Upload training/evez-kaggle-training.ipynb with huggingface_hub

training/evez-kaggle-training.ipynb

@@ -0,0 +1,50 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": ["# 🧬 EVEZ Self-Training Pipeline\n", "### Kaggle T4 GPU Fine-Tuning\n", "**Steps:** Settings → Accelerator → GPU T4 → Run All\n", "**Time:** ~20 min | **Cost:** $0 (free tier)"]
+  },
+  {
+   "cell_type": "code",
+   "metadata": {},
+   "source": ["!pip install -q transformers datasets peft trl torch accelerate\n", "import torch\n", "print(f'GPU: {torch.cuda.get_device_name(0) if torch.cuda.is_available() else \"NONE\"}')"],
+   "outputs": [],
+   "execution_count": null
+  },
+  {
+   "cell_type": "code",
+   "metadata": {},
+   "source": ["import json, urllib.request\n", "url = 'https://raw.githubusercontent.com/EVEZX/neuros/main/training/evez-alpaca.json'\n", "urllib.request.urlretrieve(url, 'evez-alpaca.json')\n", "with open('evez-alpaca.json') as f:\n", "    data = json.load(f)\n", "print(f'✅ Loaded {len(data)} instruction pairs')"],
+   "outputs": [],
+   "execution_count": null
+  },
+  {
+   "cell_type": "code",
+   "metadata": {},
+   "source": ["from transformers import AutoModelForCausalLM, AutoTokenizer\n", "from peft import LoraConfig, get_peft_model\n", "from trl import SFTTrainer, SFTConfig\n", "from datasets import Dataset\n", "\n", "formatted = []\n", "for d in data:\n", "    text = f\"### Instruction:\\n{d['instruction']}\\n### Input:\\n{d.get('input','')}\\n### Response:\\n{d['output']}\"\n", "    formatted.append({'text': text})\n", "dataset = Dataset.from_list(formatted)\n", "\n", "model_name = 'HuggingFaceTB/SmolLM2-135M'\n", "tokenizer = AutoTokenizer.from_pretrained(model_name)\n", "tokenizer.pad_token = tokenizer.eos_token\n", "model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map='auto')\n", "\n", "lora_config = LoraConfig(r=8, lora_alpha=16, lora_dropout=0.05, target_modules=['q_proj','v_proj'], task_type='CAUSAL_LM')\n", "model = get_peft_model(model, lora_config)\n", "model.print_trainable_parameters()"],
+   "outputs": [],
+   "execution_count": null
+  },
+  {
+   "cell_type": "code",
+   "metadata": {},
+   "source": ["training_args = SFTConfig(output_dir='./evez-adapter', num_train_epochs=3, per_device_train_batch_size=4, learning_rate=2e-4, logging_steps=10, max_seq_length=512, dataset_text_field='text')\n", "trainer = SFTTrainer(model=model, args=training_args, train_dataset=dataset, processing_class=tokenizer)\n", "trainer.train()"],
+   "outputs": [],
+   "execution_count": null
+  },
+  {
+   "cell_type": "code",
+   "metadata": {},
+   "source": ["model.save_pretrained('./evez-adapter')\n", "tokenizer.save_pretrained('./evez-adapter')\n", "print('✅ Training complete! Adapter saved to ./evez-adapter/')"],
+   "outputs": [],
+   "execution_count": null
+  }
+ ],
+ "metadata": {
+  "kernelspec": {"display_name": "Python 3", "language": "python", "name": "python3"},
+  "accelerator": "GPU"
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}