Phind CodeLlama 34B v2 - Mirror Model Installer

⚠️ Important Disclaimer

This repository and the tools within are provided as a workaround for users who are unable to download Ollama models directly due to restrictive corporate network policies. You are solely responsible for ensuring that using this model and the methods in this repository complies with your organization's IT, data security, and fair use policies. If you have any reason to believe that downloading, storing, or using this model on your corporate-managed device is against policy, DO NOT PROCEED. By using the contents of this repository, you acknowledge that you are doing so at your own risk and are responsible for any potential policy violations.

---

About This Repository

Want to code on your SOE while trapped in a metal tube at 35,000 feet? Or perhaps during your sardine-can commute experience? Maybe in your off-grid man-cave where even carrier pigeons can't find you? Use your favorite AI IDE coding agent like Cline, Roo, Kilo, or Kiro to leverage codellama - a "where-have-you-been-all-my-life" coding companion that works without the internet's blessing! 🧙‍♂️✈️🚇

This repository provides a robust, resumable installer for the phind-codellama-34b-v2 Ollama model, designed for users on networks where direct downloads from ollama.ai and/or cloudflarestorage.com/ollama/ are blocked. The model is split into 50MB chunks for reliable downloading, with a script to reassemble, verify, and install the model for Ollama.

Target Audience

• Developers and engineers on macOS or Windows laptops who are blocked from downloading Ollama models directly.
• This model requires at least 24GB of RAM to run effectively.

Prerequisites

• Ollama installed and run at least once (to create the models directory).
• At least 40GB of free disk space.
• At least 24GB of RAM to run the model.

Installation

macOS (Recommended: One-Line Install)

Open your terminal and run:

curl -sSL https://raw.githubusercontent.com/enelass/phind-codellama-34b-v2/main/install.sh | bash

What the Installer Does

• Checks for macOS and the presence of the Ollama models directory.
• Verifies at least 40GB of free disk space.
• Downloads the model's SHA256 file.
• Downloads all model chunk parts with a real progress bar and resumable downloads.
• Reassembles the model from chunks with a spinner and verifies the SHA256 checksum (with a spinner).
• Moves the model file to the correct Ollama models directory.
• Downloads all required manifest and metadata files (including all sha256- files) from GitHub, not the Ollama registry.*
• Prints a final message with the run command.

Using the Model

Once installation is complete, you can run the model with:

ollama run phind-codellama:34b-v2

The model will appear in Ollama automatically if the steps complete successfully.

Windows (Manual Steps)

There is no automated script for Windows, but you can install the model manually:

1. Download All Chunks

   Use a download manager or PowerShell to download all part_* files from the model-chunks directory in this repository.

   Example PowerShell (from the repo root):

   $baseUrl = "https://github.com/enelass/phind-codellama-34b-v2/raw/main/model-chunks"
   $parts = @()
   foreach ($first in 'a'..'n') {
     foreach ($second in 'a'..'z') {
       $suffix = "$first$second"
       $parts += $suffix
       if ($suffix -eq "nz") { break }
     }
     if ($suffix -eq "nz") { break }
   }
   foreach ($suffix in $parts) {
     $url = "$baseUrl/part_$suffix"
     Invoke-WebRequest -Uri $url -OutFile "part_$suffix"
   }

2. Concatenate Chunks

   In Command Prompt (CMD):

   copy /b part_aa+part_ab+part_ac+...+part_nz sha256-45488384ce7a0a42ed3afa01b759df504b9d994f896aacbea64e5b1414d38ba2

   (Replace ... with all chunk names in order. You can generate this command with a script or by using PowerShell's Get-ChildItem.)

3. Download Manifest and Metadata Files

   Download the following files from GitHub and place them in your Ollama manifests directory (usually %USERPROFILE%\.ollama\models\manifests\registry.ollama.ai\library\phind-codellama\):

   • 34b-v2
   • sha256-41774062cd349c744e8ee986c1aaf5784b7e42fbe306619536fa7386d421da78
   • sha256-62ab78abadd613cb882be1e63a1a0d843685858f5f9456e7c4e3350284245d04
   • sha256-ee041eb771eb86734e51c251e840d15b6933121777e2ca290a6dece875f3ecd1
   • sha256-fb9390528638173921c5100c17dbba3d549a651a83caf1d00ecc6ab437052a13

4. Verify SHA256 Checksum

   In Command Prompt:

   certutil -hashfile sha256-45488384ce7a0a42ed3afa01b759df504b9d994f896aacbea64e5b1414d38ba2 SHA256

   The output should match:

   45488384ce7a0a42ed3afa01b759df504b9d994f896aacbea64e5b1414d38ba2
   

5. Move the Model File

   Move the reassembled file to your Ollama models directory (usually):

   %USERPROFILE%\.ollama\models\blobs\
   

   The final path should be:

   %USERPROFILE%\.ollama\models\blobs\sha256-45488384ce7a0a42ed3afa01b759df504b9d994f896aacbea64e5b1414d38ba2
   

Troubleshooting

• Ensure you have run Ollama at least once before installing (to create the models directory).
• If you encounter disk space or permission errors, free up space or check your user permissions.
• If the install is interrupted, simply re-run the install command (macOS) or repeat the manual steps (Windows) to resume.
• The installer will always download the latest manifest and metadata files from GitHub to ensure compatibility.

License

See LICENSE for details.