Skip to content

Instantly share code, notes, and snippets.

@nymwa

nymwa/bert_nimi_prediction.ipynb

Created January 28, 2022 18:06
Show Gist options
  • Save nymwa/766b0255aa40c776d9608752616c5791 to your computer and use it in GitHub Desktop.
Save nymwa/766b0255aa40c776d9608752616c5791 to your computer and use it in GitHub Desktop.
Download ZIP
bert_nimi_prediction.ipynb
Display the source blob
Display the rendered blob
Raw
bert_nimi_prediction.ipynb
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"name": "bert_nimi_prediction.ipynb",
"provenance": [],
"authorship_tag": "ABX9TyNlHcTDS1Fbq/X7ZgfSTUPD",
"include_colab_link": true
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
},
"language_info": {
"name": "python"
},
"accelerator": "GPU"
},
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "view-in-github",
"colab_type": "text"
},
"source": [
"<a href=\"https://colab.research.google.com/gist/nymwa/766b0255aa40c776d9608752616c5791/bert_nimi_prediction.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
]
},
{
"cell_type": "markdown",
"source": [
"↑ push \"Open in Colab\"\n",
"\n",
"↓ run block (▶) one by one "
],
"metadata": {
"id": "1MaK3FCQnUC9"
}
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "EyABZ4g-eNpA",
"outputId": "12abc96a-8274-4d96-f231-0654c4491655"
},
"outputs": [
{
"output_type": "stream",
"name": "stderr",
"text": [
"Cloning into 'ilonimi'...\n",
"Note: checking out '5224a4c36c5a636afc1c925be198d0f8c17d435a'.\n",
"\n",
"You are in 'detached HEAD' state. You can look around, make experimental\n",
"changes and commit them, and you can discard any commits you make in this\n",
"state without impacting any branches by performing another checkout.\n",
"\n",
"If you want to create a new branch to retain commits you create, you may\n",
"do so (now or later) by using -b with the checkout command again. Example:\n",
"\n",
" git checkout -b <new-branch-name>\n",
"\n",
"HEAD is now at 5224a4c add ignore set\n",
"Cloning into 'ponask'...\n",
"Note: checking out 'fadc92ddca0422009e10d92d22aabbb468ab88dc'.\n",
"\n",
"You are in 'detached HEAD' state. You can look around, make experimental\n",
"changes and commit them, and you can discard any commits you make in this\n",
"state without impacting any branches by performing another checkout.\n",
"\n",
"If you want to create a new branch to retain commits you create, you may\n",
"do so (now or later) by using -b with the checkout command again. Example:\n",
"\n",
" git checkout -b <new-branch-name>\n",
"\n",
"HEAD is now at fadc92d add\n",
"Downloading...\n",
"From: https://drive.google.com/uc?id=1d3XjRHEp11jAbNiSU7mRt-f8zpu3BVxK\n",
"To: /content/bert_big.pt\n",
"\r 0%| | 0.00/153M [00:00<?, ?B/s]\r 6%|▌ | 8.91M/153M [00:00<00:04, 34.7MB/s]\r 12%|█▏ | 17.8M/153M [00:00<00:02, 53.2MB/s]\r 18%|█▊ | 26.7M/153M [00:00<00:01, 65.0MB/s]\r 29%|██▉ | 44.6M/153M [00:00<00:01, 101MB/s] \r 39%|███▉ | 59.2M/153M [00:00<00:00, 93.9MB/s]\r 46%|████▌ | 70.3M/153M [00:00<00:00, 95.3MB/s]\r 55%|█████▍ | 83.4M/153M [00:00<00:00, 104MB/s] \r 62%|██████▏ | 94.9M/153M [00:01<00:00, 106MB/s]\r 71%|███████ | 109M/153M [00:01<00:00, 113MB/s] \r 79%|███████▉ | 121M/153M [00:01<00:00, 107MB/s]\r 87%|████████▋ | 132M/153M [00:01<00:00, 93.9MB/s]\r 93%|█████████▎| 142M/153M [00:01<00:00, 90.6MB/s]\r100%|██████████| 153M/153M [00:01<00:00, 92.6MB/s]\n"
]
}
],
"source": [
"%%bash\n",
"\n",
"git clone https://github.com/nymwa/ilonimi.git\n",
"mv ilonimi ilonimi.d\n",
"cd ilonimi.d\n",
"git checkout 5224a4c36c5a636afc1c925be198d0f8c17d435a\n",
"mv ilonimi ../\n",
"cd ..\n",
"\n",
"git clone https://github.com/nymwa/ponask.git\n",
"mv ponask ponask.d\n",
"cd ponask.d\n",
"git checkout fadc92ddca0422009e10d92d22aabbb468ab88dc\n",
"mv ponask ../\n",
"cd ..\n",
"\n",
"gdown --id \"1d3XjRHEp11jAbNiSU7mRt-f8zpu3BVxK\""
]
},
{
"cell_type": "code",
"source": [
"from ilonimi import Normalizer, Tokenizer, Splitter\n",
"import torch\n",
"import torch\n",
"from torch.nn.utils.rnn import pad_sequence as pad\n",
"from ponask.vocab import Vocab\n",
"from ponask.bert import BERT\n",
"from ponask.batch import Batch\n",
"from tabulate import tabulate"
],
"metadata": {
"id": "gUKBYZ8kjrPH"
},
"execution_count": 2,
"outputs": []
},
{
"cell_type": "code",
"source": [
"def load_vocab(path):\n",
" with open(path) as f:\n",
" tokens = [x.strip() for x in f]\n",
" vocab = Vocab(tokens)\n",
" return vocab\n",
"\n",
"vocab = load_vocab('ponask.d/work/vocab.txt')\n",
"normalizer = Normalizer()\n",
"tokenizer = Tokenizer(\n",
" convert_unk = True,\n",
" convert_number = False,\n",
" convert_proper = False,\n",
" ignore_set = {'<msk>'})\n",
"splitter = Splitter(sharp = False)\n",
"\n",
"model = BERT(len(vocab), 512, 8, 2048, 0.3, 0.2, 0.2, 12, 64)\n",
"model.load_state_dict(torch.load('bert_big.pt', map_location = 'cpu'))\n",
"model = model.cuda()\n",
"model.eval()"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "VGD7_wGbeUT5",
"outputId": "9ad1bf64-a94b-40b2-f494-2406f174c009"
},
"execution_count": 3,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"BERT(\n",
" (embedding): TransformerEmbedding(\n",
" (token_embedding): Embedding(241, 512, padding_idx=0)\n",
" (position_embedding): SinusoidalPositionalEmbedding(\n",
" (embedding): Embedding(64, 512)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (encoder): TransformerEncoder(\n",
" (layers): ModuleList(\n",
" (0): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (1): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (2): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (3): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (4): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (5): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (6): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (7): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (8): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (9): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (10): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" (11): TransformerEncoderLayer(\n",
" (self_attn_layer): SelfAttentionSubLayer(\n",
" (self_attn): MultiheadAttention(\n",
" (out_proj): NonDynamicallyQuantizableLinear(in_features=512, out_features=512, bias=True)\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.3, inplace=False)\n",
" )\n",
" (feed_forward_layer): FeedForwardSubLayer(\n",
" (linear1): Linear(in_features=512, out_features=2048, bias=True)\n",
" (linear2): Linear(in_features=2048, out_features=512, bias=True)\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" (dropout): Dropout(p=0.2, inplace=False)\n",
" (activation): GELU()\n",
" )\n",
" )\n",
" )\n",
" (norm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
" )\n",
" (fc): Linear(in_features=512, out_features=241, bias=True)\n",
")"
]
},
"metadata": {},
"execution_count": 3
}
]
},
{
"cell_type": "code",
"source": [
"def infer(x):\n",
" x = x.strip()\n",
" x = normalizer(x)\n",
" x = tokenizer(x)\n",
" x = splitter(x)\n",
" x = x.split()\n",
" x = [vocab.bos] + [vocab(w) for w in x] + [vocab.eos]\n",
" mask_indices = [i for i, w in enumerate(x) if w == vocab.msk]\n",
" x = torch.tensor(x)\n",
" x = pad([x], padding_value = vocab.pad)\n",
" x = Batch(x)\n",
" x = x.cuda()\n",
" with torch.no_grad():\n",
" x = model(x)\n",
" x = x.transpose(1, 0)[0]\n",
" for i in mask_indices:\n",
" prob = torch.softmax(x[i], dim = -1)\n",
" values, indices = torch.topk(prob, 10)\n",
" tab = [\n",
" ['word'] + [vocab[index] for index in indices],\n",
" ['prediction (%)'] + ['{:.2f}'.format(value * 100) for value in values]]\n",
" tab = tabulate(tab, tablefmt = 'psql')\n",
" print('index: {}'.format(i))\n",
" print(tab)"
],
"metadata": {
"id": "tiEM8SV5kJei"
},
"execution_count": 4,
"outputs": []
},
{
"cell_type": "markdown",
"source": [
"You can guess <msk> by running \"infer('sentence')\""
],
"metadata": {
"id": "Iv-BqDcmnbVg"
}
},
{
"cell_type": "code",
"source": [
"infer('mi pilin akesi <msk> moli.')"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "-aNzVn4FerF8",
"outputId": "4fca95a1-f22a-4b79-f3ef-5da4f5c2de37"
},
"execution_count": 5,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"index: 4\n",
"+----------------+-------+------+------+------+------+------+------+------+------+------+\n",
"| word | tan | tawa | li | jan | ala | e | mi | sina | lon | pi |\n",
"| prediction (%) | 99.84 | 0.09 | 0.03 | 0.01 | 0.01 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |\n",
"+----------------+-------+------+------+------+------+------+------+------+------+------+\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"infer('mi sitelen <msk> lipu.')\n",
"infer('mi sitelen e soweli <msk> lipu.')"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "SFMbxTmHkfws",
"outputId": "41914179-6401-49b2-c70d-7dc265699698"
},
"execution_count": 6,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"index: 3\n",
"+----------------+-------+------+------+------+---------+------+------+------+------+--------+\n",
"| word | e | lon | insa | tawa | kepeken | tan | li | ala | ma | palisa |\n",
"| prediction (%) | 96.43 | 3.38 | 0.09 | 0.04 | 0.03 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |\n",
"+----------------+-------+------+------+------+---------+------+------+------+------+--------+\n",
"index: 5\n",
"+----------------+-------+------+------+---------+------+------+--------+------+------+------+\n",
"| word | lon | tomo | e | kepeken | tawa | luka | palisa | sona | toki | musi |\n",
"| prediction (%) | 99.32 | 0.28 | 0.14 | 0.09 | 0.03 | 0.03 | 0.02 | 0.02 | 0.01 | 0.01 |\n",
"+----------------+-------+------+------+---------+------+------+--------+------+------+------+\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"infer('toki <msk> li toki <msk>.')"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "IA5-jaKQl1jr",
"outputId": "7256cd3c-7dc5-4de9-9807-918f7ab8aa44"
},
"execution_count": 7,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"index: 2\n",
"+----------------+-------+------+------+------+------+------+------+------+------+------+\n",
"| word | pona | mute | ni | mi | ona | lili | nasa | sina | ala | mama |\n",
"| prediction (%) | 86.69 | 4.78 | 3.73 | 0.74 | 0.65 | 0.64 | 0.60 | 0.37 | 0.37 | 0.36 |\n",
"+----------------+-------+------+------+------+------+------+------+------+------+------+\n",
"index: 5\n",
"+----------------+-------+-------+------+------+-------+------+------+------+------+------+\n",
"| word | pona | ala | lon | mute | utala | taso | nasa | wawa | musi | ante |\n",
"| prediction (%) | 59.88 | 18.91 | 7.75 | 3.47 | 1.97 | 1.74 | 1.43 | 1.15 | 0.64 | 0.51 |\n",
"+----------------+-------+-------+------+------+-------+------+------+------+------+------+\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"infer('jan Lase li wile <msk>.')\n",
"infer('jan Lisa li wile <msk>.')\n",
"infer('jan Nanko li wile <msk>.')\n",
"infer('jan Kuli li wile <msk>.')\n",
"infer('jan Kijolo li wile <msk>.')"
],
"metadata": {
"id": "MV9mCl0_nzdI",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "1f0a8213-c682-416b-d0bf-941434812b6f"
},
"execution_count": 8,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"index: 6\n",
"+----------------+-------+-------+---------+------+-------+------+-------+------+------+------+\n",
"| word | toki | lape | sitelen | moku | utala | pali | lukin | tawa | kama | esun |\n",
"| prediction (%) | 31.40 | 18.28 | 11.57 | 9.99 | 9.00 | 8.04 | 2.45 | 1.74 | 1.39 | 1.16 |\n",
"+----------------+-------+-------+---------+------+-------+------+-------+------+------+------+\n",
"index: 6\n",
"+----------------+-------+-------+-------+-------+------+------+------+-------+------+------+\n",
"| word | unpa | lape | moli | lukin | tawa | esun | moku | utala | kama | pona |\n",
"| prediction (%) | 36.80 | 13.54 | 11.68 | 7.55 | 6.26 | 5.71 | 5.05 | 2.43 | 2.12 | 1.20 |\n",
"+----------------+-------+-------+-------+-------+------+------+------+-------+------+------+\n",
"index: 6\n",
"+----------------+-------+------+---------+------+-------+------+------+------+------+------+\n",
"| word | lape | moku | sitelen | pali | lukin | toki | esun | moli | mute | kama |\n",
"| prediction (%) | 75.70 | 6.38 | 2.87 | 2.23 | 2.02 | 1.66 | 1.65 | 1.42 | 0.91 | 0.82 |\n",
"+----------------+-------+------+---------+------+-------+------+------+------+------+------+\n",
"index: 6\n",
"+----------------+-------+-------+------+------+-------+-------+------+------+------+-------+\n",
"| word | unpa | lape | moku | toki | utala | akesi | pali | tawa | esun | lukin |\n",
"| prediction (%) | 78.09 | 12.42 | 1.63 | 1.37 | 1.04 | 0.73 | 0.72 | 0.70 | 0.56 | 0.51 |\n",
"+----------------+-------+-------+------+------+-------+-------+------+------+------+-------+\n",
"index: 7\n",
"+----------------+-------+-------+------+------+-------+------+------+------+------+------+\n",
"| word | lukin | lape | awen | moli | utala | ike | musi | mute | wan | pona |\n",
"| prediction (%) | 50.91 | 18.49 | 7.88 | 6.90 | 5.71 | 1.59 | 1.21 | 1.21 | 0.74 | 0.72 |\n",
"+----------------+-------+-------+------+------+-------+------+------+------+------+------+\n"
]
}
]
},
{
"cell_type": "code",
"source": [
""
],
"metadata": {
"id": "9FvVB4jV3AKb"
},
"execution_count": 8,
"outputs": []
}
]
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment