metadata
license: apache-2.0
base_model: google/gemma-2-12b-it
tags:
- verilog
- code-generation
- instruction-tuned
- vericoder
Gemma-3-12B-IT (VeriCoder Dataset Ablation)
This is a fine-tuned version of Gemma-3-12B-IT model trained on VeriCoder dataset.
Model Details
- Base Model: Gemma-3-12B-IT
- Training Dataset: VeriCoder dataset (126k samples)
- Model Architecture: Gemma3ForCausalLM
- Parameters: ~11.7B
- Context Length: 131,072 tokens
- Sliding Window: 1024
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "LLM4Code/VeriCoder_Gemma12b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
# Example usage
inputs = tokenizer("Your prompt here", return_tensors="pt")
outputs = model.generate(**inputs, max_length=512)
print(tokenizer.decode(outputs[0]))
Training Details
- Dataset: VeriCoder dataset ablation (126k samples)
- Commit: ae17392c
Files
The model includes:
- Model weights in SafeTensors format (5 shards)
- Tokenizer files (tokenizer.json, tokenizer.model, tokenizer_config.json)
- Model configuration (config.json)
- Generation configuration (generation_config.json)
- Chat template (chat_template.jinja)