Uploaded Model
- Developed by: Daemontatox
- License: apache-2.0
- Finetuned from model: unsloth/qwen2.5-coder-14b-instruct-bnb-4bit
Overview
This Qwen2 model has been finetuned using Unsloth and Hugging Face's TRL (Transformers Reinforcement Learning) library. The finetuning process achieved a 2x speedup compared to traditional methods.
Features
- Optimized for text generation and inference tasks.
- Lightweight with 4-bit quantization for efficient performance.
- Compatible with various NLP and code-generation applications.
Acknowledgments
This model leverages Unsloth’s advanced optimization techniques to ensure faster training and inference.
Open LLM Leaderboard Evaluation Results
Detailed results can be found here! Summarized results can be found here!
Metric | Value (%) |
---|---|
Average | 30.81 |
IFEval (0-Shot) | 66.37 |
BBH (3-Shot) | 46.48 |
MATH Lvl 5 (4-Shot) | 20.77 |
GPQA (0-shot) | 8.84 |
MuSR (0-shot) | 9.07 |
MMLU-PRO (5-shot) | 33.33 |
- Downloads last month
- 57
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for Daemontatox/CogitoZ14
Base model
Qwen/Qwen2.5-14B
Finetuned
Qwen/Qwen2.5-Coder-14B
Finetuned
Qwen/Qwen2.5-Coder-14B-Instruct
Dataset used to train Daemontatox/CogitoZ14
Evaluation results
- averaged accuracy on IFEval (0-Shot)Open LLM Leaderboard66.370
- normalized accuracy on BBH (3-Shot)test set Open LLM Leaderboard46.480
- exact match on MATH Lvl 5 (4-Shot)test set Open LLM Leaderboard20.770
- acc_norm on GPQA (0-shot)Open LLM Leaderboard8.840
- acc_norm on MuSR (0-shot)Open LLM Leaderboard9.070
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard33.330