{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "ee8cfec1",
   "metadata": {
    "papermill": {
     "duration": 0.001903,
     "end_time": "2023-08-23T19:20:40.353029",
     "exception": false,
     "start_time": "2023-08-23T19:20:40.351126",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Model Init\n",
    "\n",
    "Test that the model init code, runs without issues\n",
    "\n",
    "**L6-D512 model with**\n",
    "- Layer count: 6\n",
    "- Embed size: 512"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "7713f8b0",
   "metadata": {
    "notebookRunGroups": {
     "groupValue": ""
    },
    "papermill": {
     "duration": 0.001217,
     "end_time": "2023-08-23T19:20:40.355733",
     "exception": false,
     "start_time": "2023-08-23T19:20:40.354516",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## Preparing the init model and test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "54684bc2",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-08-23T19:20:40.360445Z",
     "iopub.status.busy": "2023-08-23T19:20:40.359491Z",
     "iopub.status.idle": "2023-08-23T19:20:41.112004Z",
     "shell.execute_reply": "2023-08-23T19:20:41.111011Z"
    },
    "papermill": {
     "duration": 0.757092,
     "end_time": "2023-08-23T19:20:41.114108",
     "exception": false,
     "start_time": "2023-08-23T19:20:40.357016",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# First lets setup the various directories\n",
    "!mkdir -p ../../model/\n",
    "!mkdir -p ../../datapath/\n",
    "!mkdir -p ../../checkpoint/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "80da2afe",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-08-23T19:20:41.119086Z",
     "iopub.status.busy": "2023-08-23T19:20:41.118584Z",
     "iopub.status.idle": "2023-08-23T19:20:50.840612Z",
     "shell.execute_reply": "2023-08-23T19:20:50.839437Z"
    },
    "papermill": {
     "duration": 9.727624,
     "end_time": "2023-08-23T19:20:50.843488",
     "exception": false,
     "start_time": "2023-08-23T19:20:41.115864",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2023-08-23 19:20:45,786] [INFO] [real_accelerator.py:133:get_accelerator] Setting ds_accelerator to cuda (auto detect)\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[RWKV.model] Running RWKV model using 'torch-jit' with torch '2.0.1+cu118'\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---- Initializing model ----\r\n",
      "No of layers: 6\r\n",
      "Embedding size: 512\r\n",
      "Output model path: ../model/L6-D512-neox-init.pth\r\n",
      "Vocab size: 50277\r\n",
      "Note: this process takes a significant time (and ram) for large models\r\n",
      "---- ----- ----\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using /root/.cache/torch_extensions/py310_cu118 as PyTorch extensions root...\r\n",
      "Detected CUDA files, patching ldflags\r\n",
      "Emitting ninja build file /root/.cache/torch_extensions/py310_cu118/wkv_1_bf16/build.ninja...\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Building extension module wkv_1_bf16...\r\n",
      "Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ninja: no work to do.\r\n",
      "Loading extension module wkv_1_bf16...\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[RWKV.model]: Finished initial model load\r\n",
      "50277 512   -0.0001 emb.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.0.att.key.weight\r\n",
      "512   512   1.0  blocks.0.att.value.weight\r\n",
      "512   512   0    blocks.0.att.receptance.weight\r\n",
      "512   512   0    blocks.0.att.output.weight\r\n",
      "2048  512   1.0  blocks.0.ffn.key.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.0.ffn.receptance.weight\r\n",
      "512   2048  0    blocks.0.ffn.value.weight\r\n",
      "512   512   0    blocks.1.att.key.weight\r\n",
      "512   512   1.0  blocks.1.att.value.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.1.att.receptance.weight\r\n",
      "512   512   0    blocks.1.att.output.weight\r\n",
      "2048  512   1.0  blocks.1.ffn.key.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.1.ffn.receptance.weight\r\n",
      "512   2048  0    blocks.1.ffn.value.weight\r\n",
      "512   512   0    blocks.2.att.key.weight\r\n",
      "512   512   1.0  blocks.2.att.value.weight\r\n",
      "512   512   0    blocks.2.att.receptance.weight\r\n",
      "512   512   0    blocks.2.att.output.weight\r\n",
      "2048  512   1.0  blocks.2.ffn.key.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.2.ffn.receptance.weight\r\n",
      "512   2048  0    blocks.2.ffn.value.weight\r\n",
      "512   512   0    blocks.3.att.key.weight\r\n",
      "512   512   1.0  blocks.3.att.value.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.3.att.receptance.weight\r\n",
      "512   512   0    blocks.3.att.output.weight\r\n",
      "2048  512   1.0  blocks.3.ffn.key.weight\r\n",
      "512   512   0    blocks.3.ffn.receptance.weight\r\n",
      "512   2048  0    blocks.3.ffn.value.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.4.att.key.weight\r\n",
      "512   512   1.0  blocks.4.att.value.weight\r\n",
      "512   512   0    blocks.4.att.receptance.weight\r\n",
      "512   512   0    blocks.4.att.output.weight\r\n",
      "2048  512   1.0  blocks.4.ffn.key.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.4.ffn.receptance.weight\r\n",
      "512   2048  0    blocks.4.ffn.value.weight\r\n",
      "512   512   0    blocks.5.att.key.weight\r\n",
      "512   512   1.0  blocks.5.att.value.weight\r\n",
      "512   512   0    blocks.5.att.receptance.weight\r\n",
      "512   512   0    blocks.5.att.output.weight\r\n",
      "2048  512   1.0  blocks.5.ffn.key.weight\r\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "512   512   0    blocks.5.ffn.receptance.weight\r\n",
      "512   2048  0    blocks.5.ffn.value.weight\r\n",
      "50277 512   0.5  head.weight\r\n"
     ]
    }
   ],
   "source": [
    "# Lets initialized the L6-D512 model with the init_model.py code\n",
    "!cd ../../RWKV-v4neo/ && python3 init_model.py --n_layer 6 --n_embd 512 --vocab_size neox ../model/L6-D512-neox-init.pth"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "rwkv-exp",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.12"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 12.198584,
   "end_time": "2023-08-23T19:20:51.170107",
   "environment_variables": {},
   "exception": null,
   "input_path": "/actions-runner/_work/RWKV-infctx-trainer/RWKV-infctx-trainer/notebook/trainer-v4-unit-test/model-init.ipynb",
   "output_path": "/actions-runner/_work/RWKV-infctx-trainer/RWKV-infctx-trainer/output/trainer-v4-unit-test/model-init.ipynb",
   "parameters": {},
   "start_time": "2023-08-23T19:20:38.971523",
   "version": "2.4.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}