refactor: restructure project - scripts at root level

train-on-this-server.sh

@@ -20,6 +20,6 @@ pip install transformers datasets trl peft accelerate bitsandbytes deepspeed
 
 # Run training
 echo "Starting training with accelerate..."
-accelerate launch --multi_gpu --num_processes 2 training/scripts/train.py --config training/configs/ornith-35b-lora.yaml
+accelerate launch --multi_gpu --num_processes 2 train.py --config training/configs/ornith-35b-lora.yaml
 
 echo "Training completed!"

train.py

@@ -110,8 +110,8 @@ def train(config_path):
     from datasets import load_dataset
     import os
     
-    # Get dataset paths - go up to project root
+    # Get dataset paths - training/data/ is relative to project root
-    repo_root = Path(__file__).parent.parent.parent
+    repo_root = Path(__file__).parent
     train_path = str(repo_root / "training" / "data" / "train.jsonl")
     test_path = str(repo_root / "training" / "data" / "test.jsonl")
     print(f"Looking for dataset at: {train_path}")
@@ -206,7 +206,7 @@ def check_setup(config_path):
     
     # Check dataset files
     print("\n3. Dataset files:")
-    repo_root = Path(__file__).parent.parent.parent  # Go up to project root
+    repo_root = Path(__file__).parent
     train_path = repo_root / "training" / "data" / "train.jsonl"
     test_path = repo_root / "training" / "data" / "test.jsonl"
     

training/scripts/run_with_accelerate.sh

@@ -1,11 +0,0 @@
-#!/bin/bash
-# Run training with distributed setup
-
-set -e
-
-# Install accelerate if not present
-pip install -q accelerate
-
-# Launch with accelerate
-# Note: Use the dedicated train-on-this-server.sh instead
-echo "See train-on-this-server.sh for full setup"