# batched
## wikipedia (a model)
epoki zwiększają się stopniowo według wzoru EPOCHS + int(p/2) gdzie p to indeks batcha. (2,12)
BATCH_SIZE = 64
EPOCHS = 2
LATENT_DIM = 400
LATENT_DIM_DECODER = 400 # idea: make it different to ensure things all fit together properly!
NUM_SAMPLES = 100000
NUM_PARTS = 20 #number of batches to divide samples to
MAX_SEQUENCE_LENGTH = 100
MAX_NUM_WORDS = 20000
EMBEDDING_DIM = 100
```
Input sentence: This problem can be overcome in a number of ways, including the use of a relatively short length of fiber that has the opposite refractive index gradient. =
Predicted translation: this problem can be difficult in a number of certain including the of a maximum of fiber that has the opposite index is
Actual translation: This problem can be overcome in a number of ways, including the use of a relatively short length of fiber that has the opposite refractive index gradient. = <eos>
Continue? [Y/n]
Input sentence: : How George Boole and Claude Shannon Create the Information Age, Princeton University Press, 2013, ISBN
Predicted translation: we george and birmingham and the secure european street prize in american american
Actual translation: : How George Boole and Claude Shannon Create the Information Age, Princeton University Press, 2013, ISBN <eos>
Continue? [Y/n]
Input sentence: In other words, when someone uses a telephone, the speaker's voice is "encoded" then reconstructed for the person on the other end.
Predicted translation: in other words, when if a telephone the medium is specified called for the person on the other end.
Actual translation: In other words, when someone uses a telephone, the speaker's voice is "encoded" then reconstructed for the person on the other end. <eos>
Continue? [Y/n]
Input sentence: The ARPANET was a research project that was communications-oriented, rather than user-oriented in design.
Predicted translation: the broadcasting was a research project that was network rather than user with technology.
Actual translation: The ARPANET was a research project that was communications-oriented, rather than user-oriented in design. <eos>
Continue? [Y/n]
Input sentence: Websites are often created using content management software with, initially, very little content.
Predicted translation: among such as uses security management software with initially very larger functions.
Actual translation: Websites are often created using content management software with, initially, very little content. <eos>
Continue? [Y/n]
Input sentence: Promotional trailers arrive on a separate hard drive and range between 200 GB and 400 GB in size.
Predicted translation: apple's cameras on a separate larger repair and range between 30 and and in earth.
Actual translation: Promotional trailers arrive on a separate hard drive and range between 200 GB and 400 GB in size. <eos>
Continue? [Y/n]
Input sentence: Kincaid is best known for his convergence model of communication, a nonlinear model of communication wherein two communicators strive to reach "mutual understanding."
Predicted translation: although is significant for his information model of communication a high model of communication
Actual translation: Kincaid is best known for his convergence model of communication, a nonlinear model of communication wherein two communicators strive to reach "mutual understanding." <eos>
Continue? [Y/n]
Input sentence: Subscriptions to the overall offering hosted on ScienceDirect, rather than to specific titles it carries, are usually acquired through a so called big deal.
Predicted translation: subscribers to the gain for cnn than rather than to specific while it called are usually through a so called
Actual translation: Subscriptions to the overall offering hosted on ScienceDirect, rather than to specific titles it carries, are usually acquired through a so called big deal. <eos>
Continue? [Y/n]
Input sentence: One additional approach which is not initiated by operators is the usage of over-the-top content (OTT) services, using applications like Skype and Google Talk to provide LTE voice service.
Predicted translation: one additional approach which is not initiated by operators is the availability of over the 2g content services, like and google talk to provide lte voice service.
Actual translation: One additional approach which is not initiated by operators is the usage of over-the-top content (OTT) services, using applications like Skype and Google Talk to provide LTE voice service. <eos>
Continue? [Y/n]
Input sentence: Europe had a total population of about 741 million (about 11% of the world population) as of 2018.
Predicted translation: europe had a total population of about million about and the population and in 2018.
Actual translation: Europe had a total population of about 741 million (about 11% of the world population) as of 2018. <eos>
Continue? [Y/n]
Input sentence: The gain of each channel is made proportional to the rms signal level and inversely proportional to the mean square noise level in that channel.
Predicted translation: the gain of each channel is made proportional to the earth signal and inversely proportional to the person square noise in the networks.
Actual translation: The gain of each channel is made proportional to the rms signal level and inversely proportional to the mean square noise level in that channel. <eos>
Continue? [Y/n]
Input sentence: In 2009, Obama signed into law the National Defense Authorization Act for Fiscal Year 2010, which contained in it the Matthew Shepard and
Predicted translation: in 2009, 2017, after cnn and the national intelligence proposal for order was the latter will the and
Actual translation: In 2009, Obama signed into law the National Defense Authorization Act for Fiscal Year 2010, which contained in it the Matthew Shepard and <eos>
Continue? [Y/n]
Input sentence: On 16 March 2011, the freighter MS Oliva ran aground on Nightingale Island, spilling tons of heavy fuel oil into the ocean.
Predicted translation: on 16 march 2011, the 1 george before 30 of heavy and the ocean into the ocean
Actual translation: On 16 March 2011, the freighter MS Oliva ran aground on Nightingale Island, spilling tons of heavy fuel oil into the ocean. <eos>
Continue? [Y/n]
Input sentence: Since UDH is part of the payload, the number of available characters per segment is lower: 153 for 7-bit encoding, 134 for 8-bit encoding and 67 for 16-bit encoding.
Predicted translation: since is part of the oscillating of the web rate is using for using high bits and and higher
Actual translation: Since UDH is part of the payload, the number of available characters per segment is lower: 153 for 7-bit encoding, 134 for 8-bit encoding and 67 for 16-bit encoding. <eos>
Continue? [Y/n]
Input sentence: In Zone 1, things are much different.
Predicted translation: in the processes are much other.
Actual translation: In Zone 1, things are much different. <eos>
Continue? [Y/n]
Input sentence: Such a handover, in which the source and the target are different cells (even if they are on the same cell site) is called inter-cell handover.
Predicted translation: such a key in which the source and the target are different cells even if they are on the same cell site is called digital
Actual translation: Such a handover, in which the source and the target are different cells (even if they are on the same cell site) is called inter-cell handover. <eos>
Continue? [Y/n]
Input sentence: Single-chip processors increase reliability because there are many fewer electrical connections that could fail.
Predicted translation: single chip devices, offer generation level because many other and the user or that
Actual translation: Single-chip processors increase reliability because there are many fewer electrical connections that could fail. <eos>
```
## experiment (brak modelu)
``` =python
for i in range(EPOCHS):
for batch in BATCHES_OF_DATA:
policz_model(epochs=1)
```
skopałam zapisywanie danych i to tylko ostatnie 5k ze 100k i ostatnia epoka, ale reszta też tak falowała koło 0.9
```
Input sentence: The DCT is the coding standard for multimedia communications devices.
Predicted translation: the dct is the coding standard for multimedia communications devices. ===
Actual translation: The DCT is the coding standard for multimedia communications devices. <eos>
Continue? [Y/n]
Input sentence: A statement by the Prime Minister summarized the situation on 8 September.
Predicted translation: a statement by the prime minister summarized the situation on 8 september.
Actual translation: A statement by the Prime Minister summarized the situation on 8 September. <eos>
Continue? [Y/n]
Input sentence: Thus, they are often gold plated.
Predicted translation: thus, they are often gold
Actual translation: Thus, they are often gold plated. <eos>
Continue? [Y/n]
Input sentence: Postman was born in New York City, where he would spend most of his life.
Predicted translation: postman was born in new york city, where he would spend most of his life. ===
Actual translation: Postman was born in New York City, where he would spend most of his life. <eos>
Continue? [Y/n]
Input sentence: Yet, the paradigm's inherent influence perspective and other central questions of the debate keep re-emerging under different headings and more advanced concepts such as hybridization and glocalisation. ==
Predicted translation: yet, the inherent influence perspective and other central questions of the debate keep under different and more advanced concepts such as and
Actual translation: Yet, the paradigm's inherent influence perspective and other central questions of the debate keep re-emerging under different headings and more advanced concepts such as hybridization and glocalisation. == <eos>
Continue? [Y/n]
Input sentence: However, McLuhan's comments are more aware of the contingent cultural context in which for instance linear perspective arose, while Panofsky's ones are more teleological.
Predicted translation: however, mcluhan's comments are more aware of the cultural context in which for instance, linear perspective while ones are more
Actual translation: However, McLuhan's comments are more aware of the contingent cultural context in which for instance linear perspective arose, while Panofsky's ones are more teleological. <eos>
Continue? [Y/n]
Input sentence: During 1790â1795, at the height of the French Revolution, France needed a swift and reliable military communications system to thwart the war efforts of its enemies.
Predicted translation: during at the height of the french revolution france needed a swift and reliable military communications system to thwart the war efforts of its enemies.
Actual translation: During 1790â1795, at the height of the French Revolution, France needed a swift and reliable military communications system to thwart the war efforts of its enemies. <eos>
Continue? [Y/n]
Input sentence: The early Swedish system was more flexible, having the ability to set up message connections between arbitrary stations.
Predicted translation: the early swedish system was more flexible, having the ability to set up message between arbitrary stations. ===
Actual translation: The early Swedish system was more flexible, having the ability to set up message connections between arbitrary stations. <eos>
Continue? [Y/n]
Input sentence: He spent the time waiting for a new voyage inventing his mirror galvanometer.
Predicted translation: he spent the time waiting for a new voyage inventing his mirror.
Actual translation: He spent the time waiting for a new voyage inventing his mirror galvanometer. <eos>
Continue? [Y/n]
Input sentence: John Lennon and Yoko Ono summered in Mill Valley in the early 1970s, having left some of his own graffiti on the wall of the residence
Predicted translation: john and in mill valley in the early 1970s, having left some of his own on the wall of the
Actual translation: John Lennon and Yoko Ono summered in Mill Valley in the early 1970s, having left some of his own graffiti on the wall of the residence <eos>
Continue? [Y/n]
Input sentence: It has been possible to carry out Denial-of-service attacks in a P2P lab environment, where users running BitTorrent clients act as amplifiers for an attack at another service.
Predicted translation: it has been possible to carry out denial of service attacks in a iot lab where users running bittorrent clients act as amplifiers for an attack at another service.
Actual translation: It has been possible to carry out Denial-of-service attacks in a P2P lab environment, where users running BitTorrent clients act as amplifiers for an attack at another service. <eos>
Continue? [Y/n]
Input sentence: Related documents were shown using citation and word based measures and an active and continuously updated bibliography is shown for each document.
Predicted translation: related documents were shown using citation and word based measures and an active and continuously updated bibliography is shown for each
Actual translation: Related documents were shown using citation and word based measures and an active and continuously updated bibliography is shown for each document. <eos>
```
## wikispoiledstruct model
BATCH_SIZE = 10
EPOCHS = 25
LATENT_DIM = 117
LATENT_DIM_DECODER = 63 # idea: make it different to ensure things all fit together properly!
NUM_SAMPLES = 100000
NUM_PARTS = 20 #number of batches to divide samples to
MAX_SEQUENCE_LENGTH = 100
MAX_NUM_WORDS = 20000
EMBEDDING_DIM = 100
poprawione rysowanie wykresu (wartości loss, acc itp uśrednione z całego datasetu dla każdej epoki oddzielnie)
wywaliło się
## wikibooks dictionary pierwszy "lepsiejszy" niż poprzednie (wikibooksidict)
BATCH_SIZE = 30
EPOCHS = 15
LATENT_DIM = 168 #encoder data shape
LATENT_DIM_DECODER = 148 # decoder data shape
NUM_SAMPLES = 90000
NUM_PARTS = 40 #number of batches to divide samples to
MAX_SEQUENCE_LENGTH = 100
MAX_NUM_WORDS = 20000
EMBEDDING_DIM = 100
zdanka trzeba przetestować na serwisie bo komórka się nie odpala po tym jak model się kończy liczyć
## wikibooks struct
Sign in with Wallet
Connect another wallet