Downloads · 30 days
22
39% of all-time downloads
TigreGotico/m2m100_418M_ibo_en_rel_news_ft-onnx
m2m100_418M_ibo_en_rel_news_ft-onnx is a translation model from TigreGotico. Use it when you need text moved from one language to another. The card lists the license as afl-3.0.
ONNX export of masakhane/m2m100418Miboenrelnewsft.
Downloads · 30 days
22
39% of all-time downloads
All-time downloads
57
Public
Repo size
7.9 GB
Likes
0
Public
Click a slice to open those files.
.onnx7.9 GB · 100%
From the Hugging Face model README
ONNX export of masakhane/m2m100_418M_ibo_en_rel_news_ft.
Translation direction: Igbo (ig) -> English (en).
The original model was trained by Masakhane
(masakhane-io/lafand-mt) as part of MAFAND-MT
(Adelani et al., NAACL 2022, "A Few Thousand Translations Go a Long Way!").
It is a fine-tune of facebook/m2m100_418M on JW300 plus the MAFAND news corpus.
Licence: AFL-3.0, as declared by the original repository. This repository only converts
the weights to ONNX. All credit for the model belongs to Masakhane.
optimum-cli export onnx --model masakhane/m2m100_418M_ibo_en_rel_news_ft \
--task text2text-generation-with-past --no-post-process <outdir>
| Path | Precision | Size |
|---|---|---|
*.onnx (root) | fp32 | 4.6 GB |
int8/*.onnx | int8 dynamic | 3.0 GB |
10 general-domain Igbo sentences (source language for this ig->en
pair), PyTorch original vs ONNX, exact string match against
AutoModelForSeq2SeqLM.generate():
parity:
fp32_greedy: 1.00 # 10/10
fp32_beam4: 1.00 # 10/10
int8_greedy: 0.80 # 8/10
int8_beam4: 0.70 # 7/10
| Decoding | fp32 | int8 |
|---|---|---|
| greedy | 100% (10/10) | 80% (8/10) |
| beam=4 | 100% (10/10) | 70% (7/10) |
fp32 is an exact reproduction of the original model. int8 disagreements were inspected by hand and are paraphrases (e.g. ref "The children play joyfully in the field." vs int8 "The kids play joyfully in the field."), not degradations. Note the base model's own English output quality is uneven on some inputs regardless of precision (e.g. "This is a test sentence for igweaking." for a sentence about Igbo translation) - that is inherited from the original Masakhane checkpoint, not introduced by this conversion.
Igbo has a native M2M100 language code (ig). The target is already fixed
via forced_bos_token_id in generation_config.json. You must still set
the source language on the tokenizer:
tokenizer.src_lang = "ig"
from transformers import AutoTokenizer
from optimum.onnxruntime import ORTModelForSeq2SeqLM
tok = AutoTokenizer.from_pretrained("TigreGotico/m2m100_418M_ibo_en_rel_news_ft-onnx")
tok.src_lang = "ig"
model = ORTModelForSeq2SeqLM.from_pretrained("TigreGotico/m2m100_418M_ibo_en_rel_news_ft-onnx", use_cache=True, use_merged=False)
enc = tok("Ihu igwe dị mma taa.", return_tensors="pt")
out = model.generate(**enc, num_beams=4, max_new_tokens=64)
print(tok.batch_decode(out, skip_special_tokens=True)[0])
# -> The weather is good today.
For the int8 build, pass subfolder="int8".