qwen-finetuned-Reasoning-Socratic-QandA/README.md

base_model, tags, license, language, datasets

base_model	tags	license	language	datasets
unsloth/Qwen2.5-7B-Instruct	text-generation-inference transformers unsloth qwen2	apache-2.0	en	Bialy17/Reasoning-Socratic-QandA

Uploaded finetuned model

• Developed by: Bialy17
• License: apache-2.0
• Finetuned from model : unsloth/Qwen2.5-7B-Instruct

This qwen2 model was trained 2x faster with Unsloth and Huggingface's TRL library.

---

This is the same model as 'Bialy17/qwen-finetuned-Reasoning-Socratic-QandA-unsloth' but as a pure standalone model;

---

Fine tuned on Runpod using RTX4090 and unsloth template (docker.io/unsloth/unsloth:latest)

Trained with the below LoRA configurations

r = 64 lora alpha = 128 lora dropout = 0 max seq length = 2048

Trained for 1875 step

''' TrainOutput(global_step=1875, training_loss=0.6351530222256978, metrics={'train_runtime': 9419.7211, 'train_samples_per_second': 3.185, 'train_steps_per_second': 0.199, 'total_flos': 9.825842690769101e+17, 'train_loss': 0.6351530222256978, 'epoch': 1.9994666666666667}) '''