---
license: apache-2.0
base_model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
tags:
  - llama
  - deepseek
  - r1
  - fine-tuned
  - media
  - blockchain
  - ai
  - purple-squirrel
  - conversational
language:
  - en
pipeline_tag: text-generation
library_name: transformers
widget:
  - text: "What is Purple Squirrel?"
    example_title: "About Platform"
  - text: "Can I mint NFTs with Purple Squirrel?"
    example_title: "NFT Minting"
  - text: "What AI capabilities does Purple Squirrel have?"
    example_title: "AI Features"
  - text: "How does Purple Squirrel handle media storage?"
    example_title: "Storage"
inference:
  parameters:
    max_new_tokens: 256
    temperature: 0.3
    do_sample: true
datasets:
  - purplesquirrelnetworks/purple-squirrel-training
---

# Purple Squirrel R1

A fine-tuned version of [DeepSeek-R1-Distill-Llama-8B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-8B) specialized for Purple Squirrel - an AI-powered media platform combining blockchain technology with intelligent content management.

## Model Details

- **Base Model:** DeepSeek-R1-Distill-Llama-8B
- **Parameters:** 8B
- **Training Method:** LoRA (Low-Rank Adaptation)
- **Training Data:** 74 curated examples covering Purple Squirrel features
- **Epochs:** 10
- **Final Loss:** 0.43
- **Token Accuracy:** 91%

## Capabilities

This model is trained to answer questions about:

- **Core Platform:** What Purple Squirrel is and its mission
- **AI Features:** Transcription, tagging, intelligent search, content analysis
- **Blockchain:** NFT minting, decentralized storage (Arweave/IPFS), ownership verification
- **Media Management:** File formats, organization, collaboration
- **Technical:** API integration, SDKs, security features

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "purplesquirrelnetworks/purple-squirrel-r1"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

messages = [{"role": "user", "content": "What is Purple Squirrel?"}]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=256,
    temperature=0.3,
    do_sample=True,
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## API Access

This model is deployed on Modal with a REST API:

```bash
curl -X POST "https://purplesquirrelnetworks--purple-squirrel-r1-api-purplesqu-1e3c01.modal.run" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is Purple Squirrel?"}'
```

## Training Details

| Parameter | Value |
|-----------|-------|
| LoRA Rank | 64 |
| LoRA Alpha | 16 |
| Learning Rate | 5e-4 |
| Batch Size | 2 |
| Gradient Accumulation | 4 |
| Optimizer | paged_adamw_8bit |
| Precision | bfloat16 |
| GPU | NVIDIA A10G |

## Example Outputs

**Q: What is Purple Squirrel?**
> Purple Squirrel is a comprehensive digital media management platform designed to help content creators streamline their workflow and maximize the value of their digital assets.

**Q: Can I mint NFTs with Purple Squirrel?**
> Yes, Purple Squirrel allows you to create and mint NFTs. The platform supports various blockchains like Ethereum, Solana, and others.

**Q: What AI capabilities does Purple Squirrel have?**
> Purple Squirrel offers AI-driven features including smart indexing, face detection, object recognition, automatic tagging, content moderation, intelligent search, and personalized recommendations.

## Limitations

- Optimized for Purple Squirrel-related queries
- May produce R1-style reasoning (`<think>` tags) in responses
- Best results with questions similar to training data

## License

Apache 2.0

## Links

- [Purple Squirrel Networks](https://purplesquirrel.io)
- [Modal API Deployment](https://modal.com/apps/purplesquirrelnetworks/main/deployed/purple-squirrel-r1-api)