flan-t5-base-opus-en-id-id-en

This model consist to be Translator in multimodal Indonesia and English only.

Model Details

Model Description

• Model type: Language model
• Language(s) (NLP): English, Indonesian
• License: Apache 2.0

Usage

Using the Pytorch model

Running the model on a CPU

Click to expand

from transformers import T5Tokenizer, T5ForConditionalGeneration
tokenizer = T5Tokenizer.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-en")
model = T5ForConditionalGeneration.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-en")
input_text = "translate English to Indonesia: How old are you?"
input_ids = tokenizer(input_text, return_tensors="pt").input_ids
outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))

Running the model on a GPU

Click to expand

# pip install accelerate
from transformers import T5Tokenizer, T5ForConditionalGeneration
tokenizer = T5Tokenizer.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-en")
model = T5ForConditionalGeneration.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-en", device_map="auto")
input_text = "translate English to Indonesia: How old are you?"
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")
outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))

Running the model on a GPU using different precisions

FP16

Click to expand

# pip install accelerate
import torch
from transformers import T5Tokenizer, T5ForConditionalGeneration
tokenizer = T5Tokenizer.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-ene")
model = T5ForConditionalGeneration.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-en", device_map="auto", torch_dtype=torch.float16)
input_text = "translate English to Indonesia: How old are you?"
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")
outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))

INT8

Click to expand

# pip install bitsandbytes accelerate
from transformers import T5Tokenizer, T5ForConditionalGeneration
tokenizer = T5Tokenizer.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-en")
model = T5ForConditionalGeneration.from_pretrained("muvazana/flan-t5-base-opus-en-id-id-en", device_map="auto", load_in_8bit=True)
input_text = "translate English to Indonesia: How old are you?"
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")
outputs = model.generate(input_ids)
print(tokenizer.decode(outputs[0]))

Training results

Training Loss	Epoch	Step	Validation Loss	Score	Counts	Totals	Precisions	Bp	Sys Len	Ref Len	Gen Len
1.6959	0.55	4000	1.5776	30.6542	[4414, 2368, 1345, 733]	[7417, 6417, 5426, 4519]	[59.511932047997846, 36.9019791179679, 24.78805750092149, 16.220402743969906]	1.0	7417	7354	10.77
1.4378	1.11	8000	1.4527	32.3772	[4526, 2538, 1483, 834]	[7567, 6567, 5576, 4666]	[59.81234306858729, 38.647784376427595, 26.596126255380202, 17.873981997428203]	1.0	7567	7354	10.885
1.3904	1.66	12000	1.3961	33.8978	[4558, 2559, 1494, 836]	[7286, 6286, 5295, 4383]	[62.55833104584134, 40.70951320394528, 28.21529745042493, 19.073693817020306]	0.9907	7286	7354	10.569
1.3035	2.21	16000	1.3758	34.9471	[4609, 2628, 1546, 880]	[7297, 6297, 5306, 4392]	[63.16294367548308, 41.73415912339209, 29.136826234451565, 20.036429872495447]	0.9922	7297	7354	10.591
1.2994	2.77	20000	1.3685	35.0259	[4617, 2627, 1550, 883]	[7288, 6288, 5297, 4382]	[63.350713501646545, 41.777989821882954, 29.261846328110252, 20.150616157005935]	0.991	7288	7354	10.556

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1
• Datasets 2.13.1
• Tokenizers 0.13.3