40 lines
1.0 KiB
Markdown
40 lines
1.0 KiB
Markdown
---
|
|
license: apache-2.0
|
|
base_model: unsloth/gemma-3-1b-it
|
|
tags:
|
|
- sales
|
|
- conversational
|
|
- fine-tuned
|
|
- gemma
|
|
language:
|
|
- en
|
|
pipeline_tag: text-generation
|
|
---
|
|
|
|
# Gemma 3 1B - Sales Conversation Fine-tuned
|
|
|
|
Fine-tuned on 100K B2B sales conversations for objection handling and deal progression.
|
|
|
|
## Usage
|
|
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
model = AutoModelForCausalLM.from_pretrained("convaiinnovations/gemma3-fine-tuned", device_map="auto")
|
|
tokenizer = AutoTokenizer.from_pretrained("convaiinnovations/gemma3-fine-tuned")
|
|
|
|
messages = [{"role": "user", "content": "Customer: It's too expensive. How to respond?"}]
|
|
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
|
|
outputs = model.generate(inputs.to(model.device), max_new_tokens=256)
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
|
```
|
|
|
|
## Training
|
|
- Base: `unsloth/gemma-3-1b-it`
|
|
- LoRA: r=64, alpha=32
|
|
- Epochs: 3
|
|
- Data: 100K synthetic sales conversations
|
|
|
|
## License
|
|
Apache 2.0 | ConvAI Innovations 2026
|