Instructions to use OptimaMT/fr-ar with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers.js
How to use OptimaMT/fr-ar with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('translation', 'OptimaMT/fr-ar');
Add OptimaMT fr-ar
Browse files- .gitattributes +2 -0
- README.md +59 -0
- config.json +60 -0
- generation_config.json +16 -0
- onnx/decoder_model_merged_quantized.onnx +3 -0
- onnx/encoder_model_quantized.onnx +3 -0
- source.spm +3 -0
- special_tokens_map.json +5 -0
- target.spm +3 -0
- tokenizer.json +0 -0
- tokenizer_config.json +12 -0
- vocab.json +0 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
source.spm filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
target.spm filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- fr
|
| 5 |
+
- ar
|
| 6 |
+
pipeline_tag: translation
|
| 7 |
+
library_name: transformers.js
|
| 8 |
+
base_model: Helsinki-NLP/opus-mt-fr-ar
|
| 9 |
+
tags:
|
| 10 |
+
- translation
|
| 11 |
+
- marian
|
| 12 |
+
- opus-mt
|
| 13 |
+
- onnx
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# OptimaMT fr → ar (French to Arabic)
|
| 17 |
+
|
| 18 |
+
Drop-in copy of an ONNX export of [Helsinki-NLP/opus-mt-fr-ar](https://huggingface.co/Helsinki-NLP/opus-mt-fr-ar) at revision `5469dc759cd9` (same layout as the Xenova exports) with a lighter merged decoder: same weights, same tokens, less memory.
|
| 19 |
+
Part of [OptimaMT](https://github.com/AmerAlreyahi/OptimaMT) (49 pairs, `OptimaMT/<src>-<tgt>`).
|
| 20 |
+
|
| 21 |
+
| Memory after load | Linux | Windows | Identical tokens |
|
| 22 |
+
|---|---:|---:|---|
|
| 23 |
+
| `disable_quant_qdq = 1` | 525 → 281 MiB (-47 %) | 500 → 257 MiB (-49 %) | yes |
|
| 24 |
+
| ONNX Runtime defaults | 281 → 220 MiB (-22 %) | 196 → 165 MiB (-16 %) | yes |
|
| 25 |
+
|
| 26 |
+
ONNX Runtime 1.31.1, CPU, 4 threads (Linux; Windows in a VM). Interactive results: [OptimaMT/results](https://huggingface.co/spaces/OptimaMT/results) ·
|
| 27 |
+
method: [REPORT.md](https://github.com/AmerAlreyahi/OptimaMT/blob/main/REPORT.md).
|
| 28 |
+
|
| 29 |
+
## Use
|
| 30 |
+
|
| 31 |
+
```js
|
| 32 |
+
import { pipeline } from "@huggingface/transformers";
|
| 33 |
+
|
| 34 |
+
const translator = await pipeline("translation", "OptimaMT/fr-ar", {
|
| 35 |
+
dtype: "q8",
|
| 36 |
+
session_options: { extra: { session: { disable_quant_qdq: "1" } } },
|
| 37 |
+
});
|
| 38 |
+
const [out] = await translator(">>ara<< Le phare se dresse sur la côte.");
|
| 39 |
+
console.log(out.translation_text);
|
| 40 |
+
```
|
| 41 |
+
|
| 42 |
+
Python ([translate.py](https://github.com/AmerAlreyahi/OptimaMT/blob/main/scripts/translate.py): onnxruntime, sentencepiece, numpy):
|
| 43 |
+
|
| 44 |
+
```bash
|
| 45 |
+
python translate.py OptimaMT/fr-ar "Le phare se dresse sur la côte." --lang ">>ara<<"
|
| 46 |
+
```
|
| 47 |
+
|
| 48 |
+
Start every input with `>>ara<<` (the target variant). Use transformers.js 3.x: 4.3.2 splits this token, with the original export too.
|
| 49 |
+
|
| 50 |
+
## Files
|
| 51 |
+
|
| 52 |
+
`onnx/decoder_model_merged_quantized.onnx` is rebuilt; `tokenizer.json`, `special_tokens_map.json` and `tokenizer_config.json` were written in the Xenova layout; every other file is the original, byte for byte.
|
| 53 |
+
Vocabulary: 62,419 tokens.
|
| 54 |
+
|
| 55 |
+
## Licence and credits
|
| 56 |
+
|
| 57 |
+
- Weights: [OPUS-MT](https://huggingface.co/Helsinki-NLP/opus-mt-fr-ar) by the University of Helsinki, CC BY 4.0.
|
| 58 |
+
- ONNX export: an ONNX export of [Helsinki-NLP/opus-mt-fr-ar](https://huggingface.co/Helsinki-NLP/opus-mt-fr-ar) at revision `5469dc759cd9` (same layout as the Xenova exports).
|
| 59 |
+
- Modified: the decoder's output projection was moved out of its two branches (graph change only). CC BY 4.0.
|
config.json
ADDED
|
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "/tmp/Helsinki-NLP/opus-mt-fr-ar",
|
| 3 |
+
"activation_dropout": 0.0,
|
| 4 |
+
"activation_function": "swish",
|
| 5 |
+
"add_bias_logits": false,
|
| 6 |
+
"add_final_layer_norm": false,
|
| 7 |
+
"architectures": [
|
| 8 |
+
"MarianMTModel"
|
| 9 |
+
],
|
| 10 |
+
"attention_dropout": 0.0,
|
| 11 |
+
"bad_words_ids": [
|
| 12 |
+
[
|
| 13 |
+
62418
|
| 14 |
+
]
|
| 15 |
+
],
|
| 16 |
+
"bos_token_id": 0,
|
| 17 |
+
"classif_dropout": 0.0,
|
| 18 |
+
"classifier_dropout": 0.0,
|
| 19 |
+
"d_model": 512,
|
| 20 |
+
"decoder_attention_heads": 8,
|
| 21 |
+
"decoder_ffn_dim": 2048,
|
| 22 |
+
"decoder_layerdrop": 0.0,
|
| 23 |
+
"decoder_layers": 6,
|
| 24 |
+
"decoder_start_token_id": 62418,
|
| 25 |
+
"decoder_vocab_size": 62419,
|
| 26 |
+
"dropout": 0.1,
|
| 27 |
+
"encoder_attention_heads": 8,
|
| 28 |
+
"encoder_ffn_dim": 2048,
|
| 29 |
+
"encoder_layerdrop": 0.0,
|
| 30 |
+
"encoder_layers": 6,
|
| 31 |
+
"eos_token_id": 0,
|
| 32 |
+
"extra_pos_embeddings": 62419,
|
| 33 |
+
"forced_eos_token_id": 0,
|
| 34 |
+
"id2label": {
|
| 35 |
+
"0": "LABEL_0",
|
| 36 |
+
"1": "LABEL_1",
|
| 37 |
+
"2": "LABEL_2"
|
| 38 |
+
},
|
| 39 |
+
"init_std": 0.02,
|
| 40 |
+
"is_encoder_decoder": true,
|
| 41 |
+
"label2id": {
|
| 42 |
+
"LABEL_0": 0,
|
| 43 |
+
"LABEL_1": 1,
|
| 44 |
+
"LABEL_2": 2
|
| 45 |
+
},
|
| 46 |
+
"max_length": 512,
|
| 47 |
+
"max_position_embeddings": 512,
|
| 48 |
+
"model_type": "marian",
|
| 49 |
+
"normalize_before": false,
|
| 50 |
+
"normalize_embedding": false,
|
| 51 |
+
"num_beams": 4,
|
| 52 |
+
"num_hidden_layers": 6,
|
| 53 |
+
"pad_token_id": 62418,
|
| 54 |
+
"scale_embedding": true,
|
| 55 |
+
"share_encoder_decoder_embeddings": true,
|
| 56 |
+
"static_position_embeddings": true,
|
| 57 |
+
"transformers_version": "4.22.0.dev0",
|
| 58 |
+
"use_cache": true,
|
| 59 |
+
"vocab_size": 62419
|
| 60 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bad_words_ids": [
|
| 3 |
+
[
|
| 4 |
+
62418
|
| 5 |
+
]
|
| 6 |
+
],
|
| 7 |
+
"bos_token_id": 0,
|
| 8 |
+
"decoder_start_token_id": 62418,
|
| 9 |
+
"eos_token_id": 0,
|
| 10 |
+
"forced_eos_token_id": 0,
|
| 11 |
+
"max_length": 512,
|
| 12 |
+
"num_beams": 4,
|
| 13 |
+
"pad_token_id": 62418,
|
| 14 |
+
"renormalize_logits": true,
|
| 15 |
+
"transformers_version": "4.32.0.dev0"
|
| 16 |
+
}
|
onnx/decoder_model_merged_quantized.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:83e0d1115910d4830bc98bc28da5b3dadacc993de130f72c74c812d2740573f3
|
| 3 |
+
size 58880688
|
onnx/encoder_model_quantized.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f7c318194076e252bcd028a91f7534bc05e65181842ceda203fc56f8f6d9b3f0
|
| 3 |
+
size 51577879
|
source.spm
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f32f818f2a89130ada3ab6b09a69cae104ae20ca5c9f1c309c41dae86a7dcc4e
|
| 3 |
+
size 826663
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"eos_token": "</s>",
|
| 3 |
+
"pad_token": "<pad>",
|
| 4 |
+
"unk_token": "<unk>"
|
| 5 |
+
}
|
target.spm
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ccaf1383c4d5296fff63006aaba595931a0da0edf3e2c1a0a3a85cc1e757aafd
|
| 3 |
+
size 924824
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"clean_up_tokenization_spaces": true,
|
| 3 |
+
"eos_token": "</s>",
|
| 4 |
+
"model_max_length": 512,
|
| 5 |
+
"pad_token": "<pad>",
|
| 6 |
+
"separate_vocabs": false,
|
| 7 |
+
"source_lang": "fra",
|
| 8 |
+
"sp_model_kwargs": {},
|
| 9 |
+
"target_lang": "ara",
|
| 10 |
+
"tokenizer_class": "MarianTokenizer",
|
| 11 |
+
"unk_token": "<unk>"
|
| 12 |
+
}
|
vocab.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|