Hugging Face's logo

WindyTranslate
/
translate-windy-hplt-mr-en

Translation
ctranslate2
marian
int8
mr
en
Model card Files
xet
 Community
sneakyfree commited on
Commit
2df3e0d
·
verified ·
1 Parent(s): 71b2f81

Promote from WindstormLabs/translate-windy-hplt-mr-en

Browse files
Files changed (10) hide show
  1. .gitattributes +2 -0
  2. README.md +54 -0
  3. config.json +10 -0
  4. model.bin +3 -0
  5. shared_vocabulary.json +0 -0
  6. source.spm +3 -0
  7. special_tokens_map.json +23 -0
  8. target.spm +3 -0
  9. tokenizer_config.json +38 -0
  10. 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 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ base_model: HPLT/translate-mr-en-v2.0-hplt_opus
4
+ library_name: ctranslate2
5
+ pipeline_tag: translation
6
+ tags: [translation, marian, ctranslate2, int8, mr, en]
7
+ ---
8
+
9
+ # translate-windy-hplt-mr-en — mr to en
10
+
11
+ Bilingual machine translation in **CTranslate2 INT8** for fast CPU inference.
12
+ Fine-tuned by **Windstorm Labs** from [`HPLT/translate-mr-en-v2.0-hplt_opus`](https://huggingface.co/HPLT/translate-mr-en-v2.0-hplt_opus).
13
+
14
+ ## Attribution
15
+
16
+ Derived from [`HPLT/translate-mr-en-v2.0-hplt_opus`](https://huggingface.co/HPLT/translate-mr-en-v2.0-hplt_opus), from the **HPLT** project, licensed under
17
+ **CC-BY-4.0**. **Modified by Windstorm Labs.** Attribution is retained as the licence requires.
18
+
19
+ ## What we did
20
+
21
+ Converted from the upstream Marian release, LoRA fine-tuned on parallel data for this
22
+ language pair, merged into the base weights, then quantized to INT8.
23
+
24
+ | | |
25
+ |---|---|
26
+ | Rank / alpha | 8 / 16 |
27
+ | Learning rate / steps | 2.5e-06 / 50 |
28
+ | Target modules | `q_proj`, `v_proj` |
29
+ | Seed | 42 |
30
+ | Training data | opus100_mr-en (clinic cache), 2,000 sentence pairs |
31
+ | Tensors modified | 36 |
32
+
33
+ ## Provenance
34
+
35
+ ```
36
+ base model.bin sha256 7014cf14366ae8ef8bc83459feb265ed8bdfb71f8884741ca85cb79dc3ee1cf9
37
+ this model.bin sha256 4b424ffbbcf29b2514c8a1d60d6bacea9cfaa6c7341c4e451fc40987682b1734
38
+ ```
39
+
40
+ Verified after INT8 quantization, on the published artifact.
41
+
42
+ ## Evaluation
43
+
44
+ FLORES-200 devtest, 200 sentences, beam 4, CTranslate2 INT8 on CPU.
45
+
46
+ | | spBLEU | chrF |
47
+ |---|---:|---:|
48
+ | this model | **28.39** | 55.13 |
49
+
50
+ ## Notes
51
+
52
+ - Single language pair: mr to en.
53
+ - FLORES-200 is news and encyclopedic prose.
54
+ - No human evaluation was performed.
config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_source_bos": false,
3
+ "add_source_eos": false,
4
+ "bos_token": "<s>",
5
+ "decoder_start_token": "</s>",
6
+ "eos_token": "</s>",
7
+ "layer_norm_epsilon": null,
8
+ "multi_query_attention": false,
9
+ "unk_token": "<unk>"
10
+ }
model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b424ffbbcf29b2514c8a1d60d6bacea9cfaa6c7341c4e451fc40987682b1734
3
+ size 79047635
shared_vocabulary.json ADDED
The diff for this file is too large to render. See raw diff
 
source.spm ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:151a516aad031749f180e4380b0c8e9da98933d36293df72803a826143ecce1e
3
+ size 1810251
special_tokens_map.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "eos_token": {
3
+ "content": "</s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "pad_token": {
10
+ "content": "<pad>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "unk_token": {
17
+ "content": "<unk>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ }
23
+ }
target.spm ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:151a516aad031749f180e4380b0c8e9da98933d36293df72803a826143ecce1e
3
+ size 1810251
tokenizer_config.json ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "</s>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "1": {
12
+ "content": "<unk>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "64000": {
20
+ "content": "<pad>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ }
27
+ },
28
+ "clean_up_tokenization_spaces": false,
29
+ "eos_token": "</s>",
30
+ "model_max_length": 512,
31
+ "pad_token": "<pad>",
32
+ "separate_vocabs": false,
33
+ "source_lang": "",
34
+ "sp_model_kwargs": {},
35
+ "target_lang": "marian",
36
+ "tokenizer_class": "MarianTokenizer",
37
+ "unk_token": "<unk>"
38
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff