Upload folder using huggingface_hub

.ipynb_checkpoints/README-checkpoint.md

@@ -0,0 +1,101 @@
+---
+language:
+- en
+- vi
+tags:
+- translation
+license: cc-by-4.0
+datasets:
+- quickmt/quickmt-train.vi-en
+model-index:
+- name: quickmt-en-vi
+  results:
+  - task:
+      name: Translation eng-vie
+      type: translation
+      args: eng-vie
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: eng_Latn vie_Latn devtest
+    metrics:
+    - name: BLEU
+      type: bleu
+      value: 43.68
+    - name: CHRF
+      type: chrf
+      value: 60.75
+    - name: COMET
+      type: comet
+      value: 87.52
+---
+
+
+# `quickmt-en-vi` Neural Machine Translation Model 
+
+`quickmt-en-vi` is a reasonably fast and reasonably accurate neural machine translation model for translation from `en` into `vi`.
+
+
+## Model Information
+
+* Trained using [`eole`](https://github.com/eole-nlp/eole)
+* 195M parameter transformer 'big' with 8 encoder layers and 2 decoder layers
+* 20k sentencepiece vocabularies
+* Exported for fast inference to [CTranslate2](https://github.com/OpenNMT/CTranslate2) format
+* Training data: https://huggingface.co/datasets/quickmt/quickmt-train.fa-en/tree/main
+
+See the `eole` model configuration in this repository for further details and the `eole-model` for the raw `eole` (pytorch) model.
+
+
+## Usage with `quickmt`
+
+You must install the Nvidia cuda toolkit first, if you want to do GPU inference.
+
+Next, install the `quickmt` python library and download the model:
+
+```bash
+git clone https://github.com/quickmt/quickmt.git
+pip install ./quickmt/
+
+quickmt-model-download quickmt/quickmt-en-vi ./quickmt-en-vi
+```
+
+Finally use the model in python:
+
+```python
+from quickmt import Translator
+
+# Auto-detects GPU, set to "cpu" to force CPU inference
+t = Translator("./quickmt-en-vi/", device="auto")
+
+# Translate - set beam size to 5 for higher quality (but slower speed)
+sample_text = 'Dr. Ehud Ur, professor of medicine at Dalhousie University in Halifax, Nova Scotia and chair of the clinical and scientific division of the Canadian Diabetes Association cautioned that the research is still in its early days.'
+t(sample_text, beam_size=5)
+```
+
+> 'Tiến sĩ Ehud Ur, giáo sư y khoa tại Đại học Dalhousie ở Halifax, Nova Scotia và chủ tịch bộ phận lâm sàng và khoa học của Hiệp hội Tiểu đường Canada cảnh báo rằng nghiên cứu vẫn còn trong những ngày đầu.'
+```python
+# Get alternative translations by sampling
+# You can pass any cTranslate2 `translate_batch` arguments
+t([sample_text], sampling_temperature=1.2, beam_size=1, sampling_topk=50, sampling_topp=0.9)
+```
+
+> 'Tiến sĩ Ehud Ur, Giáo sư Y tại Đại học Dalhousie ở Halifax, Nova Scotia và là chủ tịch bộ phận lâm sàng và khoa học của Hiệp hội Tiểu đường Canada đã cảnh báo rằng nghiên cứu mới này vẫn còn trong những ngày đầu.'
+
+
+The model is in `ctranslate2` format, and the tokenizers are `sentencepiece`, so you can use `ctranslate2` directly instead of through `quickmt`. It is also possible  to get this model to work with e.g. [LibreTranslate](https://libretranslate.com/) which also uses `ctranslate2` and `sentencepiece`.
+
+
+## Metrics
+
+`bleu` and `chrf2` are calculated with [sacrebleu](https://github.com/mjpost/sacrebleu) on the [Flores200 `devtest` test set](https://huggingface.co/datasets/facebook/flores) ("eng_Latn"->"vie_Latn"). `comet22` with the [`comet`](https://github.com/Unbabel/COMET) library and the [default model](https://huggingface.co/Unbabel/wmt22-comet-da). "Time (s)" is the time in seconds to translate (using `ctranslate2`) the flores-devtest dataset (1012 sentences) on an RTX 4070s GPU with batch size 32 (faster speed is possible with a larger batch size).
+
+|                                  |   bleu |   chrf2 |   comet22 |   Time (s) |
+|:---------------------------------|-------:|--------:|----------:|-----------:|
+| quickmt/quickmt-en-vi            |  43.68 |   60.75 |     87.52 |       1.5  |
+| Helsinki-NLP/opus-mt-en-vi       |  27.57 |   46.45 |     76.22 |       4.67 |
+| facebook/nllb-200-distilled-600M |  39.06 |   57.27 |     86.95 |      24.12 |
+| facebook/nllb-200-distilled-1.3B |  41.27 |   59.04 |     88.11 |      42.3  |
+| facebook/m2m100_418M             |  33.95 |   52.82 |     82.53 |      20.32 |
+| facebook/m2m100_1.2B             |  39.46 |   57.45 |     86.56 |      38.65 |
+

.ipynb_checkpoints/eole-config-checkpoint.yaml

@@ -0,0 +1,96 @@
+## IO
+save_data: data
+overwrite: True
+seed: 1234
+report_every: 100
+valid_metrics: ["BLEU"]
+tensorboard: true
+tensorboard_log_dir: tensorboard
+
+### Vocab
+src_vocab: en.eole.vocab
+tgt_vocab: vi.eole.vocab
+src_vocab_size: 20000
+tgt_vocab_size: 20000
+vocab_size_multiple: 8
+share_vocab: false
+n_sample: 0
+
+data:
+    corpus_1:
+        path_src: hf://quickmt/quickmt-train.vi-en/en
+        path_tgt: hf://quickmt/quickmt-train.vi-en/vi
+        path_sco: hf://quickmt/quickmt-train.vi-en/sco
+    valid:
+        path_src: valid.en
+        path_tgt: valid.vi
+
+transforms: [sentencepiece, filtertoolong]
+transforms_configs:
+  sentencepiece:
+    src_subword_model: "en.spm.model"
+    tgt_subword_model: "vi.spm.model"
+  filtertoolong:
+    src_seq_length: 256
+    tgt_seq_length: 256
+
+training:
+    # Run configuration
+    model_path: quickmt-en-vi-eole-model
+    #train_from: model
+    keep_checkpoint: 4
+    train_steps: 100000
+    save_checkpoint_steps: 5000
+    valid_steps: 5000
+    
+    # Train on a single GPU
+    world_size: 1
+    gpu_ranks: [0]
+
+    # Batching 10240
+    batch_type: "tokens"
+    batch_size: 8000
+    valid_batch_size: 4096
+    batch_size_multiple: 8
+    accum_count: [10]
+    accum_steps: [0]
+
+    # Optimizer & Compute
+    compute_dtype: "fp16"
+    optim: "adamw"
+    #use_amp: False
+    learning_rate: 2.0
+    warmup_steps: 4000
+    decay_method: "noam"
+    adam_beta2: 0.998
+
+    # Data loading
+    bucket_size: 128000
+    num_workers: 4
+    prefetch_factor: 32
+    
+    # Hyperparams
+    dropout_steps: [0]
+    dropout: [0.1]
+    attention_dropout: [0.1]
+    max_grad_norm: 0
+    label_smoothing: 0.1
+    average_decay: 0.0001
+    param_init_method: xavier_uniform
+    normalization: "tokens"
+
+model:
+    architecture: "transformer"
+    share_embeddings: false
+    share_decoder_embeddings: false
+    hidden_size: 1024
+    encoder:
+        layers: 8
+    decoder:
+        layers: 2
+    heads: 8
+    transformer_ff: 4096
+    embeddings:
+        word_vec_size: 1024
+        position_encoding_type: "SinusoidalInterleaved"
+

README.md

@@ -1,3 +1,101 @@
----
-license: cc-by-4.0
----
+---
+language:
+- en
+- vi
+tags:
+- translation
+license: cc-by-4.0
+datasets:
+- quickmt/quickmt-train.vi-en
+model-index:
+- name: quickmt-en-vi
+  results:
+  - task:
+      name: Translation eng-vie
+      type: translation
+      args: eng-vie
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: eng_Latn vie_Latn devtest
+    metrics:
+    - name: BLEU
+      type: bleu
+      value: 43.68
+    - name: CHRF
+      type: chrf
+      value: 60.75
+    - name: COMET
+      type: comet
+      value: 87.52
+---
+
+
+# `quickmt-en-vi` Neural Machine Translation Model 
+
+`quickmt-en-vi` is a reasonably fast and reasonably accurate neural machine translation model for translation from `en` into `vi`.
+
+
+## Model Information
+
+* Trained using [`eole`](https://github.com/eole-nlp/eole)
+* 195M parameter transformer 'big' with 8 encoder layers and 2 decoder layers
+* 20k sentencepiece vocabularies
+* Exported for fast inference to [CTranslate2](https://github.com/OpenNMT/CTranslate2) format
+* Training data: https://huggingface.co/datasets/quickmt/quickmt-train.fa-en/tree/main
+
+See the `eole` model configuration in this repository for further details and the `eole-model` for the raw `eole` (pytorch) model.
+
+
+## Usage with `quickmt`
+
+You must install the Nvidia cuda toolkit first, if you want to do GPU inference.
+
+Next, install the `quickmt` python library and download the model:
+
+```bash
+git clone https://github.com/quickmt/quickmt.git
+pip install ./quickmt/
+
+quickmt-model-download quickmt/quickmt-en-vi ./quickmt-en-vi
+```
+
+Finally use the model in python:
+
+```python
+from quickmt import Translator
+
+# Auto-detects GPU, set to "cpu" to force CPU inference
+t = Translator("./quickmt-en-vi/", device="auto")
+
+# Translate - set beam size to 5 for higher quality (but slower speed)
+sample_text = 'Dr. Ehud Ur, professor of medicine at Dalhousie University in Halifax, Nova Scotia and chair of the clinical and scientific division of the Canadian Diabetes Association cautioned that the research is still in its early days.'
+t(sample_text, beam_size=5)
+```
+
+> 'Tiến sĩ Ehud Ur, giáo sư y khoa tại Đại học Dalhousie ở Halifax, Nova Scotia và chủ tịch bộ phận lâm sàng và khoa học của Hiệp hội Tiểu đường Canada cảnh báo rằng nghiên cứu vẫn còn trong những ngày đầu.'
+```python
+# Get alternative translations by sampling
+# You can pass any cTranslate2 `translate_batch` arguments
+t([sample_text], sampling_temperature=1.2, beam_size=1, sampling_topk=50, sampling_topp=0.9)
+```
+
+> 'Tiến sĩ Ehud Ur, Giáo sư Y tại Đại học Dalhousie ở Halifax, Nova Scotia và là chủ tịch bộ phận lâm sàng và khoa học của Hiệp hội Tiểu đường Canada đã cảnh báo rằng nghiên cứu mới này vẫn còn trong những ngày đầu.'
+
+
+The model is in `ctranslate2` format, and the tokenizers are `sentencepiece`, so you can use `ctranslate2` directly instead of through `quickmt`. It is also possible  to get this model to work with e.g. [LibreTranslate](https://libretranslate.com/) which also uses `ctranslate2` and `sentencepiece`.
+
+
+## Metrics
+
+`bleu` and `chrf2` are calculated with [sacrebleu](https://github.com/mjpost/sacrebleu) on the [Flores200 `devtest` test set](https://huggingface.co/datasets/facebook/flores) ("eng_Latn"->"vie_Latn"). `comet22` with the [`comet`](https://github.com/Unbabel/COMET) library and the [default model](https://huggingface.co/Unbabel/wmt22-comet-da). "Time (s)" is the time in seconds to translate (using `ctranslate2`) the flores-devtest dataset (1012 sentences) on an RTX 4070s GPU with batch size 32 (faster speed is possible with a larger batch size).
+
+|                                  |   bleu |   chrf2 |   comet22 |   Time (s) |
+|:---------------------------------|-------:|--------:|----------:|-----------:|
+| quickmt/quickmt-en-vi            |  43.68 |   60.75 |     87.52 |       1.5  |
+| Helsinki-NLP/opus-mt-en-vi       |  27.57 |   46.45 |     76.22 |       4.67 |
+| facebook/nllb-200-distilled-600M |  39.06 |   57.27 |     86.95 |      24.12 |
+| facebook/nllb-200-distilled-1.3B |  41.27 |   59.04 |     88.11 |      42.3  |
+| facebook/m2m100_418M             |  33.95 |   52.82 |     82.53 |      20.32 |
+| facebook/m2m100_1.2B             |  39.46 |   57.45 |     86.56 |      38.65 |
+

config.json

@@ -0,0 +1,10 @@
+{
+  "add_source_bos": false,
+  "add_source_eos": false,
+  "bos_token": "<s>",
+  "decoder_start_token": "<s>",
+  "eos_token": "</s>",
+  "layer_norm_epsilon": 1e-06,
+  "multi_query_attention": false,
+  "unk_token": "<unk>"
+}

eole-config.yaml

@@ -0,0 +1,96 @@
+## IO
+save_data: data
+overwrite: True
+seed: 1234
+report_every: 100
+valid_metrics: ["BLEU"]
+tensorboard: true
+tensorboard_log_dir: tensorboard
+
+### Vocab
+src_vocab: en.eole.vocab
+tgt_vocab: vi.eole.vocab
+src_vocab_size: 20000
+tgt_vocab_size: 20000
+vocab_size_multiple: 8
+share_vocab: false
+n_sample: 0
+
+data:
+    corpus_1:
+        path_src: hf://quickmt/quickmt-train.vi-en/en
+        path_tgt: hf://quickmt/quickmt-train.vi-en/vi
+        path_sco: hf://quickmt/quickmt-train.vi-en/sco
+    valid:
+        path_src: valid.en
+        path_tgt: valid.vi
+
+transforms: [sentencepiece, filtertoolong]
+transforms_configs:
+  sentencepiece:
+    src_subword_model: "en.spm.model"
+    tgt_subword_model: "vi.spm.model"
+  filtertoolong:
+    src_seq_length: 256
+    tgt_seq_length: 256
+
+training:
+    # Run configuration
+    model_path: quickmt-en-vi-eole-model
+    #train_from: model
+    keep_checkpoint: 4
+    train_steps: 100000
+    save_checkpoint_steps: 5000
+    valid_steps: 5000
+    
+    # Train on a single GPU
+    world_size: 1
+    gpu_ranks: [0]
+
+    # Batching 10240
+    batch_type: "tokens"
+    batch_size: 8000
+    valid_batch_size: 4096
+    batch_size_multiple: 8
+    accum_count: [10]
+    accum_steps: [0]
+
+    # Optimizer & Compute
+    compute_dtype: "fp16"
+    optim: "adamw"
+    #use_amp: False
+    learning_rate: 2.0
+    warmup_steps: 4000
+    decay_method: "noam"
+    adam_beta2: 0.998
+
+    # Data loading
+    bucket_size: 128000
+    num_workers: 4
+    prefetch_factor: 32
+    
+    # Hyperparams
+    dropout_steps: [0]
+    dropout: [0.1]
+    attention_dropout: [0.1]
+    max_grad_norm: 0
+    label_smoothing: 0.1
+    average_decay: 0.0001
+    param_init_method: xavier_uniform
+    normalization: "tokens"
+
+model:
+    architecture: "transformer"
+    share_embeddings: false
+    share_decoder_embeddings: false
+    hidden_size: 1024
+    encoder:
+        layers: 8
+    decoder:
+        layers: 2
+    heads: 8
+    transformer_ff: 4096
+    embeddings:
+        word_vec_size: 1024
+        position_encoding_type: "SinusoidalInterleaved"
+

eole-model/config.json

@@ -0,0 +1,132 @@
+{
+  "tgt_vocab": "vi.eole.vocab",
+  "vocab_size_multiple": 8,
+  "src_vocab": "en.eole.vocab",
+  "report_every": 100,
+  "tensorboard": true,
+  "transforms": [
+    "sentencepiece",
+    "filtertoolong"
+  ],
+  "share_vocab": false,
+  "n_sample": 0,
+  "tgt_vocab_size": 20000,
+  "src_vocab_size": 20000,
+  "tensorboard_log_dir_dated": "tensorboard/Jul-11_15-14-32",
+  "seed": 1234,
+  "valid_metrics": [
+    "BLEU"
+  ],
+  "overwrite": true,
+  "tensorboard_log_dir": "tensorboard",
+  "save_data": "data",
+  "training": {
+    "bucket_size": 128000,
+    "num_workers": 0,
+    "batch_type": "tokens",
+    "save_checkpoint_steps": 5000,
+    "keep_checkpoint": 4,
+    "accum_count": [
+      10
+    ],
+    "decay_method": "noam",
+    "param_init_method": "xavier_uniform",
+    "adam_beta2": 0.998,
+    "gpu_ranks": [
+      0
+    ],
+    "world_size": 1,
+    "valid_batch_size": 4096,
+    "max_grad_norm": 0.0,
+    "prefetch_factor": 32,
+    "accum_steps": [
+      0
+    ],
+    "warmup_steps": 4000,
+    "optim": "adamw",
+    "compute_dtype": "torch.float16",
+    "dropout_steps": [
+      0
+    ],
+    "valid_steps": 5000,
+    "normalization": "tokens",
+    "attention_dropout": [
+      0.1
+    ],
+    "model_path": "quickmt-en-vi-eole-model",
+    "label_smoothing": 0.1,
+    "train_steps": 100000,
+    "batch_size_multiple": 8,
+    "average_decay": 0.0001,
+    "batch_size": 8000,
+    "learning_rate": 2.0,
+    "dropout": [
+      0.1
+    ]
+  },
+  "transforms_configs": {
+    "filtertoolong": {
+      "src_seq_length": 256,
+      "tgt_seq_length": 256
+    },
+    "sentencepiece": {
+      "src_subword_model": "${MODEL_PATH}/en.spm.model",
+      "tgt_subword_model": "${MODEL_PATH}/vi.spm.model"
+    }
+  },
+  "data": {
+    "corpus_1": {
+      "transforms": [
+        "sentencepiece",
+        "filtertoolong"
+      ],
+      "path_src": "train.en",
+      "path_align": null,
+      "path_tgt": "train.vi"
+    },
+    "valid": {
+      "transforms": [
+        "sentencepiece",
+        "filtertoolong"
+      ],
+      "path_src": "valid.en",
+      "path_align": null,
+      "path_tgt": "valid.vi"
+    }
+  },
+  "model": {
+    "heads": 8,
+    "hidden_size": 1024,
+    "position_encoding_type": "SinusoidalInterleaved",
+    "share_embeddings": false,
+    "transformer_ff": 4096,
+    "share_decoder_embeddings": false,
+    "architecture": "transformer",
+    "decoder": {
+      "decoder_type": "transformer",
+      "hidden_size": 1024,
+      "heads": 8,
+      "layers": 2,
+      "position_encoding_type": "SinusoidalInterleaved",
+      "transformer_ff": 4096,
+      "tgt_word_vec_size": 1024,
+      "n_positions": null
+    },
+    "encoder": {
+      "hidden_size": 1024,
+      "heads": 8,
+      "encoder_type": "transformer",
+      "layers": 8,
+      "position_encoding_type": "SinusoidalInterleaved",
+      "src_word_vec_size": 1024,
+      "transformer_ff": 4096,
+      "n_positions": null
+    },
+    "embeddings": {
+      "tgt_word_vec_size": 1024,
+      "src_word_vec_size": 1024,
+      "word_vec_size": 1024,
+      "position_encoding_type": "SinusoidalInterleaved"
+    }
+  }
+}

eole-model/en.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82efd76a945ec4574b2ab3d2476b0043a9737c8dbe693cd20a7fa036d273c0eb
+size 586058

eole-model/model.00.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5ef7d951c721c41fc83116194400cd7a70fa072c1a29862598384036d6c3098
+size 823882912

eole-model/vi.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03be7f771990d2bee00d7ff09219baad95b53fef1dcd387b3048c156d665e046
+size 553521

model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4aa5184d9f3788fb1dde788b14acfd60741b3e0cefc24c4d9e771e8242e177e8
+size 401699775

src.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82efd76a945ec4574b2ab3d2476b0043a9737c8dbe693cd20a7fa036d273c0eb
+size 586058

tgt.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03be7f771990d2bee00d7ff09219baad95b53fef1dcd387b3048c156d665e046
+size 553521