docs: updates to absolute path following comments

Author: MStokluska

examples/fine-tuning/lora/lora_sft-distributed.ipynb

@@ -215,7 +215,7 @@
         "print(f\"Training examples: {len(train_data)}\")\n",
         "\n",
         "# Save training data to JSONL format\n",
-        "OUTPUT_DIR = Path(f\"{PVC_PATH}/lora_text_sql_output\")\n",
+        "OUTPUT_DIR = Path(f\"/opt/app-root/src/{PVC_PATH}/lora_text_sql_output\")\n",
         "OUTPUT_DIR.mkdir(parents=True, exist_ok=True)\n",
         "\n",
         "training_file = OUTPUT_DIR / \"train_data.jsonl\"\n",
@@ -227,7 +227,7 @@
         "print(f\"Training data saved to: {training_file}\")\n",
         "print(f\"File size: {training_file.stat().st_size / 1024:.1f} KB\")\n",
         "\n",
-        "data_path = f\"{PVC_PATH}/lora_text_sql_output/train_data.jsonl\"\n",
+        "data_path = f\"/opt/app-root/src/{PVC_PATH}/lora_text_sql_output/train_data.jsonl\"\n",
         "print(data_path)"
       ],
       "execution_count": null,
@@ -473,7 +473,7 @@
         "\n",
         "import torch\n",
         "\n",
-        "CHECKPOINTS_DIR = f\"{PVC_PATH}/checkpoints\"\n",
+        "CHECKPOINTS_DIR = f\"/opt/app-root/src/{PVC_PATH}/checkpoints\"\n",
         "checkpoint_dirs = sorted(\n",
         "    glob.glob(os.path.join(CHECKPOINTS_DIR, \"checkpoint-*\")), key=os.path.getctime\n",
         ")\n",

examples/fine-tuning/osft/osft-distributed.ipynb

@@ -354,7 +354,7 @@
         "print(f\"Subset size: {len(subset_data)}\")\n",
         "\n",
         "# Define the output file path\n",
-        "output_dir = f\"{PVC_PATH}/table-gpt-data/train\"\n",
+        "output_dir = f\"/opt/app-root/src/{PVC_PATH}/table-gpt-data/train\"\n",
         "output_file = f\"{output_dir}/train_All_5000.jsonl\"\n",
         "\n",
         "# Save the subset to a JSONL file\n",

examples/fine-tuning/sft/sft-distributed.ipynb

@@ -258,16 +258,16 @@
         "\n",
         "# Save the subset to a JSONL file\n",
         "\n",
-        "os.makedirs(\"/opt/app-root/src/shared/table-gpt-data/train\", exist_ok=True)\n",
+        "os.makedirs(f\"/opt/app-root/src/{PVC_PATH}/table-gpt-data/train\", exist_ok=True)\n",
         "\n",
         "with open(\n",
-        "    \"/opt/app-root/src/shared/table-gpt-data/train/train_All_5000.jsonl\", \"w\"\n",
+        "    f\"/opt/app-root/src/{PVC_PATH}/table-gpt-data/train/train_All_5000.jsonl\", \"w\"\n",
         ") as f:\n",
         "    for example in subset_data:\n",
         "        f.write(json.dumps(example) + \"\\n\")\n",
         "\n",
         "print(\n",
-        "    \"Subset saved to /opt/app-root/src/shared/table-gpt-data/train/train_All_5000.jsonl\"\n",
+        "    f\"Subset saved to /opt/app-root/src/{PVC_PATH}/table-gpt-data/train/train_All_5000.jsonl\"\n",
         ")"
       ],
       "execution_count": null,
@@ -280,7 +280,7 @@
       "source": [
         "## Downloading the Qwen2.5-1.5B model\n",
         "\n",
-        "We need to download a copy of the model to the shared storage directory in `/opt/app-root/src/shared/`. We can do this with the `hf` CLI. "
+        "We need to download a copy of the model to the shared storage directory. We can do this with the `hf` CLI. "
       ],
       "id": "774c386d-94e9-4b1c-a084-2c6a51f63fce"
     },
@@ -393,7 +393,7 @@
       "source": [
         "model_path = download_model_snapshot(\n",
         "    model_id=\"Qwen/Qwen2.5-1.5B-Instruct\",\n",
-        "    output_dir=\"/opt/app-root/src/shared/Qwen/Qwen2.5-1.5B-Instruct\",\n",
+        "    output_dir=f\"/opt/app-root/src/{PVC_PATH}/Qwen/Qwen2.5-1.5B-Instruct\",\n",
         "    token=\"\",\n",
         ")"
       ],