Upload folder using huggingface_hub

.ipynb_checkpoints/README-checkpoint.md

@@ -0,0 +1,101 @@
+---
+language:
+- en
+- it
+tags:
+- translation
+license: cc-by-4.0
+datasets:
+- quickmt/quickmt-train.it-en
+model-index:
+- name: quickmt-en-it
+  results:
+  - task:
+      name: Translation eng-ita
+      type: translation
+      args: eng-ita
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: eng_Latn ita_Latn devtest
+    metrics:
+    - name: BLEU
+      type: bleu
+      value: 30.52
+    - name: CHRF
+      type: chrf
+      value: 59.72
+    - name: COMET
+      type: comet
+      value: 87.55
+---
+
+
+# `quickmt-en-it` Neural Machine Translation Model 
+
+`quickmt-en-it` is a reasonably fast and reasonably accurate neural machine translation model for translation from `en` into `it`.
+
+
+## Model Information
+
+* Trained using [`eole`](https://github.com/eole-nlp/eole)
+* 185M 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.it-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-it ./quickmt-en-it
+```
+
+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-it/", 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)
+
+> 'Il dottor Ehud Ur, professore di medicina alla Dalhousie University di Halifax, in Nuova Scozia e presidente della divisione clinica e scientifica della Canadian Diabetes Association, ha avvertito che la ricerca è ancora agli inizi.'
+
+# 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)
+
+> "Il Dott. Ehud Ur, professore di medicina alla Dalhousie University di Halifax, Nuova Scozia e presidente del dipartimento scientifico e clinica dell'Associazione Canada-del diabete ha avvertito che questa ricerca si trova ancora in pieno giorno."
+
+```
+
+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"->"ita_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 using a large batch size).
+
+|                                  |   bleu |   chrf2 |   comet22 |   Time (s) |
+|:---------------------------------|-------:|--------:|----------:|-----------:|
+| quickmt/quickmt-en-it            |  30.52 |   59.72 |     87.55 |       1.06 |
+| Helsink-NLP/opus-mt-en-it        |  27.58 |   57.66 |     85.1  |       4.42 |
+| facebook/nllb-200-distilled-600M |  28.58 |   58.3  |     87.3  |      24.93 |
+| facebook/nllb-200-distilled-1.3B |  30.92 |   59.89 |     88.52 |      43.8  |
+| facebook/m2m100_418M             |  23.81 |   54.34 |     82.64 |      20.34 |
+| facebook/m2m100_1.2B             |  27.22 |   57.19 |     86.23 |      39.98 |
+
+`quickmt-en-it` is the fastest and second highest quality next to `nllb-200-distilled-1.3B`.

README.md

@@ -0,0 +1,101 @@
+---
+language:
+- en
+- it
+tags:
+- translation
+license: cc-by-4.0
+datasets:
+- quickmt/quickmt-train.it-en
+model-index:
+- name: quickmt-en-it
+  results:
+  - task:
+      name: Translation eng-ita
+      type: translation
+      args: eng-ita
+    dataset:
+      name: flores101-devtest
+      type: flores_101
+      args: eng_Latn ita_Latn devtest
+    metrics:
+    - name: BLEU
+      type: bleu
+      value: 30.52
+    - name: CHRF
+      type: chrf
+      value: 59.72
+    - name: COMET
+      type: comet
+      value: 87.55
+---
+
+
+# `quickmt-en-it` Neural Machine Translation Model 
+
+`quickmt-en-it` is a reasonably fast and reasonably accurate neural machine translation model for translation from `en` into `it`.
+
+
+## Model Information
+
+* Trained using [`eole`](https://github.com/eole-nlp/eole)
+* 185M 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.it-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-it ./quickmt-en-it
+```
+
+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-it/", 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)
+
+> 'Il dottor Ehud Ur, professore di medicina alla Dalhousie University di Halifax, in Nuova Scozia e presidente della divisione clinica e scientifica della Canadian Diabetes Association, ha avvertito che la ricerca è ancora agli inizi.'
+
+# 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)
+
+> "Il Dott. Ehud Ur, professore di medicina alla Dalhousie University di Halifax, Nuova Scozia e presidente del dipartimento scientifico e clinica dell'Associazione Canada-del diabete ha avvertito che questa ricerca si trova ancora in pieno giorno."
+
+```
+
+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"->"ita_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 using a large batch size).
+
+|                                  |   bleu |   chrf2 |   comet22 |   Time (s) |
+|:---------------------------------|-------:|--------:|----------:|-----------:|
+| quickmt/quickmt-en-it            |  30.52 |   59.72 |     87.55 |       1.06 |
+| Helsink-NLP/opus-mt-en-it        |  27.58 |   57.66 |     85.1  |       4.42 |
+| facebook/nllb-200-distilled-600M |  28.58 |   58.3  |     87.3  |      24.93 |
+| facebook/nllb-200-distilled-1.3B |  30.92 |   59.89 |     88.52 |      43.8  |
+| facebook/m2m100_418M             |  23.81 |   54.34 |     82.64 |      20.34 |
+| facebook/m2m100_1.2B             |  27.22 |   57.19 |     86.23 |      39.98 |
+
+`quickmt-en-it` is the fastest and second highest quality next to `nllb-200-distilled-1.3B`.

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 +1,97 @@
+## 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: it.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.ja-en/ja
+        # path_tgt: hf://quickmt/quickmt-train.ja-en/en
+        # path_sco: hf://quickmt/quickmt-train.ja-en/sco
+        path_src: train.en
+        path_tgt: train.it
+    valid:
+        path_src: valid.en
+        path_tgt: valid.it
+
+transforms: [sentencepiece, filtertoolong]
+transforms_configs:
+  sentencepiece:
+    src_subword_model: "en.spm.model"
+    tgt_subword_model: "it.spm.model"
+  filtertoolong:
+    src_seq_length: 256
+    tgt_seq_length: 256
+
+training:
+    # Run configuration
+    model_path: quickmt-en-it-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"
+    learning_rate: 2.0
+    warmup_steps: 4000
+    decay_method: "noam"
+    adam_beta2: 0.998
+
+    # Data loading
+    bucket_size: 512000
+    num_workers: 4
+    prefetch_factor: 64
+    
+    # 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 @@
+{
+  "src_vocab": "en.eole.vocab",
+  "n_sample": 0,
+  "seed": 1234,
+  "tensorboard": true,
+  "src_vocab_size": 20000,
+  "vocab_size_multiple": 8,
+  "tensorboard_log_dir_dated": "tensorboard/Apr-21_19-00-53",
+  "tgt_vocab_size": 20000,
+  "tensorboard_log_dir": "tensorboard",
+  "share_vocab": false,
+  "save_data": "data",
+  "transforms": [
+    "sentencepiece",
+    "filtertoolong"
+  ],
+  "valid_metrics": [
+    "BLEU"
+  ],
+  "tgt_vocab": "it.eole.vocab",
+  "report_every": 100,
+  "overwrite": true,
+  "training": {
+    "optim": "adamw",
+    "average_decay": 0.0001,
+    "keep_checkpoint": 4,
+    "model_path": "quickmt-en-it-eole-model",
+    "gpu_ranks": [
+      0
+    ],
+    "valid_steps": 5000,
+    "batch_size": 8000,
+    "bucket_size": 512000,
+    "adam_beta2": 0.998,
+    "label_smoothing": 0.1,
+    "warmup_steps": 4000,
+    "param_init_method": "xavier_uniform",
+    "normalization": "tokens",
+    "batch_type": "tokens",
+    "save_checkpoint_steps": 5000,
+    "num_workers": 0,
+    "accum_count": [
+      10
+    ],
+    "train_steps": 100000,
+    "dropout": [
+      0.1
+    ],
+    "learning_rate": 2.0,
+    "max_grad_norm": 0.0,
+    "attention_dropout": [
+      0.1
+    ],
+    "batch_size_multiple": 8,
+    "prefetch_factor": 64,
+    "compute_dtype": "torch.float16",
+    "decay_method": "noam",
+    "valid_batch_size": 4096,
+    "world_size": 1,
+    "accum_steps": [
+      0
+    ],
+    "dropout_steps": [
+      0
+    ]
+  },
+  "data": {
+    "corpus_1": {
+      "transforms": [
+        "sentencepiece",
+        "filtertoolong"
+      ],
+      "path_tgt": "train.it",
+      "path_align": null,
+      "path_src": "train.en"
+    },
+    "valid": {
+      "transforms": [
+        "sentencepiece",
+        "filtertoolong"
+      ],
+      "path_tgt": "valid.it",
+      "path_align": null,
+      "path_src": "valid.en"
+    }
+  },
+  "model": {
+    "transformer_ff": 4096,
+    "architecture": "transformer",
+    "heads": 8,
+    "share_embeddings": false,
+    "hidden_size": 1024,
+    "share_decoder_embeddings": false,
+    "position_encoding_type": "SinusoidalInterleaved",
+    "embeddings": {
+      "position_encoding_type": "SinusoidalInterleaved",
+      "src_word_vec_size": 1024,
+      "tgt_word_vec_size": 1024,
+      "word_vec_size": 1024
+    },
+    "decoder": {
+      "tgt_word_vec_size": 1024,
+      "layers": 2,
+      "transformer_ff": 4096,
+      "n_positions": null,
+      "heads": 8,
+      "decoder_type": "transformer",
+      "hidden_size": 1024,
+      "position_encoding_type": "SinusoidalInterleaved"
+    },
+    "encoder": {
+      "src_word_vec_size": 1024,
+      "layers": 8,
+      "transformer_ff": 4096,
+      "n_positions": null,
+      "encoder_type": "transformer",
+      "heads": 8,
+      "hidden_size": 1024,
+      "position_encoding_type": "SinusoidalInterleaved"
+    }
+  },
+  "transforms_configs": {
+    "filtertoolong": {
+      "tgt_seq_length": 256,
+      "src_seq_length": 256
+    },
+    "sentencepiece": {
+      "src_subword_model": "${MODEL_PATH}/en.spm.model",
+      "tgt_subword_model": "${MODEL_PATH}/it.spm.model"
+    }
+  }
+}

eole-model/en.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:352edbef8a5896ac750aa1318696d770c15c85d79116c94b4872086a1de94b2d
+size 586485

eole-model/it.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47b20d73c2224922ffef2d6f4bbe72810d7f75c80281dcea17515fc25a7a7bb9
+size 598389

eole-model/model.00.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1164a0b5ebb78a58e4e10563aeab594b3633d5e306c3e7e5760699984c0e9076
+size 823882912

model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:def57a69faad1db092d1ac8ad7db1a9ee056e77625ab3dca0d2d1ac22d6c1f37
+size 401699775

src.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:352edbef8a5896ac750aa1318696d770c15c85d79116c94b4872086a1de94b2d
+size 586485

tgt.spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47b20d73c2224922ffef2d6f4bbe72810d7f75c80281dcea17515fc25a7a7bb9
+size 598389