add model

.ipynb_checkpoints/convert-from-malaya-checkpoint.ipynb

@@ -0,0 +1,184 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "58d45708",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from transformers import XLNetTokenizer, XLNetModel, XLNetConfig, AutoTokenizer, AutoModelWithLMHead, pipeline"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "e0314358",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "model.ckpt-320000.data-00000-of-00001  model.ckpt-320000.meta\r\n",
+      "model.ckpt-320000.index\r\n"
+     ]
+    }
+   ],
+   "source": [
+    "# !tar -zxf xlnet-large-2021-09-06.tar.gz\n",
+    "# !rm xlnet-large-2021-09-06.tar.gz\n",
+    "!ls xlnet-large"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "59d2c8b5",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# !wget https://raw.githubusercontent.com/huseinzol05/malaya/master/pretrained-model/xlnet/tokenizer/sp10m.cased.v9.vocab\n",
+    "# !wget https://raw.githubusercontent.com/huseinzol05/malaya/master/pretrained-model/xlnet/tokenizer/sp10m.cased.v9.model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "f35e09f4",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "('./tokenizer_config.json',\n",
+       " './special_tokens_map.json',\n",
+       " './spiece.model',\n",
+       " './added_tokens.json')"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "tokenizer = XLNetTokenizer('sp10m.cased.v9.model', do_lower_case = False)\n",
+    "tokenizer.save_pretrained('./')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "4438ff5c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import json\n",
+    "\n",
+    "config = {\n",
+    "    \"d_head\": 64,\n",
+    "    \"d_inner\": 4096,\n",
+    "    \"d_model\": 1024,\n",
+    "    \"ff_activation\": \"gelu\",\n",
+    "    \"n_head\": 16,\n",
+    "    \"n_layer\": 20,\n",
+    "    \"n_token\": 32000,\n",
+    "    \"untie_r\": True\n",
+    "}\n",
+    "\n",
+    "with open('config.json', 'w') as fopen:\n",
+    "    json.dump(config, fopen)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "a265f23c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# !transformers-cli convert --model_type xlnet \\\n",
+    "#   --tf_checkpoint xlnet-large/model.ckpt-320000 \\\n",
+    "#   --config config.json \\\n",
+    "#   --pytorch_dump_output ./"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "22b94055",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "config = XLNetConfig(f'./config.json')\n",
+    "config.vocab_size = 32000\n",
+    "config.d_inner = 4096\n",
+    "config.d_model = 1024\n",
+    "config.n_head = 16\n",
+    "config.n_layer = 20"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "17c6d447",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Some weights of the model checkpoint at ./ were not used when initializing XLNetModel: ['lm_loss.weight', 'lm_loss.bias']\n",
+      "- This IS expected if you are initializing XLNetModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
+      "- This IS NOT expected if you are initializing XLNetModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
+     ]
+    }
+   ],
+   "source": [
+    "model = XLNetModel.from_pretrained('./', config = config)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "d0fc0138",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "tokenizer = XLNetTokenizer.from_pretrained('./',do_lower_case = False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "ec2c0661",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.save_pretrained('./')"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "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": "3.6.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

README.md

@@ -0,0 +1,32 @@
+---
+language: ms
+---
+
+# xlnet-large-bahasa-cased
+
+Pretrained XLNET large language model for Malay.
+
+## Pretraining Corpus
+
+`xlnet-large-bahasa-cased` model was pretrained on ~1.4 Billion words. Below is list of data we trained on,
+
+1. [cleaned local texts](https://github.com/huseinzol05/malay-dataset/tree/master/dumping/clean).
+2. [translated The Pile](https://github.com/huseinzol05/malay-dataset/tree/master/corpus/pile).
+
+## Pretraining details
+
+- All steps can reproduce from here, [Malaya/pretrained-model/xlnet](https://github.com/huseinzol05/Malaya/tree/master/pretrained-model/xlnet).
+
+## Load Pretrained Model
+
+You can use this model by installing `torch` or `tensorflow` and Huggingface library `transformers`. And you can use it directly by initializing it like this:  
+
+```python
+from transformers import XLNetModel, XLNetTokenizer
+
+model = XLNetModel.from_pretrained('malay-huggingface/xlnet-large-bahasa-cased')
+tokenizer = XLNetTokenizer.from_pretrained(
+    'malay-huggingface/xlnet-large-bahasa-cased',
+    do_lower_case = False,
+)
+```

config.json

@@ -0,0 +1,37 @@
+{
+  "_name_or_path": "./",
+  "architectures": [
+    "XLNetModel"
+  ],
+  "attn_type": "bi",
+  "bi_data": false,
+  "bos_token_id": 1,
+  "clamp_len": -1,
+  "d_head": 64,
+  "d_inner": 4096,
+  "d_model": 1024,
+  "dropout": 0.1,
+  "end_n_top": 5,
+  "eos_token_id": 2,
+  "ff_activation": "gelu",
+  "initializer_range": 0.02,
+  "layer_norm_eps": 1e-12,
+  "mem_len": 512,
+  "model_type": "xlnet",
+  "n_head": 16,
+  "n_layer": 20,
+  "pad_token_id": 5,
+  "reuse_len": null,
+  "same_length": false,
+  "start_n_top": 5,
+  "summary_activation": "tanh",
+  "summary_last_dropout": 0.1,
+  "summary_type": "last",
+  "summary_use_proj": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.10.0",
+  "untie_r": true,
+  "use_mems_eval": true,
+  "use_mems_train": false,
+  "vocab_size": 32000
+}

convert-from-malaya.ipynb

@@ -0,0 +1,184 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "58d45708",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from transformers import XLNetTokenizer, XLNetModel, XLNetConfig, AutoTokenizer, AutoModelWithLMHead, pipeline"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "e0314358",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "model.ckpt-320000.data-00000-of-00001  model.ckpt-320000.meta\r\n",
+      "model.ckpt-320000.index\r\n"
+     ]
+    }
+   ],
+   "source": [
+    "# !tar -zxf xlnet-large-2021-09-06.tar.gz\n",
+    "# !rm xlnet-large-2021-09-06.tar.gz\n",
+    "!ls xlnet-large"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "59d2c8b5",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# !wget https://raw.githubusercontent.com/huseinzol05/malaya/master/pretrained-model/xlnet/tokenizer/sp10m.cased.v9.vocab\n",
+    "# !wget https://raw.githubusercontent.com/huseinzol05/malaya/master/pretrained-model/xlnet/tokenizer/sp10m.cased.v9.model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "f35e09f4",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "('./tokenizer_config.json',\n",
+       " './special_tokens_map.json',\n",
+       " './spiece.model',\n",
+       " './added_tokens.json')"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "tokenizer = XLNetTokenizer('sp10m.cased.v9.model', do_lower_case = False)\n",
+    "tokenizer.save_pretrained('./')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "4438ff5c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import json\n",
+    "\n",
+    "config = {\n",
+    "    \"d_head\": 64,\n",
+    "    \"d_inner\": 4096,\n",
+    "    \"d_model\": 1024,\n",
+    "    \"ff_activation\": \"gelu\",\n",
+    "    \"n_head\": 16,\n",
+    "    \"n_layer\": 20,\n",
+    "    \"n_token\": 32000,\n",
+    "    \"untie_r\": True\n",
+    "}\n",
+    "\n",
+    "with open('config.json', 'w') as fopen:\n",
+    "    json.dump(config, fopen)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "a265f23c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# !transformers-cli convert --model_type xlnet \\\n",
+    "#   --tf_checkpoint xlnet-large/model.ckpt-320000 \\\n",
+    "#   --config config.json \\\n",
+    "#   --pytorch_dump_output ./"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "22b94055",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "config = XLNetConfig(f'./config.json')\n",
+    "config.vocab_size = 32000\n",
+    "config.d_inner = 4096\n",
+    "config.d_model = 1024\n",
+    "config.n_head = 16\n",
+    "config.n_layer = 20"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "17c6d447",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Some weights of the model checkpoint at ./ were not used when initializing XLNetModel: ['lm_loss.weight', 'lm_loss.bias']\n",
+      "- This IS expected if you are initializing XLNetModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
+      "- This IS NOT expected if you are initializing XLNetModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
+     ]
+    }
+   ],
+   "source": [
+    "model = XLNetModel.from_pretrained('./', config = config)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "d0fc0138",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "tokenizer = XLNetTokenizer.from_pretrained('./',do_lower_case = False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "ec2c0661",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.save_pretrained('./')"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "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": "3.6.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1b26afddfeb5ebe45d8a6ebbb0acd8e39367cf1f3e32d97e4419552bdd40c30
+size 1222853333

sp10m.cased.v9.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0caa407e56cc60c5a74aad6f90d3c3a7d25231b0c0d92211df9f4c7442b839a
+size 778744

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "sep_token": "<sep>", "pad_token": "<pad>", "cls_token": "<cls>", "mask_token": {"content": "<mask>", "single_word": false, "lstrip": true, "rstrip": false, "normalized": true}, "additional_special_tokens": ["<eop>", "<eod>"]}

spiece.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0caa407e56cc60c5a74aad6f90d3c3a7d25231b0c0d92211df9f4c7442b839a
+size 778744

tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": false, "remove_space": true, "keep_accents": false, "bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "sep_token": "<sep>", "pad_token": "<pad>", "cls_token": "<cls>", "mask_token": {"content": "<mask>", "single_word": false, "lstrip": true, "rstrip": false, "normalized": true, "__type": "AddedToken"}, "additional_special_tokens": ["<eop>", "<eod>"], "sp_model_kwargs": {}, "tokenizer_class": "XLNetTokenizer"}