Google Colab Notebook: Gemma 4 - 31B | Full Quantized

Here is the JSON configuration for running this notebook on Google Colab:

{
  "project": "google-colab",
  "job_name": "gemma-4-31b-fq",
  "runtime_version": "2.12.0",
  "region": "us-central1",
  "machine_type": "N1_HIGH_PERFORMANCE",
  "accelerator_type": "GPU",
  "image_project_id": "us-core-1",
  "image_version": "latest",
  "environment_version": "2.12.0",
  "extra_resources": {
    "gpu_memory_training_mode": true,
    "num_cpu_cores": 16,
    "cpu_threads_per_worker": 1
  },
  "network_config": {
    "network": "my-project"
  },
  "service_account_email": "your-account@my-project.iamcredentials.com",
  "additional_flags": [
    "--use_torch_xla",
    "--disable_client_library"
  ],
  "dockerfile": "FROM nvidia/cuda:11.2-base\nRUN apt-get update && apt-get install -y --no-install-recommends python3-pip\nWORKDIR /opt/python\nENV http_proxy=:8080\nENV https_proxy=:8080\nCOPY requirements.txt requirements.txt\nRUN pip3 install --no-cache-dir -r requirements.txt\nCMD [\"python3\", \"run.py\"]"
  },
  "requirements": [
    "torch==1.12.1+cu112",
    "transformers==4.25.0",
    "einops==0.3.0",
    "bitsandbytes==0.4.0",
    "unzip",
    "git", 
    "pyyaml"
  ]
}
        

Copy to Clipboard