error converting to Ollama: {ollama_push_result.stderr}")
                    print("Model pushed to Ollama library successfully!")
                    if maintainer == True:
                        llama_push_latest =  f"ollama push  {OLLAMA_USERNAME}/{model_name}:latest"
                    else:    
                        ollama_push_latest =  f"ollama push  {OLLAMA_USERNAME}/{ollama_model_name}:latest"
                    ollama_push_latest_result = subprocess.run(ollama_push_latest, shell=True, capture_output=True)
                    print(ollama_push_latest_result)
                    if ollama_push_latest_result.returncode != 0:
                        raise Exception(f"Error converting to Ollama: {ollama_push_result.stderr}")
                    print("Model pushed to Ollama library successfully!")
            
       
    except Exception as e:
        return (f"Error: {e}", "error.png")
    finally:
        shutil.rmtree(model_name, ignore_errors=True)
        print("Folder cleaned up successfully!")

        
css="""/* Custom CSS to allow scrolling */
.gradio-container {overflow-y: auto;}
"""
# Create Gradio interface
with gr.Blocks(css=css) as demo: 
    gr.LoginButton(min_width=250)
    gr.Markdown("You must be logged in to use Ollamafy.")
    gr.Markdown(ollama_pubkey.read().rstrip())
    ollama_pubkey.close()

    model_id = HuggingfaceHubSearch(
        label="Hub Model ID",
        placeholder="Search for model id on Huggingface",
        search_type="model",
    )

    ollama_q_method
        latest = gr.Dropdown(
        ollama_q_methods,
        label="Ollama Lastest Quantization Method",
        info="Chose which quantization will be labled with the latest tag in the Ollama Library",
        value="FP16", 
        filterable=False,
        visible=False
    )
    latest = gr.Checkbox(
        value=False,
        label="Latest",
        info="Copy Model to Ollama Library with the :latest tag"
    )
    
    maintainer = gr.Checkbox(
        value=False,
        label="Maintainer",
        info="This is your original repository on both Hugging Face and Ollama. (DO NOT USE!!!)"
    )
    
    iface = gr.Interface(
        fn=process_model,
        inputs=[
            model_id,
            ollama_q_method,
            latest,
            maintainer
        ],
        outputs=[
            gr.Markdown(label="output"),
            gr.Image(show_label=False),
        ],
        title="Create your own Ollama Models and Push them to the Ollama Library, blazingly fast ⚡!",
        description="Sampled from https://huggingface.co/spaces/ggml-org/gguf-my-repo and https://huggingface.co/spaces/gingdev/ollama-server",
        api_name=False
    )
    
#username = whoami(oauth_token.token)["name"]
def restart_space():
    HfApi().restart_space(repo_id="unclemusclez/ollamafy", token=HF_TOKEN, factory_reboot=True)

scheduler = BackgroundScheduler()
scheduler.add_job(restart_space, "interval", seconds=21600)
scheduler.start()

# Launch the interface
demo.queue(default_concurrency_limit=1, max_size=5).launch(debug=True, show_api=False)