llama-3.2-3b-instruct-obfuscated-reviewer

Llama 3.2 3B Instruct Obfuscated Reviewer Package

This repository contains the obfuscated Llama 3.2 3B Instruct reviewer package.

Anonymous reviewer link:

https://anonymous-hf.up.railway.app/a/mzd746oc1d9p/

Repository Structure

.
|-- README.md
|-- LICENSE.txt
|-- USE_POLICY.md
|-- SHA256SUMS
|-- config.json
|-- generation_config.json
|-- model.safetensors.index.json
|-- model-00001-of-00009.safetensors
|-- model-00002-of-00009.safetensors
|-- model-00003-of-00009.safetensors
|-- model-00004-of-00009.safetensors
|-- model-00005-of-00009.safetensors
|-- model-00006-of-00009.safetensors
|-- model-00007-of-00009.safetensors
|-- model-00008-of-00009.safetensors
|-- model-00009-of-00009.safetensors
|-- tokenizer.json
|-- tokenizer_config.json
|-- special_tokens_map.json
|-- chat_template.jinja
`-- client/
    `-- client_secret.pt

Usage

Download the repository contents from the anonymous reviewer link and place the directory under:

models/llama-3.2-3b-instruct-obfuscated-reviewer/

The runtime configuration should use:

model_path=models/llama-3.2-3b-instruct-obfuscated-reviewer
client_secret_path=models/llama-3.2-3b-instruct-obfuscated-reviewer/client/client_secret.pt

Minimal dependency set:

pip install torch transformers accelerate safetensors

Basic local loading check:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_path = "models/llama-3.2-3b-instruct-obfuscated-reviewer"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(
    model_path,
    device_map="auto",
    torch_dtype="auto",
)
client_secret_path = f"{model_path}/client/client_secret.pt"

Included Files

• sharded safetensors model weights (model-00001-of-00009.safetensors ... model-00009-of-00009.safetensors)
• model.safetensors.index.json
• tokenizer files
• model configuration files
• chat template
• client/client_secret.pt for reviewer-side input mapping and output recovery
• upstream license and use policy files

Excluded Files

• deployment .env files
• internal hostnames, private IP addresses, logs, and databases