phi3-mini-5g-faults-specialist

Phi-3 Mini 5G Faults Specialist

Model Description

This model is a specialized fine-tuned version of johnlam90/phi3-mini-eni-specialist that has been further trained on 5G network fault diagnosis scenarios. It combines the general ENI (Evolved Network Intelligence) domain expertise of the base model with specialized knowledge for diagnosing and troubleshooting 5G network issues.

Base Model

• Base Model: johnlam90/phi3-mini-eni-specialist
• Architecture: Microsoft Phi-3 Mini (3.8B parameters)
• Original Base: microsoft/Phi-3-mini-4k-instruct

Training Details

Dataset

• Training Dataset: greenwich157/5G_Faults_Full
• Dataset Size: 1,993 5G fault diagnosis examples
• Format: Instruction-following format (Alpaca style)

Training Configuration

• Method: LoRA (Low-Rank Adaptation) fine-tuning
• LoRA Rank: 16
• LoRA Alpha: 32
• LoRA Dropout: 0.05
• Target Modules: qkv_proj, o_proj, gate_proj, up_proj, down_proj
• Trainable Parameters: 15.2M (0.4% of total model parameters)

Training Parameters

• Epochs: 3
• Batch Size: 8 (per device)
• Gradient Accumulation Steps: 4
• Effective Batch Size: 32
• Learning Rate: 5e-6
• Weight Decay: 0.0
• Warmup Steps: 20
• Max Gradient Norm: 0.3
• Max Sequence Length: 1024
• Precision: bfloat16
• Hardware: NVIDIA H100 80GB HBM3

Capabilities

This model specializes in:

1. 5G RAN (Radio Access Network) Issues

   • Connection failures
   • Signal strength problems
   • Coverage optimization

2. 5G Core Network Problems

   • Latency issues
   • Performance degradation
   • Network congestion

3. 5G Handover Issues

   • Inter-cell handover failures
   • Mobility management problems
   • Connection drops during movement

4. 5G Network Slicing

   • Slice performance issues
   • SLA compliance problems
   • Resource allocation optimization

Usage

Basic Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-4k-instruct", trust_remote_code=True)

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    "microsoft/Phi-3-mini-4k-instruct",
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)

# Load fine-tuned model
model = PeftModel.from_pretrained(
    base_model,
    "johnlam90/phi3-mini-5g-faults-specialist",
    torch_dtype=torch.bfloat16
)

# Format prompt
def format_prompt(instruction, input_text=""):
    if input_text.strip():
        return f"### Instruction:\n{instruction}\n\n### Input:\n{input_text}\n\n### Response:\n"
    else:
        return f"### Instruction:\n{instruction}\n\n### Response:\n"

# Example usage
instruction = "Diagnose the following 5G network issue and provide troubleshooting steps."
input_text = "User reports unable to connect to 5G network. Device shows 'No Service' despite being in a 5G coverage area."

prompt = format_prompt(instruction, input_text)
inputs = tokenizer(prompt, return_tensors="pt")

with torch.no_grad():
    outputs = model.generate(
        **inputs,
        max_new_tokens=256,
        temperature=0.7,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id
    )

response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
print(response)

Example Outputs

5G RAN Connection Issue:

Input: "User reports unable to connect to 5G network. Device shows 'No Service' despite being in a 5G coverage area."

Output: "This could be due to several reasons:

1. Check if the device's 5G feature is activated. If not, enable it.
2. Reboot the device to refresh network connections.
3. Verify the device is not in Airplane Mode or has connectivity restrictions turned on.
4. Check if the network provider has issued a service outage.
5. If possible, try connecting to a different 5G network to determine if the issue is device-specific or network-specific."

5G Core Network Latency:

Input: "5G network experiencing high latency (>100ms) during peak hours. Users report slow data speeds and connection timeouts."

Output: "To address the high latency in the 5G network during peak hours, consider the following solutions:

1. Network optimization: Review and optimize the network configuration, including resource allocation and traffic prioritization.
2. Capacity planning: Assess the network's capacity to handle peak traffic volumes, and if necessary, increase the number of cell sites.
3. Traffic management: Implement traffic shaping and prioritization techniques to manage congestion.
4. Load balancing: Distribute user traffic evenly across multiple base stations to prevent overloading."

Performance

The model demonstrates strong performance in 5G fault diagnosis scenarios:

• Provides structured, actionable troubleshooting steps
• Shows understanding of 5G-specific terminology and concepts
• Offers comprehensive solutions covering multiple aspects of network issues
• Maintains technical accuracy in recommendations

Limitations

• Specialized for 5G network fault diagnosis; may not perform as well on general tasks
• Inherits limitations from the base Phi-3 Mini model
• Training data focused on specific 5G fault scenarios
• Requires domain expertise to validate technical recommendations

Ethical Considerations

This model is designed for technical troubleshooting and should be used by qualified network engineers and technicians. Users should validate all recommendations against their specific network configurations and safety protocols.

Citation

If you use this model, please cite:

@misc{phi3-5g-faults-specialist,
  title={Phi-3 Mini 5G Faults Specialist},
  author={johnlam90},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/johnlam90/phi3-mini-5g-faults-specialist}
}

License

This model is released under the MIT License, following the licensing of the base models.