IridiumLlama-72B-v0.1

IridiumLlama-72B-v0.1

Model Description

IridiumLlama is a 72B parameter language model created through a merge of Qwen2-72B-Instruct, calme2.1-72b, and magnum-72b-v1 using model_stock.

This is converted from leafspark/Iridium-72B-v0.1 (currently private)

Features

• 72 billion parameters
• Sharded in 31 files (unlike Iridium, which has 963 shards due to the merging process)
• Combines Magnum prose with Calam smarts
• Llamaified for easy use

Technical Specifications

Architecture

• LlamaForCasualLM
• Models: Qwen2-72B-Instruct (base), calme2.1-72b, magnum-72b-v1
• Merged layers: 80
• Total tensors: 1,043
• Context length: 32k

Tensor Distribution

• Attention layers: 560 files
• MLP layers: 240 files
• Layer norms: 160 files
• Miscellaneous (embeddings, output): 162 files

Merging

Custom script utilizing safetensors library.

Usage

Loading the Model

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained("leafspark/IridiumLlama-72B-v0.1", 
                                             device_map="auto", 
                                             torch_dtype=torch.float16)
tokenizer = AutoTokenizer.from_pretrained("leafspark/IridiumLlama-72B-v0.1")

GGUFs

Find them here: mradermacher/IridiumLlama-72B-v0.1-GGUF

Hardware Requirements

• At least ~150GB of free space
• ~150GB VRAM