|
|
--- |
|
|
license: llama3.1 |
|
|
base_model: meta-llama/Llama-3-1-8B-Instruct |
|
|
library_name: transformers |
|
|
language: |
|
|
- en |
|
|
pipeline_tag: text-generation |
|
|
extra_gated_prompt: "Please provide answers to the below questions to gain access to the model" |
|
|
extra_gated_fields: |
|
|
Company: text |
|
|
Full Name: text |
|
|
Requests from Personal email IDs will be rejected by default Provide college or business email ID: text |
|
|
I want to use this model for: |
|
|
type: select |
|
|
options: |
|
|
- Research |
|
|
- Education |
|
|
- Commercial |
|
|
- label: Other |
|
|
value: other |
|
|
tags: |
|
|
- telecom |
|
|
- telecommunications |
|
|
- 5g |
|
|
- networking |
|
|
- domain-specific |
|
|
- agentic-ai |
|
|
- customer-support |
|
|
- network-operations |
|
|
--- |
|
|
|
|
|
<div align="center"> |
|
|
<img src="logo.png" alt="NetoAI Logo" width="500"/> |
|
|
|
|
|
# TSLAM-8B-L31 |
|
|
|
|
|
### Telecom-Specific Large Action Model for Real-Time Intelligent Agents |
|
|
|
|
|
[](https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE) |
|
|
[](https://huggingface.co/NetoAISolutions/TSLAM-8B-L31) |
|
|
</div> |
|
|
|
|
|
--- |
|
|
|
|
|
## Overview |
|
|
|
|
|
**TSLAM-8B-L31** is a production-ready, domain-specialized language model engineered for telecommunications operations. Built on an optimized version of Llama-3.1-8B-Instruct and fine-tuned on telecom-specific data, this model delivers SME-level expertise for real-time agent deployments, network operations, and customer support systems. |
|
|
|
|
|
**Key Capabilities:** |
|
|
- 🎯 Real-time customer support with technical accuracy |
|
|
- 🔧 Network troubleshooting and diagnostics |
|
|
- 📊 Service provisioning and activation workflows |
|
|
- 🤖 Autonomous agent operations |
|
|
- 📱 Multi-turn conversational intelligence |
|
|
|
|
|
--- |
|
|
|
|
|
## Model Details |
|
|
|
|
|
| Property | Value | |
|
|
|----------|-------| |
|
|
| **Base Model** | Llama-3.1-8B-Instruct (optimized) | |
|
|
| **Parameters** | 8 Billion | |
|
|
| **Context Window** | 128K tokens | |
|
|
| **Optimization** | Flash Attention 2, BF16 precision | |
|
|
| **License** | Llama 3.1 Community License | |
|
|
|
|
|
--- |
|
|
|
|
|
## Use Cases |
|
|
|
|
|
### 1. Autonomous Customer Support Agent |
|
|
|
|
|
Deploy AI agents that handle complex customer inquiries with technical precision: |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
import torch |
|
|
|
|
|
model_id = "NetoAISolutions/TSLAM-8B-L31" |
|
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
|
model = AutoModelForCausalLM.from_pretrained( |
|
|
model_id, |
|
|
torch_dtype=torch.bfloat16, |
|
|
device_map="auto" |
|
|
) |
|
|
|
|
|
messages = [ |
|
|
{ |
|
|
"role": "system", |
|
|
"content": "You are an expert telecommunications support agent helping customers resolve technical issues." |
|
|
}, |
|
|
{ |
|
|
"role": "user", |
|
|
"content": "My 5G connection keeps dropping every few minutes. I'm using a Samsung Galaxy S23 in downtown Chicago." |
|
|
} |
|
|
] |
|
|
|
|
|
inputs = tokenizer.apply_chat_template( |
|
|
messages, |
|
|
tokenize=True, |
|
|
add_generation_prompt=True, |
|
|
return_tensors="pt" |
|
|
).to(model.device) |
|
|
|
|
|
outputs = model.generate( |
|
|
inputs, |
|
|
max_new_tokens=512, |
|
|
temperature=0.7, |
|
|
top_p=0.9, |
|
|
do_sample=True, |
|
|
pad_token_id=tokenizer.eos_token_id |
|
|
) |
|
|
|
|
|
response = tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True) |
|
|
print(response) |
|
|
``` |
|
|
|
|
|
|
|
|
|
|
|
### 2. Network Operations Assistant |
|
|
|
|
|
Enable NOC teams to diagnose and resolve issues faster: |
|
|
|
|
|
```python |
|
|
messages = [ |
|
|
{ |
|
|
"role": "system", |
|
|
"content": "You are a network operations expert assisting NOC engineers with diagnostics and troubleshooting." |
|
|
}, |
|
|
{ |
|
|
"role": "user", |
|
|
"content": "Cell tower ID 2847 is showing high RACH failures. Current RACH success rate: 73%. What should I check?" |
|
|
} |
|
|
] |
|
|
|
|
|
``` |
|
|
|
|
|
### 3. Field Technician Copilot |
|
|
|
|
|
Provide real-time guidance for on-site installations and repairs: |
|
|
|
|
|
```python |
|
|
messages = [ |
|
|
{ |
|
|
"role": "system", |
|
|
"content": "You are an expert field technician assistant providing step-by-step guidance for installations and repairs." |
|
|
}, |
|
|
{ |
|
|
"role": "user", |
|
|
"content": "I'm installing a small cell unit. The fiber connection is established but I'm not seeing any signal propagation. What are the likely causes?" |
|
|
} |
|
|
] |
|
|
``` |
|
|
|
|
|
### 4. Service Provisioning Automation |
|
|
|
|
|
Streamline activation and configuration workflows: |
|
|
|
|
|
```python |
|
|
messages = [ |
|
|
{ |
|
|
"role": "system", |
|
|
"content": "You are a service provisioning specialist helping with device activation and configuration." |
|
|
}, |
|
|
{ |
|
|
"role": "user", |
|
|
"content": "I need to activate VoLTE for a customer on an iPhone 14 Pro. Walk me through the provisioning checklist." |
|
|
} |
|
|
] |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Inference Optimization |
|
|
|
|
|
### Using Transformers Pipeline |
|
|
|
|
|
```python |
|
|
import transformers |
|
|
import torch |
|
|
|
|
|
pipeline = transformers.pipeline( |
|
|
"text-generation", |
|
|
model="NetoAISolutions/TSLAM-8B-L31", |
|
|
model_kwargs={"torch_dtype": torch.bfloat16}, |
|
|
device_map="auto" |
|
|
) |
|
|
|
|
|
messages = [ |
|
|
{"role": "system", "content": "You are a helpful telecom expert."}, |
|
|
{"role": "user", "content": "Explain the difference between NSA and SA 5G deployments."} |
|
|
] |
|
|
|
|
|
outputs = pipeline(messages, max_new_tokens=256) |
|
|
print(outputs[0]["generated_text"][-1]["content"]) |
|
|
``` |
|
|
|
|
|
### Deployment with vLLM |
|
|
|
|
|
For production deployments requiring high throughput: |
|
|
|
|
|
```bash |
|
|
pip install vllm |
|
|
|
|
|
python -m vllm.entrypoints.openai.api_server \ |
|
|
--model NetoAISolutions/TSLAM-8B-L31 \ |
|
|
--dtype bfloat16 \ |
|
|
--max-model-len 8192 |
|
|
``` |
|
|
|
|
|
### Quantization with BitsAndBytes |
|
|
|
|
|
For memory-constrained environments: |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig |
|
|
|
|
|
quantization_config = BitsAndBytesConfig( |
|
|
load_in_4bit=True, |
|
|
bnb_4bit_compute_dtype=torch.bfloat16, |
|
|
bnb_4bit_use_double_quant=True, |
|
|
bnb_4bit_quant_type="nf4" |
|
|
) |
|
|
|
|
|
model = AutoModelForCausalLM.from_pretrained( |
|
|
"NetoAISolutions/TSLAM-8B-L31", |
|
|
quantization_config=quantization_config, |
|
|
device_map="auto" |
|
|
) |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Prompt Template |
|
|
|
|
|
TSLAM-8B-L31 uses the standard **Llama 3.1 chat template**: |
|
|
|
|
|
``` |
|
|
<|begin_of_text|><|start_header_id|>system<|end_header_id|> |
|
|
|
|
|
You are a helpful telecom expert assistant.<|eot_id|><|start_header_id|>user<|end_header_id|> |
|
|
|
|
|
How do I configure APN settings for LTE?<|eot_id|><|start_header_id|>assistant<|end_header_id|> |
|
|
``` |
|
|
|
|
|
The template is automatically applied when using `tokenizer.apply_chat_template()`. |
|
|
|
|
|
--- |
|
|
|
|
|
## Hardware Requirements |
|
|
|
|
|
| Deployment Type | GPU | VRAM | Precision | |
|
|
|----------------|-----|------|-----------| |
|
|
| **Full Precision** | A100 40GB | 40GB | BF16 | |
|
|
| **Recommended** | A10 / RTX 4090 | 24GB | BF16 | |
|
|
| **Quantized (4-bit)** | RTX 3090 / 4080 | 16GB | INT4 | |
|
|
| **CPU Inference** | 64GB RAM | - | FP32 | |
|
|
|
|
|
--- |
|
|
|
|
|
--- |
|
|
|
|
|
## Limitations & Considerations |
|
|
|
|
|
- **Domain Specificity**: Optimized for telecommunications use cases. Performance on general tasks may vary. |
|
|
- **Language**: Primarily trained on English telecom data. Multilingual support inherits from base Llama 3.1 model. |
|
|
- **Safety**: Deploy with appropriate content filtering for production customer-facing applications. |
|
|
- **Context**: While supporting 128K context, optimal performance is observed with contexts under 8K tokens. |
|
|
|
|
|
--- |
|
|
|
|
|
## Responsible AI |
|
|
|
|
|
This model should be deployed as part of a complete AI system with appropriate safeguards: |
|
|
- Implement content moderation for customer-facing applications |
|
|
- Monitor outputs for accuracy in critical operations |
|
|
- Maintain human oversight for network configuration changes |
|
|
- Follow industry compliance standards (GDPR, CCPA, etc.) |
|
|
|
|
|
--- |
|
|
|
|
|
## Citation |
|
|
|
|
|
If you use TSLAM-8B-L31 in your research or applications, please cite: |
|
|
|
|
|
```bibtex |
|
|
@misc{tslam-8b-l31, |
|
|
title={TSLAM-8B-L31: Telecom-Specific Large Action Model}, |
|
|
author={NetoAI Solutions}, |
|
|
year={2025}, |
|
|
publisher={HuggingFace}, |
|
|
howpublished={\url{https://huggingface.co/NetoAISolutions/TSLAM-8B-L31}} |
|
|
} |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Community & Support |
|
|
|
|
|
- 🤗 **Hugging Face**: [NetoAISolutions](https://huggingface.co/NetoAISolutions) |
|
|
- 🌐 **Website**: [netoai.ai](https://netoai.ai) |
|
|
- 📧 **Enterprise**: support@netoai.ai |
|
|
- 💼 **LinkedIn**: [NetoAI Solutions](https://www.linkedin.com/company/netoai/) |
|
|
|
|
|
--- |
|
|
|
|
|
## License |
|
|
|
|
|
Licensed under the [Llama 3.1 Community License](https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE). |
|
|
|
|
|
For commercial deployments exceeding 700M MAU, additional licensing may be required per Meta's terms. |
|
|
|
|
|
--- |
|
|
|
|
|
## Acknowledgments |
|
|
|
|
|
Built on top of [Meta's Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct) and developed by the NetoAI Solutions team |
|
|
|
|
|
--- |
|
|
|
|
|
<div align="center"> |
|
|
<sub>Built with ❤️ by NetoAI Solutions | Empowering the Future of Telecom with AI</sub> |
|
|
</div> |
