ilkeileri/gemma-sales-comprehensive

🎯 Gemma Sales Comprehensive

A Gemma 1.1 2B model fine-tuned specifically for sales conversations and customer communication.

📊 Model Details

• Base Model: google/gemma-1.1-2b-it
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Parameter Size: 22.3 MB (adapter weights only)
• LoRA Rank: 8
• LoRA Alpha: 16
• Target Modules: q_proj, k_proj, v_proj, o_proj
• Languages: English & Turkish

🚀 Usage

Installation

pip install transformers peft torch

Loading the Model

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

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    "google/gemma-1.1-2b-it",
    torch_dtype=torch.float16,
    device_map="auto"
)

# Load fine-tuned adapter
model = PeftModel.from_pretrained(
    base_model,
    "YOUR_USERNAME/gemma-sales-comprehensive"
)

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained("YOUR_USERNAME/gemma-sales-comprehensive")

Example Usage

# Simple usage
prompt = "How should I approach a potential customer?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(
    **inputs,
    max_length=300,
    temperature=0.7,
    do_sample=True,
    top_p=0.9
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Interactive Chat

while True:
    prompt = input("\nYour question (type 'quit' to exit): ")
    if prompt.lower() == 'quit':
        break

    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_length=300, temperature=0.7)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    print(f"\nResponse: {response}")

💡 Use Cases

• ✅ Sales conversation strategies
• ✅ Handling customer objections
• ✅ Product/service presentations
• ✅ Customer relationship management
• ✅ Communication techniques
• ✅ Sales pitch generation

📈 Training Details

The model was fine-tuned on a specialized dataset containing sales scenarios and customer communication examples.

Training Hyperparameters

• LoRA Rank (r): 8
• LoRA Alpha: 16
• LoRA Dropout: 0.1
• Target Modules: q_proj, k_proj, v_proj, o_proj
• Task Type: Causal Language Modeling

⚙️ Technical Specifications

Model Architecture

• Base: Gemma 1.1 2B Instruction-Tuned
• Adapter Type: LoRA (Low-Rank Adaptation)
• Adapter Size: ~22 MB
• Total Parameters (with base): ~2B

Inference

# For faster inference
model.eval()
with torch.no_grad():
    outputs = model.generate(**inputs)

⚠️ Limitations

• The model is specialized for sales and customer communication contexts
• May not perform as well as the base model on general-purpose tasks
• Promotes ethical and professional sales practices
• Responses should be reviewed before use in production

🔒 Safety & Ethics

This model is designed to assist with professional sales communication. It should:

• Promote honest and transparent sales practices
• Respect customer autonomy and consent
• Avoid manipulative or deceptive tactics
• Comply with relevant sales regulations and guidelines

📝 License

Apache 2.0

🤝 Contributors

[Add team member names here]

📧 Contact

For questions or feedback: [email/contact info]

🙏 Acknowledgments

• Built with Hugging Face Transformers
• Uses PEFT for efficient fine-tuning
• Based on Google's Gemma

---

Note: This is a LoRA adapter and requires the base Gemma model to function. The adapter weights are ~22 MB, while the full model with base weights is ~5 GB.