End of training

README.md

@@ -3,6 +3,8 @@ license: apache-2.0
 base_model: google-t5/t5-small
 tags:
 - generated_from_trainer
+metrics:
+- bleu
 model-index:
 - name: t5-small-finetuned-iwslt2037
   results: []
@@ -14,6 +16,10 @@ should probably proofread and complete it, then remove this comment. -->
 # t5-small-finetuned-iwslt2037
 
 This model is a fine-tuned version of [google-t5/t5-small](https://huggingface.co/google-t5/t5-small) on an unknown dataset.
+It achieves the following results on the evaluation set:
+- Loss: 2.4200
+- Bleu: 0.2502
+- Gen Len: 26.2162
 
 ## Model description
 

all_results.json

@@ -0,0 +1,23 @@
+{
+    "epoch": 100.0,
+    "eval_bleu": 0.2502,
+    "eval_gen_len": 26.2162,
+    "eval_loss": 2.4199626445770264,
+    "eval_runtime": 5.4423,
+    "eval_samples": 888,
+    "eval_samples_per_second": 163.166,
+    "eval_steps_per_second": 1.286,
+    "predict_bleu": 0.1921,
+    "predict_gen_len": 20.8805,
+    "predict_loss": 3.0906615257263184,
+    "predict_runtime": 5.6211,
+    "predict_samples": 1138,
+    "predict_samples_per_second": 202.453,
+    "predict_steps_per_second": 1.601,
+    "total_flos": 3.157662139522744e+17,
+    "train_loss": 2.261507166926833,
+    "train_runtime": 62730.7046,
+    "train_samples": 116654,
+    "train_samples_per_second": 185.96,
+    "train_steps_per_second": 2.906
+}

eval_results.json

@@ -0,0 +1,10 @@
+{
+    "epoch": 100.0,
+    "eval_bleu": 0.2502,
+    "eval_gen_len": 26.2162,
+    "eval_loss": 2.4199626445770264,
+    "eval_runtime": 5.4423,
+    "eval_samples": 888,
+    "eval_samples_per_second": 163.166,
+    "eval_steps_per_second": 1.286
+}

generated_predictions.txt

@@ -0,0 +1,1138 @@
+That's James Rises.
+You may know it because it won the New York CityTime re-engineering owner.
+For a long time, before anybody's heard of Snow, he wrote a book that he published spectacular, that the adult phones had stopped illegal phones from America.
+But it's another chapter that should leave a sexy impression behind it.
+He describes a catastrophic local U.S. secret surgery where Iran literally grew up in a nuclear bomb for a nuclear weapon.
+If that sounds crazy, read it.
+It's an incredible story.
+But you know, who didn't like the chapter at all?
+U.S. government administration.
+Almost 10 years, government has been able to draw the government against Rise and asked him to use its mental sources.
+In this train, he became a symbol of the pattern of government, Whisblowers and journalists.
+And the first thing is that the media has the right to publish secret information.
+But it's impossible to apply that right to media if that information doesn't get information and if they can't protect identity that they're surrounded by.
+So when the government arrived at Risearyt, he did something that a lot of brave reporters had already done before him: he would rather be crying, and he would rather walk in jail.
+So from 2007, to 2015 Rise lived with risk where prison was going.
+But then only days before the process, something extraordinary happened.
+Suddenly, even though they gave years more reliable for their case, the researchers were emitted to do it.
+The reason: in the time electronic growth can hide the reporter and sources and a few times.
+Instead of putting their digital foot and Riseing to the statement, it could also do this with its digital foot for him.
+And so the programs, without his averständnis, has secret to his telephone.
+It's as much as it's got to -- financial information and bank information, its loan, and even travel to the list of its wings.
+In the middle of this information, they found evidence that they used to CIA Sterling, a CIAist and Rise sources of Riseing.
+Unfortunately, it's just one of many cases.
+President Obama talked to protect Whisnblobs, but instead, the convicted of convicted of convicted of all of American government had gone up to him.
+Now, you can imagine how this could be a problem, especially because the government of work was secretly turned into a secret.
+Since 9/11, almost every article on national security has been about the result that a Whirblower went to a journalist.
+So we're putting the press on play, which is to protected through the first explosion, because the government has more and more opportunities to put everyone out.
+But just as much as technology allows government to avoid the rights of reporters, the press can also use technologies to protect their sources better.
+And that's the moment they can take that off, instead of having your contact, rather than when they're in the stuff.
+Today there's communication software that haven't been written as Rise's book, and it's a lot more safe than normal email or phone calls.
+One of the technologies is geo-engineerroprop, an open-source system for Whisblower, which was taken by the Internet blob of us, was developed by Aaron Swartz, and is now by the Freedom of the press Foundation that I work with.
+Instead of sending an email, go to a message set, like the Washington Post.
+And you can send documents up or information, like you can get at any ordinary contact.
+And they're then putting them together, and on a servers, to the only message that's ever gotten access to the news set.
+So the government can't really get more information, and many of the information that it would challenge them to do would not be available from ahead.
+But consciousnessroprop is just a small part of the whole, to protect press press in the 21st century.
+Unfortunately governments around the world are always developing new squatter technologies that are all endangered.
+It's safe to make sure that not just technology know how Edward Snow did a possibility to bring Miss Snows, it's as important that we protect the next Whirlwind in time.
+The tiniading of soldiers knows about the two-way hospitals, or the next environmental worker who takes a tinkering home for Flint -- or Wall Street, who warns us from the next financial crisis.
+After all, these technologies were not only done for those who want to cover crime, but to protect our very rights of all.
+Thank you.
+[Am 3.14 2016] [The largest data published the story.] [The "The theft and M.P.] [S.] [When have slammograms in the copy shit.] [When have we asked the money to do this?] [We asked Robert Palmer a week by Global Palmer, we got a news million people about a news documents of fada lawyer frog's frog's in Panama frog.
+The publication of these documents is actually a little bit of a small view of the secret world of tax.
+We get a sense of how clients and banks and rightel companies like Mossacks, frog's and say, "Okay, we need an anonymous company. Can you do that?"
+We actually see the email that exchange of news, like the whole system works like it works.
+This already led to the first very direct consequences.
+Islands Prime Minister is back.
+Also, there is a report that a brutal diet of brutality base of Basin's worst-plus letter has been a taxation.
+It's said that a trace of 2 billion dollars in Russia's president lead to Russia Wladimir Putin, versus a friend of children, a famous cellist.
+Now there will be a lot of rich people out there and others who wait nervous in the next publication, and on the next higher document.
+Now this sounds like the tinial pymah or a John-grisham-like ad.
+It seems to be crying from you, ordinary people.
+Why would we do that?
+But the truth is, if the rich and powerful able to put their wealth into tax levels, not all pay off their tax, it means less money for important public services like health services, education.
+And that's what we're all concerned with.
+For my organization, global, Witness, these are the Enthic phenomena.
+In all the world, media and politicians are putting together, like secret taxes, are used to hide some of the things that are used to happen, to reduce their wealth -- something that we've been able to discuss for 10 years and ethnicize.
+I think a lot of people think the whole world is very confused, and it's hard to understand how these taxes are working.
+I always think of it like a rjoking.
+So you have a company in another company, which is almost impossible to understand who's behind these forms.
+It can be very difficult for being a ticket and tax-minded, and for journalists or to really understand civil society, what's the thing.
+I think it's also interesting that there are less reports in the United States than that.
+Because, of course, there were no prominent Americans who were in this Enthic, this sculpture came out.
+Now, it's not like that there's no rich American who are putting their wealth into taxoaseing.
+But by the principle, after that taxoase, Mossackon fseca less American customers working.
+If we had a piece of data from Cayman Islands or even from Delaware, Wyoming or Nevada, we would see a lot more cases and examples that connect the connections in the U.S.
+In fact, it's so much less information you need in some U.S., less information to give you less information to create a company than they can know a library.
+This kind of broader type in the United States has allowed school employees to enlist school children.
+It allows concrete to end especially in randomness.
+It's this kind of behavior that's happening all of us.
+So here in global Witness, we wanted to figure out what all this looks like in practice.
+How does this actually work?
+So we sent a discovery in the office of 13 lawyer companies in Manhattan.
+Our lesson was that when African ministers wanted to get money in the United States to buy a house, a squat of desire to buy a private Soviet.
+We really shocked it all the way to an lawyer does our experience as he could make money.
+These were all pre-doughs. None of these lawyers took us as a client, and of course no one was surrounded with money, but it shows the problem with this system.
+It's also important to think of this as individuals.
+It's not about a single lawyer who's spreads in our covered beam.
+It's not about a single tipping policy, turning into a scandal.
+It's about how the system works, by corruption, taxation, poverty and instability.
+And to cope with that, we have to change the game.
+We need to change the rules to make these kinds of behaviors happen.
+This may seem very obated, as though we could not do anything about it; as though there's always going to be the rich and the richness.
+But as a optimism I see through the fact that there's a lot of stuff happening.
+We've seen more transparency years ago, which is true for the owner of companies.
+The theme of the British prime ministers found David Cameron on the G8i apple in Northern Ireland.
+Since then, the E.U. has been able to do with central feet at a national level, who really put our companies in Europe, and who controls them.
+One of the things that is insecure is that the United States is going to go after that.
+And both parties have brought a laws into both parliament chambers, but this is not the progress that we'd like to see.
+We would really like to see how this Panamathic, this giant glimpse of the world's tax taxation is used as a means to worry about transparence in the United States and global transparency.
+For us, globalization is a moment for change.
+We need ordinary people who are angry when they see how other people can hide their true identity.
+We need leaders in business, who stand up and say, "We're not good for business."
+We need politicians who recognize the problem and replace it by laws of reducing these kinds of inflection.
+We can zoom in on the dragged, which is now a kind of taxation of taxation, corruption, and make money laundering finally.
+That's the story I did almost do, and in the suitcase of a red Mazda.
+One day after I finished my design school, I made a back-of-the-box.
+A guy in the red Mazda stayed there, and he looked at my things.
+He bought one of my artworks.
+He was alone in the city, and he was making a road trip across the entire country, and then he would go to peacekeepers.
+I invited him to a beer. He told me how he wanted to change the world.
+It was late. I was a tired.
+While we paid the calculation, I made the mistakes to ask him, "Where are you going to sleep at night?"
+He did the whole thing worse: "I don't know yet."
+And I thought, "Oh, man!
+So what do I do?"
+Who don't know the situation?
+I now had to offer a sleep in sleep?
+But I just met him. He said, he would go to peace, but I don't know if he really want to land this in a suitcase room.
+This is a little squatter room.
+Then I heard myself saying, "I've got an airt. You can sleep in my living room."
+A voice in my head said, "Are you?"
+And at night I was in bed, I died the ceiling and I thought, "Oh, man! What did I dry up there?
+A wild aliens in my living room.
+What if he's crazy?"
+I got a fear that I got out of bed, up on the door, sat down on the door, and then I decided to have my bedroom door.
+But he wasn't crazy.
+We're still in contact.
+The art, the piece that he bought from me was in his classroom today. He's now teachers.
+This was my first experience as a Gastgeber. It changed my perspective completely.
+Maybe the people who sold me in childhood were actually sold as foreign, actually, friends who were waiting to be discovered?
+People on my air dissent, normal to me, when I moved to San Francisco, I took the air damage.
+Let's take a leap, two years later...
+I'm not working, almost reverberation, and my habit is going to be able to increase rent.
+I learned that there was a design conference in the city, all hotels were made of.
+I think creativity can make it fear of fun.
+I wrote my best friend and new compassion in Brian Chesky: "Brian, I thought we could make something like our apartment, our designer-and-a-half-half-half-half-half-half-half-half-half-half-half ink, diabetes-a-half in the morning, including a Matr.
+AB: Hah!
+We made a website, and we set up a website called breakfast," and the cluster of mobile paper, and it was a piece of paper that was called the fourth-powered.
+Three happy Gäste for us in 20 dollars on the air floor on wooden floor.
+They found it great, and we also.
+I'm sure our cheese sandwich savanry schmeckts completely different because we made them for our crops.
+We sat through all the city with them all the time. When we first walked through our last hospitality and the door, the door fell in, Brian and I got to us.
+Are we just discovered that we could find new friends at the same time and pay our rent?
+Things came to the role.
+My former living, Nate leadczyk, decided we were going to a developers.
+We wanted to find out if you could make a business conceptual tool out of it.
+So we presented with investors: we want to create a website where people are building public images from their private, their bedrooms, their B, their bedrooms, their room, the species that come by.
+They can invite the Internet to people in their rate to them.
+That's going to be the next big thing!"
+We were waiting for the rocket to be artificial.
+But they don't.
+No one who's even more drum, would be invested in a business that allows strangers to sleep in apartments of other people.
+Why?
+Because when we've learned that all the strangers are dangerous.
+If you've got a problem, you're focusing on things that you can do well. We could design.
+At the art ofaka, we learned that design is much more than just looking and Hapamine -- it's the total experience.
+We've learned how to design objects, but now we wanted to create huge trust between people who had never met before.
+Can design do this?
+Is it possible to create trust?
+I want to give you a copy of what degrees of trust we've been doing.
+It's a 30-second experiment. It's force you out of your comfort zone.
+You can get up if you're ready.
+Take your cell phone.
+Now I want you tosper your cell phone.
+Give your tinker phone to your left seats.
+This quiet trip that you're feeling now -- -- this is exactly what's going on when you open the door for the first time.
+Because the only thing that is personal than your cell phone is your home.
+You can't just read your text, you can see your bedroom, your kitchen, your toilet.
+How does it feel to hold cell phone a stranger in your hands?
+Most of them feel responsibility.
+So most people feel like they're near where they're above.
+Just that way, it can exist our company.
+Just like, who's now got Al Gore's cell phone?
+Could you just pay Twitter that it made for presidents?
+You can give back the cell phones now.
+Now, you've experienced what kind of trust we want to build. I want to talk about some discoveries.
+What if we had a little detail about this experiment?
+What if your neighbor had introduced himself to his name, if he told where he comes from, how his dog or his children?
+Imagine if you had 150 convicted that everybody said, "He can really get a very good cell phone suck!"
+How would you feel if you had to spend your cell phone?
+Because a good thing was doing well-known feedback system, mainly.
+We've also done some things wrong.
+It was hard for people to get negatives.
+And we decided to wait until both the host and hospitality had shifted their prejudice before we put them online.
+We've discovered something new last week.
+We did a study at Stanford. We studied how likely it is to trust people, depending on how similar to them, how much they're in age and housing.
+We trust that the people that are most similar to us.
+The more differences we find, the less trust we trust.
+This is a natural socialness.
+walk it will add if you add the reputation of a person -- in our case, through judge.
+If you have less than three judges, nothing changes.
+But you have more than 10, and all changes.
+A good call is a good reputation.
+So the right design can help us overcome some of our deep ancestors.
+And we also learned that trust is dependent on how much you're going to deal with it.
+You can see here the reaction to the first message of a hospitality.
+If you tell too little, like, "Hi!", you don't get any answer.
+Narrator: For example, I have a problem with my mother -- -- I won't be suppose the requests is too much.
+So there's a optimal degree of openness, like, "To't do art in your apartment! I do vacation."
+How do you create it with design, make this degree of openness?
+We use the size of the text to recommend the right amount of text, and we also give people a sense of what things should be.
+Our entire company built on hope that the right design can help help teach the people to overcome our prejudices.
+Where we didn't take the idea of what was that the great amount of people who were willing to take it off.
+Here you see how many people are taking our supply in a way that we do.
+You can see three things.
+One: incredibly lucky.
+Second, our team's work.
+Third: one need not been covered before.
+It's very good for us, business is going on.
+Of course there are times when it's not smooth, too.
+There were Gäste that believed party with or apartments.
+They have re-engineered anyone in the rain.
+At the beginning of the project, I was working with clients, all the phone calls came directly to my cell phone.
+I was at the front of the line when trust was broken.
+There's nothing worse than that phone call, and it's something I think if I just think about it.
+The delusion that you heard in the voice was and will always be our biggest motivation to improve us.
+Luckily, the injected million ancestor had a very largely adapted problem.
+Because people are familiar with each other.
+If trust works, you can create miracles.
+One of our Gäste was doing vacation in Uruguay, where he was doing a heart attack.
+His Gastgeber went to the hospital.
+He even gave blood to have surgery.
+Here's his pity.
+"The desks tend to be able to travel for travels because of the activity of heart attack.
+The area is beautiful, and with enough hospitals.
+Yes, it's the Alejandra and the most challenging protection that saves you, even though you don't really know it.
+They drive a car in their own hospital when you die, and wait as you get a bypass.
+Because they don't want one to feel it, they bring books over.
+They actually stay longer without a extra dimensions.
+I can recommend it!"
+And of course, not everybody's going to get off.
+But these relationships behind the money station are exactly what you want to achieve with the Share.
+When I first stumbled across that term, I asked myself.
+How does the idea of a part of money fit together?
+It's about an economic trade.
+But it's just what we call "the industry" is not just about the thing.
+Share Economy describes a hand that talks about human relationships.
+People put a part of it, and it changes everything.
+Now, if you compare verreists today, it's fast food. It's efficient and reliable, but it's less authentic.
+But what if travels would be a rich Buffet in local local one?
+What if, in every place you were to visit, a group of waiting for a very green one, show a touch and a bar neighborhood that you've never heard of.
+Is it if you could learn cooking from a five-year-old stars?
+These days, apartments are designed by the private sphere.
+What if we were to create apartments from the bottom of the other?
+What would that look like?
+What if cities would assume the idea of common parts?
+I imagine cities that allow us to allow us to do society and relationships instead of a way of resemblance and purpose.
+In South Korea, in South Korea, this project was already started, and a lot of parking lots of parking lots of people heard about government.
+Now, people were offered. student who were looking for a lead.
+People were raised, and they were launched in incubator programs.
+To fund new economy in the Share for a team -- over our platform alone, we're going to be about seven85,000 people in 191 countries in a stranger or even a self-organization.
+So the idea doesn't seem as crazy as it was taught us.
+We didn't re-engineer the wheel.
+And hospitality gave it a long time ago.
+There were similar website before us.
+So why has our work?
+We saw happiness and timing that we realized that you can find the pieces of trust the right design.
+Design can help us overcome deeply engulf of adolescence.
+I think that's amazing.
+Just overwhelming.
+I have to think about it every time a red Mazda goes by me.
+We know that design can't solve any problem.
+But if it could help us, if it had this big impact, I wonder what we can do for design.
+Thank you.
+What do you think about me?
+Was there a liar? A expert?
+Maybe even a sister.
+Or does a brain-class brain cords, a terrorist?
+Or, just a reversal of the airport.
+That's actually true.
+I'm not going to give you the blame for your negative people.
+So the media is looking like me.
+One study made that 80 percent of the reports is over Islam and Muslims.
+studies show you, Americans will tell you that most people would not know Muslims.
+And if you don't talk to people about their own individual driver.
+For those who have never met Muslims, it's great to learn.
+I'll tell you who I am.
+I'm a mother, a coffee love -- double-scale, extra extraterrestrial.
+I am an introvert.
+One-fearness, anatiker.
+And I'm a practiced, spiritual Muslims.
+But not like Lady Gaga singing, because baby, I wasn't born that way.
+I decided to do this.
+When I was 17 years old, I met the decision to come out.
+No, not as a homosex, as some of my friends, but as Muslim, I decided to wear the obstacle, to my head a bit.
+My feminist friends were very upset, "Why are you underflies?"
+The fun thing was, for me, was that it was a feminist chance for me, and that was the pressure I felt as a 17-year-old, feeling a perfect and un-old beautyside.
+I didn't just take the belief of my parents.
+I rescued with the Koran.
+I read him, he thought, you know, he thought, sat down, and I thought, you know, two years ago.
+My relationship with God wasn't love on the first look.
+She was trust and slowly commitment to the copying of the readers with anyone.
+The beauty of rhythms sometimes brings me to wine.
+I know myself in it. I feel that God knows me.
+Did you ever feel someone's mind you're completely understand, and yet somehow?
+That's how it feels.
+I later got married, and I started to get married, and I started to get married as good Egyptian, my career as an engineer.
+I had later, after my shirt, a child, and basically lived the "gyptic dream."
+Then the terrible morning in September 2001.
+Many of you may remember exactly where they were there.
+I was sitting in my kitchen, at the end of the screen, looking at the screen and saw the word "Eilungarian."
+There was smoke, planes flew in buildings, people sat down and they would sat in the buildings.
+What was that?
+Is it a crash?
+It's a technical disorder?
+My shock was transformed quickly.
+Why would you do that?
+I changed television and I heard, "mus terrorist terrorists" -- "The Islam is," "The Lording... " "NoNo... "NoNo... "NoNo... "NoNo... "NoNo... "NoNo... "NoNoNo...
+Oh, my God.
+It's not only been attacked my country, but the fact that really turned me from another citizen into a citizen.
+And the same day, we had to go through the West in the United States to move to a new city and start building students.
+I remember being -- as we were driving in my seats -- hardly far in my seat, and I was watching the first time being a Muslims.
+We moved to this night in our apartment, in a new city where it felt like a completely different world.
+And then it heard and I read the wars of the national Islamic organizations saying, "Sed it carefully," "Sed carefully," "Borted areas," "snol on the "smetic."
+I stayed inside the week.
+Then it was blown Friday every week -- the day that Muslims blew up to the defuse.
+And again, the waiting said, "Don't look at this first Friday that might be a goal."
+I looked at the fathers instead of a report.
+The feelings were madly entic and I heard of attacks on Muslims or of people who thought were free, sasnorted and saped by Muslims.
+There were really Brand based on mosques.
+I thought we should stay home.
+But something didn't feel right.
+Because the people who were attacked this country were attacked our country.
+I figured out the anger of the people who rely on terrorists.
+Imagine! I was angry too.
+It's not easy to explain themselves differently.
+I have nothing against questions, I love questions.
+They are the excuses that are hard.
+We can actually hear people saying, "There's a problem in this country -- Muslims are hearing it.
+When are we going to get rid of them?"
+Some people want Muslims, and they want to finish Muslims and mosques.
+They talk about my community like a tumor in the body of America.
+It's just the question: are we suck or good?
+You know, it's a great tumor removed when it comes to a whole, and it's a great tumor you just think of as observation.
+The alternative isn't there, because the question is wrong.
+Muslims, as all Americans, not tumor in the body of the United States, but are a living organ.
+Thank you.
+Muslims are inventors and teachers, firsthels and oil spills.
+Will the molecular America make sure that we make?
+It may hold a few parking lots, but it doesn't stop the terrorist.
+The regular visit led to a mosque, looking at other people tolerant faith, and showing greater citizens of citizens.
+And as I was a leader in Washington, DC recently told me, people aren't actually radically radically radicalized.
+They're radically radically radically radically radically radically radically radically radically radically radically radically radically radicallyized in their basement or in bed in front of a computer.
+In the radicalization, you've noticed that it starts online. It's the first thing that the person in their community is cut off their communities, even from their family, can make the extreme group of brain systems so that the person believed that the terrorists are the true Muslims, and everybody who has gotten their behavior and their behaviors and their faith, or faith.
+If we want to prevent radius, we need to stop people from going to the mosque.
+Some of them are still going to be a claim that Islam is a violent religion.
+Finally, a group like the war on her brother-graders have started with the Koran.
+As a Muslims, as a mother, as a person, I think we need to do everything we need to stop a group like the pee.
+But we would put ourselves in their notions of the world if one is a 1.3 billion in 4.64.6.
+Thank you.
+The IS has so much to do with Islam, like the Ku-Kluxlanlanlan.
+Both groups have found their cosmology on their holy book.
+But if you look at them, it doesn't take them to what they read in their sacred writer.
+It's their brutality that makes them read these things into writing.
+And a wonderful story told me recently.
+A girl came to him because she was ugly, to close to the transition.
+I was really surprised, and I asked him if she was in contact with radical religious leaders.
+He said the problem was that the opposite. Every spirit that she talked to the sulfur and said she knew her that her anger, her re-oracle, she would only bring her to the injustice in the world, she would only bring her to trouble.
+From nothing inspired and something that their anger would have given sense of, she was a major goal for the instrumentalization through extremes that forget her.
+This is the connection to God and to her community.
+Instead of guilty for her anger, he showed her constructed ways for a real change in the world.
+What she learned in the mosque, she pulled before they got to the pee.
+This was a glimpse of how Islamophng me and my family's family's Islam.
+But how does it look like normal Americans?
+How does it look like everybody else does?
+What does the 24-hour consumption look like every day of our democracy, on our thoughts?
+A study -- actually, several neurology scientific studies -- let's start by saying, if we're real at least three things.
+We accept rather a authoritarian government system, a form and judge.
+A study shows: if you took tests in the negative news that was reported to Muslims, more based on military attacks, and the rise of the United States as a result of the Muslims Muslims.
+It's not just a academic problem.
+If you look, when the mood took up -- in the Muslims and 2013 -- this happened three times, but never in context of terror attacks.
+It happened in the Iraq War, and it happened twice the experiment.
+So Islamophphobia isn't simply the natural response to Mosle, as I expected it.
+It can actually be a tool for public services to be a foundation of free society that has linked and good citizens.
+The way we think about Muslims is an early warning.
+We may feel it as the first one, but the toxic air of fear is frightened.
+In the instructions of collective, it's not just about explaining itself.
+Deah and his wife were a young, married pair of Hill, North Carolina, where they were going to school.
+Deah was a sportsman.
+He was a perfect medical defensive, a lot promise...
+His sister would tell me that he was the most sweet, generous human being that she knew.
+She went there and he showed her life. She was amazing.
+"What has it become like a brothers like this young man?"
+Just a few weeks after Suzanne's visit at Suzanne and his wife, Craig Stephen Hick, she had mornings, as well as the sister of Yusor, Raz's sister, Razean, who was there in the afternoon, and he was in her apartment, after he had gotten on Muslim direction on his Facebook page.
+He shot eight times.
+Now, Fanism can't be moral, but it can also be deadly.
+So back to the beginning.
+What happened after 9/11?
+Are we going to the mosque, or are we going to make sure that we were going to go home?
+We talked and it wasn't a very easy decision, because it was about the America that we want to leave our children, one that would control us through fear, or one that we could get free from our religion.
+We decided to go for the mosque.
+With our son, we drove, we went to the high pressures with high pressures.
+I took him out, I moved my shoes out, went to the prayer, and what I saw was screaming me.
+The hall was filled with it.
+Then the one an assumption was to be a re-construct, and our people who were willing to come together because half of the people who were Christians, Buddhist, Buddhist, atheists, Gs and non-zero-sumness, who hadn't been able to get to get involved in us.
+And I'm broken together at that moment.
+These people were there because they were a courage and compassion and aurteil.
+What are you going to choose?
+What are you going to do right now, frightened and choose for the fantasy?
+Are you going to make sure that you're going to go?
+Or will you join the ones that mean: we're better than those.
+Thank you very much.
+Thank you.
+Helen Walters: So Dalia, you seem to have met a nerve.
+But I wonder what to say to you, those who may say you may say that you think you're a TEDTalk, a thinker with deep-sea thinker, in a noble-skn factory, so one exception and not the rule.
+What would you say to those people?
+Dalia Mogad: I wouldn't tell you. I'm totally normal.
+I'm not an exception.
+My story is not unusual.
+I'm so common as it allows you to.
+If you look around the world Muslims, and I did the biggest study of ever done in Muslims all over the world, people would like to normal things.
+They want prosperity for the family; they want to live in peace.
+So I'm not an exception in any way.
+When people appear like a rule to the rule, the rule has been broken, and they're not the exception to the policy.
+HW: Thank you. Dalia Mogad.
+What started as a platform for Basan, is about to become a billion business in front of it.
+control, environmental surveillance, photographs, films and journalism: This is some of the possible applications for free. M. M. M. Ms are made this into research worldwide.
+Before air-made air-sand-a-half years ago, an autonomous fet, an autonomous frog engine built in the flies Centre in France in France, a six-foot tower from 500 meters ago, they started flying off with Seila.
+By focusing on each other's high speed and accelerating space.
+They can also build self-replicating structures.
+They've learned to wear trucks, to deal with  Turbulence to respond and respond to nature generally.
+Today we want to show you a few of our new projects.
+Our goal is to prevent the limits of the autonomous flight.
+So a system of autonomously does have to know where in the room the mobile objects are.
+In our lab, we often use external cameras to find objects, and that allows us to focus on the quick development of dynamic development.
+For today's demonstration, we're using a new localization of reversal technology, a culosis of our lab.
+There's no external cameras.
+Every flight machine has internal sensors to determine the position in the room, computations from what the machine should do.
+There are command commands on the highest level. "P. "P. and open."
+This is what's called a tail re-in-the-wall.
+It's a aircraft aircraft trying to beat two flies with a force.
+As other stars are in front of it, in the flight, far more efficient, much more efficient than helicopters in all of their variations.
+Unlike most starving crops, it can float through a lot of oil benefits and land, and it's very vielseitig.
+However, there's always a cliche side.
+A limitation of tail star is they're sensitive to Turbulence like wind.
+We're developing new taxities and algorithms to improve this.
+The idea is that the flight device doesn't matter what the optimal position can be improved, and it can be improved by evil.
+Okay.
+As we often ask ourselves, we're very fundamentally asking the key questions about the core of the thing.
+So, for example, would be a question like this: what is the small impossible number of things to controlled parts?
+There are practical reasons to want to know the answer to this question.
+helicopters are about as machines familiar with thousands of pieces that are related to hurt you.
+Over decades ago, a dead pilot's able to fly away, the only two squatters had a prop and a tail block of a tail.
+We discovered in the longest that fly only works with one.
+This is the monosopin, the most simple, simple flight-powered flight. It was invented just before the monolithic flight.
+It just has a trawling part, a propeller.
+No class, California and cross-section, no more resemblance or control, just a prophedron.
+Although it's mechanically simple, in interior life, so it can stable it, and move around in the room all over the place.
+Yet it doesn't really mean it about the elegant algorithm in the squirt algorithm, so to fly it, I have to throw it right.
+The probability of suck it right to see me, is humbling, so I'm going to show you a video that was filmed last night.
+If the monospin is a exercise in gene, this machine here, the  Omniopter with his eight Propeller, an practice flow in over time.
+What about all this flow?
+You can see that it's dead, symmetrical.
+That's why he's ambivalent in his engineering.
+And so this gets the audacity to get the audacity.
+It turns out that in space at all, no matter what direction it's rotated, or even like it's red.
+Of course, he's complex, mostly in the realm of interactive rivers of eight Props.
+Some of them are imaginable in models that will learn directly as flies.
+Let's see.
+If the machines are supposed to be part of our everyday life, they have to be extremely safe and reliable.
+This machine is made up of two separateprop machines.
+This is a clock in the clock.
+And the other one is slowed down the clock.
+If you build them together, they behave like high-speed high-speed helicopter.
+But a little wrong -- a motor or a propeller, or electronics or a a cliche -- can also go on a machine if it's limited.
+We're going to demonstrate that by going to half a trip.
+The last demonstration of synthetic swarms.
+The large number of autonomous, koornorassic units enabled a paleoore of aesthetic expression.
+We were taking Micro Micro Microopteropter -- each one of them as a Sche bread -- and instead they were taking local technologies and algorithms.
+Every unit knows where it is in the room, and is self-organizing. That's not a top line.
+Hopefully you're going to try to think of these demos of new revolutionary ideas.
+The particularly safe machine would like to be a Broadway-like wing light.
+Of course it's difficult to predict the influence of this technology.
+For guys, the lung is in development and the creativity.
+It's as memory, as beautiful and ecstasy our universe, and it allows you to shape the creative minds that it allows you to form in such spectacular ways.
+The fact that this technology has such massive and economic potential is that it's a bag on which.
+Thank you very much.
+1.3 billion years ago, orbiting into a very distant galaxy, two black holes were always based faster, and they turned the mass three suns into a tenth of a second.
+These short moments, they sat brighter than all stars together in all galaxies of the universe, for all of us.
+It was a very big space.
+But they didn't put energy into the form of light.
+We finally talk about black holes.
+All of the energy was self-assembled into space-time, and the universe exploded in gravitational waves.
+Or we first do this, we start becoming a time.
+1.3 billion years ago, life had just been sex and 1.3 billion years ago.
+Since then, the Earth has taught some of the Earth: corals, fish, plants, oxygen, people and -- God, even the Internet.
+About 25 years ago, a group of courage -- Rai creatures, Kip Thor and Ronaldever came from Caltech to build a huge laser detector, looking for gravity, which emerges about a black holes.
+Most of them thought they were crazy.
+But enough people knew them as crazy genius, so that the U.S. National Science Foundation gave this idea.
+After 10 years of pregnant development, concepts and extremely hard labor made them a lot of hard labor: the laser-scale gravitational state wave.
+In the consequence of LIGO, the recognition of far-time power was improved.
+So now you call it Advanced LIGO.
+At the beginning of September 2015GO, the last test started to explain some smaller, hard-to-day problems.
+On 14 14th of September, only a few days after the detectonics of the detectonic, the gravitational waves of both lions of the black holes dragged through the Earth.
+They went through you and me.
+And also through the detector.
+Scott Hughes: Only two moments in my life were emotional than this one.
+The birth of my daughter.
+And the father of my sick father.
+Basically, these were the fruits of my life.
+All that I've been working for, which is no longer science fiction -- AA's had a very good friend, AA and Scott Hughes, theoretical physicists at MIT, looked at gravity for 23 years, where they've been looking at gravity and their observaal signals like LIGO signals, but what are gravity?
+A gravitational wave is a sound in the midst of space and time.
+And then, as we move across the wave of space and its entire content, we're going to have to go into a direction and re-eyed space.
+There's a lot of talk about dopamine in course, with relativity, to relativity, often a silly dance.
+"Deuld and dive, deplete and come."
+The problem is that gravitational waves are extremely affordable, even ridiculous.
+On 14 14th, for example, every one of us was raised on the waves and stolen the average person, and the average person was wore 10 to three.
+That means 20 zeros after the community, followed by a Dar Dar explains the chance for crazy.
+With a five-mile laser, and that's absurd, it has to be already less than a thousand-inch atomic-like atom. That's a big deal.
+At the end of his classic text, gravity goes across.
+And I described Kip Thors, a compassion LI re-eyed community, who said, "The way to treat these technological problems are enormously violent.
+But physicists are inventord, and with the public support, all the obstacles are thrown over caps."
+Thorned this in 1973, 42 years before his success.
+Back to LIGO. Scott would like to say that LIGO is more than one eye is.
+I want to explain what that means.
+And the kind of light has a wavelength that's much smaller than the things around us: face suits, the size of your cell phone.
+That's quite practical, because you can make a picture or a map of things by coming out of several points of light around you.
+It's different in sound.
+Listening sound has a wavelength of about 15 meters.
+So, it's very hard to actually make an image of things that tell you very much.
+The face of your child is about.
+Instead, we're laughing at certain traits like clay and Italy, rhythm and loud, to close that story down.
+"As just talks Alice."
+"And Bob is bribed."
+"Dmer Bob."
+Same thing for gravitational waves.
+We can't make them simple pictures of objects in space.
+But by looking at changes in the '60s and writing frequency of waves, we can listen to their stories.
+At least for LIGO measured the frequency of the listen.
+So if we turn waves into sound, we can literally hear the whole thing.
+The gravitational can tell us about the collision two about the collisions of black holes where my colleague is going on Scott for a long time.
+SH: Two black, not sludge, just suck.
+The two bodies very quickly, you heard the same quote with an extra clay, and it sounds like that:w-wo-wp-wo-w.
+It's a kind of a motion-based, spit into the waves.
+AA: On September 14 14th of September -- one of my date -- at least I never forget -- the sounds of law, the sound knows what it's like -- SHcening, recognizes it as sound -- SHing from -- SHers from any black holes -- from any black holes that are about as fast as the tribes are doing.
+AA: Let's think about what that means.
+Two black holes that are interconnections in space with a mass of 29 sun, a mass of mass of 36 times a second, whereas hundreds of times a second before they work.
+Imagine that forces.
+Fantastic.
+And we know about it because we heard it.
+That's what LIGO are.
+Unfortunately, a whole new way of exploring space as it never had before.
+So, in this way, we can listen to space and hear the invisible.
+A lot of things in space, we can basically see -- not basically.
+A supernova, for example. I'd like to know why massive stars exploded in supernovae.
+They're very useful. From those, we've learned a lot about space.
+The exciting physical activity is that the core of the tens of thousands of miles of iron, carbon and a half miles away is hidden.
+We'll never see through it because these Pakistan are Pakistan.
+gravitational waves of iron, as if it were transparent. The Big Bang: I would like to explore the first few minutes of space, but we will never see them because the Big Bang is being covered by its own neighbor.
+With gravitational waves, it should be possible to look back up to the beginning.
+And I think the most important thing is, I'm optimistic that in space we've never seen, that we'll never see and we don't even have any idea about it. Things that we're just hearing.
+In fact, LIGO found the first thing we didn't expect.
+My colleague at MIT, Matt Evansul, an important member of the LIGO project, says this topic: ME: the kinds of stars that produce black holes, as LIGO were seen, are the dinosaurs of the universe.
+They're enormous, ancient historical body, presumption over time. They're kind of black holes that are sort of the so-called "eyed bones for our archaeological work."
+DuringGO, we were allowed to look at a completely different perspective on space, in the emergence of stars and ultimately to come out of this chaos.
+AA: The challenge is, now, as brave as we can.
+Thank you very much. We know how to build great detector and the noise of the cosmos.
+We need ideas for new history -- a whole new generation of extremes on Earth and space.
+Because what could be more beautiful than laughing the Big Bang itself?
+Now, the time is great dreams.
+Don't stop us.
+Thank you.
+I tried some time ago.
+For a year I would say "Yes!" I would say.
+No matter if I was nervous, I was able to get into unpleasant situations, I forced myself to say "Yes."
+I mean, in public, what am I talking about?
+No, but yeah!
+Is it going to be live on TV?
+No, but yeah!
+I want to start with what I was doing to the look at?
+No, no, no, no, but yeah, yeah.
+And one crazy thing happened: that's what happened before I was scared of what I was scared.
+My fear of keeping Red fear -- puffing away my social fear, yelling.
+The power of a word is impressive.
+"Yes," changed my life.
+"Yes changed me."
+But there was a certain time that changed my life deeply unexpectedly, in an unexpected way. It began with a question of my smallpox.
+I have three incredible daughters, adapted, incredibly sophisticated and Emerson, and Emerson, called the incredibly unleavable reasons.
+as if she was a waiter from the south.
+"Look, I need milk for my shit."
+She asked her if I could play with her, when I was on the jump, and I said, "Yes."
+This is the beginning of the beginning.
+A new life of my family. From there, I always sat down with them, whenever you ask me about them, whatever I do, or where I'm going, I always say yes.
+Almost. I'm not perfect, but I'm very trying.
+It has a magical effect on me, on my children, on our family.
+But it also has an amazing side effect, which I really understood, that the "Yes" thing about my career, is that it was completely re-sembling my career with my children.
+I have a real dream job.
+I'm a author. I think he's a squeaking of things to life.
+The dream job.
+No.
+I'm a Titan.
+The dream job.
+I'm making television. I'm making televisions.
+I make TV, big style.
+In this TV season I'm responsible for 70 hours to get out to the world.
+Four TV programs, 70 hours television. Three to four show are at the same time in production.
+Each show of hundreds of jobs that don't exist before that.
+The budget for a television can be between three and six million dollars.
+Let's tell five.
+A new one, nine days every four-day show, four million dollars, four TV programs, 70 TV, three TVs, three people at the same time, 16 Episode: Grey's, 21 Episode: Scns, 15 Episode: How How on Away, 10 to 70-years, 70-olds, 70-olds, 70-olds, 70-day television, 70-year-olds, 70-olds, 70-olds, 70-old television, 70 hours, 70-olds, 70-old television;
+350 million dollars for a season.
+In America, my TV series goes on a thunder day.
+My series around the world are running in 256 areas in 67 languages for 30 million auditory audience.
+My brain is global, and 45 of these 70 TV studies that I have self-replicating, not only, so I have to find, really silent time, to relegate my fans to savor my stories and tell my stories.
+Four TVs, 70 hours of television, three, sometimes four, four, show at the same time, 350 million dollars, campfire, running around the world.
+You know who's doing it yet?
+No one, I'm a Titan.
+The dream job.
+I don't want to impress you with that.
+I say it because I know what you think when the word falls.
+I'm telling you all of you who work so hard, whether you're a company or a classroom or a business, or a business room, when I talk about work so that you don't just type around the computer and fantastic it, and I'm saying it's not about a dream job.
+It's all a job, everything, all reality, everything, blood, everything -- no tears.
+I work a lot, hard, and I love it.
+When I'm deep in work, there's no other feeling.
+My work always creates a country from nothing.
+It's like I put troops on my hand, when I was on a screen.
+As a high audio, you walked a marathon.
+You feel like Beyoncé.
+And all that at the same time.
+I love working.
+It's creative, mechanical, hard and hard, funny, clinically and tired, cruel, cruel, and sludge, and best of that is the sum.
+There's a change in me when the work is fine.
+A sum of my head starts in my head, and it grows and grows, and the sums listened to how a living road I could drive for forever.
+Many people take the money that I explain to you, that I talk about writing is the writing I'm ready to go.
+Don't get me wrong, it does.
+But the sum -- when I started working with TV production, I started working and building it, building it, I discovered this sum, this amount of energy.
+The sum is more than writing.
+The sum of action is and activity. The sum is a kite.
+The sum is music. The sum is light and air.
+The sum of God's voice in my ear.
+And if you have a sum like this, you can't help but a different way to scale.
+The feeling, not different from the size of the size, any size of the price, can't matter what price.
+That's called the sum.
+Or maybe it means being a Workaholic.
+Maybe it's genius.
+Maybe it's called ego.
+Maybe it's fear of failures.
+I don't know.
+I just know I'm not made for failure, and I only know I love the sum.
+I just want to tell you, I'm a Titan, and I know I don't want to ask it.
+To put one clear: the more successful I'm, the more 11ths, the more Episode boundaries, the more things that the more things that I see at the same time, the more I write history, the more expectations.
+The more I work to be successful, the more I have to work.
+And what did I say about work?
+I love work, right?
+The country I'm making is the chaotic, the marathon I'm walking, the army, the canvas, the high clay, the sum, the sum, the sum, the sum, the sum.
+I like this money. I love the sum.
+I need the sum. I'm the sum.
+I only want to make this sum?
+And then the sums stop.
+Over work, more sophisticated, exaggerated, pushed over.
+The sum of the sum.
+Now my three daughters are used to the truth that their mom is a single job.
+Harper told people, "My mom won't be there, but you can write my Nanny."
+And Emerson says, "Look, I want to go to Shondaand."
+It's the kids of a Titan.
+They're baby-Tits.
+They were 12, and 1.2 to the sum.
+The sum of the motor drew.
+I didn't love my work anymore. The motor was out.
+The sum wasn't back.
+My sum was broken.
+I used to do the same things as always: the same Titan squirt, 15 hours of working weekends, no race, no depressed, no Titan, no shits, no whole heart-to-sense eyes, whatever.
+But there was no sums.
+I was silence.
+Four TV programs, 70 hours, three productions at the same time, sometimes four.
+Four TV programs, 70 hours, three production production at the same time...
+I was the perfect Titan.
+I was a front-to-shoulder.
+Everything was gray, I just had no fun.
+And that was my life.
+Everything I did.
+I was the summen, and the sum of the money I was.
+So what do you do when you do what you do, the work that you like to leave?
+I know some people like, "Hey, just suck you out, shit authoritan."
+But you know, you do it, you make it, you work, what you do, you do, a teacher, to be a banker, to be a Bill Gate, you love one, you love another, and you just give a money to the sum of the sum, and you know the sum of the sum, how many times if you feel the sum of who you're doing, who you're doing?
+What are you?
+What am I?
+I still give a Titan?
+If the song stopped my heart, I can survive in silence?
+And then my "Sirn states" would ask me a question.
+I'm on the way outside, late, and she says, "Mom, do you like playing?"
+And I'm like, no, as a result, two things are aware of.
+One, I have to say yes to everything and second, they didn't call me "dism."
+They're not called any more "Bay."
+When did this happen?
+I miss it when I'm Titan and looking at my sum, and here's all changing before my own eyes.
+And so she says, "Mom, do you like playing?"
+And I say, "Yes."
+That's not special.
+We're playing, and we're laughing at this, we're reading a lot, and I'm reading all dramatic from the book "Everybody's "Everybodyrops."
+Nothing extraordinary.
+But it's figured out because my pain and panic and my microbialness and in the failure of the sum, I can't do anything but just make it.
+I'm focused on it.
+I'm empty.
+The country I make, the marathon, the marathon I run, the army, the canvas, the high clay -- they don't exist anymore.
+All they have is nuclear fingers and moist,zar Voices, crocod voices, and the song is leaving in the Lose, or whatever the girl is going on in "The ice."
+All over is peace and simplicity.
+The air in this place is so scarce that I can barely breathe.
+I can't believe I can breathe.
+play is the opposite of work.
+And I'm happy.
+Something that's in me.
+A mental door goes up and a energy blob comes in.
+And it's not immediately, but it's happening.
+I feel it.
+The sum of the sum goes slowly.
+No full volume, no one's there, not there, it's empty, but it's there.
+Not the sum, but a sum.
+And now I feel like I'm familiar with a magical secret.
+But let's stay with the thing.
+It's love. It's all.
+No magic. No secret. Just love.
+It's something we forget about.
+The sum of money that works, the Titan organ, is just the practice.
+If I ask you who you are, if I tell you who I am, if I write, if I write, and television hours and how functional my brain is, then I forgot what that really is.
+The sum is not power and it's not work-specific.
+It's dependent on the joy.
+The real sum is dependent on love.
+The sum is the stream coming from life.
+The real sum of self is subconscious and peace.
+The real sum of the pressures ignored the history of doing tasks, the expectation and the pressure.
+The real sum is simple and origine.
+The real sum of God is God's voice in my ear, but maybe Godsters me the wrong words, because what God told me that I am an Titan?
+It's just love.
+We all need a little bit more love, much more love.
+Once my kid play with me, I'll say yes.
+I do this to the solid rule so that I can free as aholic of any school.
+It's law, I don't have any choice. I don't have any choice, so I want to hear the sum.
+I wish it would be so simple. I'm not good at play; I don't like it.
+Don't play that way, I'm not really working.
+The truth is going to hurt.
+But I don't like playing it.
+I always work because I love it.
+I'm rather than there on the workplace.
+This one is painful, because what does a human work better than being in a house?
+Well, I.
+Let me say, "Titan."
+I have to have problems.
+That's me really relaxed, is not one of them.
+We're going around in the garden, going back and forth, back and forth.
+We make little dance party.
+We sing and play ball.
+We re-arar bubble.
+I feel adapted and confused.
+I'm always putting my cell phone on my cell phone.
+But it'so.
+My kids show me how to live, and the sum of the universe is moving away.
+I play, until I ask, why did we ever stop playing with the game?
+You can say it! Don't always tell if your kid wants to play with you.
+Maybe you're naive, for me, a day's dreaming.
+You're right, but you can do it too.
+They have time!
+And you know why? They're not a iad-wry, or a Mu-day show.
+Your child finds less interesting than you think.
+It's only 15 minutes.
+My little, the highest-minute thing I'd like to play with me until they have a fall in there, they want to do something different.
+They're wonderful 15 minutes, but only 15 minutes.
+After 15 minutes, a Marie would give me a Marie or cookie.
+And I talked to my tinid-rayer for about 15 minutes long, I am mother of the year.
+It's only 15 minutes, it's not going to take it.
+Every 15 minutes on the piece, even on a bad-go-occup.
+15 minutes on the piece.
+No mobile phone, no laundry, no distraction.
+The day is short: dinner, children are making the children finished.
+But 15 minutes it's in.
+My kids are compassionate, my world. It doesn't have to be the kids, it's true to have a place for his soul.
+It's not about play with your own children. It's about joy.
+It's a "resy" in general.
+Have you to 15 minutes!
+Find out what's doing good.
+Find it out, and hold it.
+I'm not perfect at it. I make them, friends meet, reading books that day enjoy.
+"You're going to play a little bit for everything I gave up when I first got my first TV show, when I was a Titan in education, when I wanted to meet myself and meet more.
+15 minutes on the piece, why not having 15 minutes?
+What can it be wrong about that? Nothing!
+The sum of my leisure came back in my leisure. The sum appears to come back when I don't work.
+It doesn't work without play.
+It takes a while, but after a door, a door opened, the energy starts coming in, and I think I'm in my office again. I hear an unknown me, and I'm listening to my soul, and my soul, led to new ideas, which is what sums up again, and I love my work again.
+I like the money, but I don't love it.
+I don't need it.
+I'm not the sum, the sum is not me -- no more.
+soap bubbles and sticky fingers, dinner with friends.
+Now, this is my sum.
+The sum of life.
+The sum of love.
+The sum of work is part of me, but only one part of it is, and I'm so grateful.
+It's my sne soup that I'm a Titan. I've never seen a Titan, playing a journey to Jerusalem.
+I said yes to work and to do more play.
+And yet, I have everything in the handle. My brain is still functional. My camps are running.
+The more I play, the more happy I are, and my children.
+The more I play, the more I feel good than good mother.
+The more I play, the more I play the more clear the head.
+The more I play, the better I work.
+The more I play, the more I hear, the money I make, the land, the marathon I'm going, the marathon, the troops, the high clay, the sum, the sum, the sum, the other, the sum of life.
+The more I feel this sum, the more unusual, receptaclesing, naked and new life to me, as opposed to the widen Titan -- more I!
+The more I feel that money, the more I know who I am.
+I'm a writer. I think I'm a picture of things, I'm putting you on life.
+That's one job, which means live its dream.
+This is the dream of this jobs.
+Because a dream was supposed to be a little bit dreams.
+I said, "Yes too much work and too much more play.
+Titans don't have a chance here.
+"Do you play?"
+Thank you.
+I'm a neurosurgeon.
+As most of my colleagues have been able to do with human tragedy every day.
+I know how your life can change from one second to another, after a heavy stroke, or a car accident.
+For us neurosurgeon, it's very frustrating for us to find that the brain, unlike other body organs, has a very small skill to heal itself.
+After a heavy injury of the central nervous system, patients often have a sexy disability.
+That's why I became functional neurosin.
+What's a functional neurosurgeon?
+A doctor who is trying to improve neurons through different surgical things.
+Certainly, they've heard about one of the most famous brain systems called "Tietulation," which is implanted by an electron in the brain to influence the circuitry of neurons so that the neurologic function improves.
+It's really an amazing technology. It's improved the destiny of patients with Parkinson's patients who have suffered under a scuba divers and heavy pain.
+But neuroscientation means not neurology.
+The dream function of neurosurgeon is the re-constructive of the brain.
+And I think we're closer to that dream.
+I want to show you that we're very close to it.
+With some help, the brain can help themselves.
+The stories started 15 years ago.
+At the time I was a senior and worked day and night in the emergency room.
+I often figured out what I did with patients with skull-like room.
+You have to imagine, with skull rocks, the brain is getting excited and sludge over.
+To save life, you have to fight your skull.
+You have to sometimes remove a snucky brain mass sometimes.
+Instead of sneoseating the swollen of the brain, we decided to start with jeans well-being, one of my colleagues, a biologist to analyze brain mass.
+What do I mean by that?
+We wanted to grow cells out of that tissue.
+That's not a easy task.
+If you grow up from a tissue piece of tissue, it's equivalent to a very small kids that come out of your families.
+You have to have the right diet, temperature, moisture and the environments to let them go.
+And that's exactly what we needed to do with these cells.
+After many attempt, jeans grew.
+This is what he saw under his microscope.
+That was a big surprise for us.
+Why?
+It looked like a stem cell culture, with large green cells, which are surrounded by small, unsustainable cells.
+You may know from biology science, that stem cells are able to turn cells into each cell type of organism.
+The adult brain has stem cells, but very few. They're in deep, little nutrients of the brain.
+It's surprising to get these kinds of stem cells from the surface tissue in the brain of surgery.
+We did another fascinating observation: normal stem cells are very active -- they share very fast.
+They never die; they're unsterable.
+But these cells were devoted to each other.
+They were slowly divided and they died after a few weeks.
+So we saw a new, odd cell population, which looked like stem cells looked like, but differently grew.
+We needed a long time to understand where they came from.
+You're going to take you from these cells.
+These blue and red cells are called  Doublecortin-positive cells.
+We all have it in our brains.
+They make four percent of our brain cells out of the world.
+They are playing a very important role in our development.
+At the stage of the axe, it's worried about the fold of the brain.
+But why do they stay alive?
+We don't know.
+We believe they're part of the healing brain because we're finding them in higher concentrations of brain restaurants.
+But that's not that safe.
+But one thing is clear -- of these cells, we got our stem cells.
+We're in front of an potential cell source that allows brain to heal.
+We had to prove that.
+So we decided to develop an experiment.
+We wanted to take a piece of brain out of not a single area, and then pull the cells out exactly the same way that jeans did.
+And they are unable to mark them in the brain.
+We've been taking it to the same individual over the same stage.
+We call the autoologist transplantation -- car transplantation.
+One of the first questions we did was, what happens when we put these cells in normal brain tissue, and what will happen if we put the same cells in damaged brain?
+Thank you very much for a help from the help of Profy, Eric routines we could work with monkeys.
+The first scenario we put in a healthy brain, and we see that they were completely disappeared for weeks, when they were re-intentioned, they go back home. The room is already sat down, so they won't go away.
+The second scenario, we put in a injury, and we re-construct the same cells, and now the cells are remained the cells -- they grew up to neurons.
+Here you see what we saw under the microscope.
+These are the cells that have been re-resolution.
+The evidence they show are these little dots. These are the cells that we face in vitro, in culture.
+And of course, we couldn't stop there.
+Do these cells help you get from a monkey to catch up from a brain.
+So we trained monkeys with a manualnessness.
+They had to take food from a tray.
+They did very good.
+When they reached a stable level of economic level, we'd be able to do motor cortex in the motor Cormoto factory.
+And so the monkeys were paralyzed, and they couldn't move their hand anymore.
+And just as it would happen in humans, they would spontaneously slow down to some degree, just like a stroke.
+The patients are paralyzed, and then they try to bring themselves with plastic mechanisms, and they get in some degree, just like the monkeys.
+When we were sure that the monkey had reached the spontaneous recovery, we put his own cells on the ground.
+On the left you see the monkeys that spontaneously recovered.
+It can bring about 40 percent of its original performance from injury to 50 percent.
+It's not that precise and it's not so fast.
+Now, look at the cells that we've been re-construct: the same monkey, two months after reimplantation.
+I can tell you, these were also very exciting for us.
+Since that time, we've found a lot more about these cells.
+We can engineer them and use later.
+We can use them in other neuropathological models, for example, in Parkinson's disease.
+But our dream is still, in people's dream, to implant it.
+I really hope that, soon, I'll show you how the human brain gives you the tools to heal itself.
+Thank you very much.
+Bruno Giussani: Jocenelyne, that's wonderful, I'm sure that just now, several dozen people in the audience, perhaps the majority, think, "I know someone who needs that."
+I'm definitely.
+Of course, the question is, what the biggest obstacles are before you can begin clinical trials on the people.
+Jocen Bloch: The biggest obstacle are the authorities. So, one of those great results, you have to fill up about two kilograms of paper and form to run this kind of studies.
+BG: That's pretty much the brain is very devoid of etc.
+JB: Yeah, but it takes a long time to take a lot of patience and a very professional team.
+BG: Look into the future -- you've been making research and trying to get permission to the clinical trials going, and if you go on to the future, as many years it can come to the hospital, and that therapies will be available.
+JB: That's hard to say.
+First of all, it depends on the random study of clinical trial.
+Will it allow us to start the listener, soon?
+Then you have to do this study with a small group of patients.
+It takes a long time to choose patients to get treatment, and to suggest whether it's useful to make that kind of treatment.
+And then you have to apply that to a multi-touch study.
+You have to really prove first that it's useful before you can offer those treatment to everyone.
+BG: And that's safe, of course. JB: Of course.
+BG: Jocene, thank you for coming to TED, and that you told us.
+BG: Thank you very much.
+democracy.
+We make a big mistake in the West, of course, as a sign of it.
+We see democracy as the fragile plant that is actually in reality, but as aInvent of our society.
+We tend to think of them as a unthinkable thing.
+We believe that capitalism leads to democracy.
+That's not true.
+Lee Kuan Yewapur, and his great descendants in Peking have proven that it's possible to have a re-conservation of capitalism and amazing growth while the politics is completely unthinkable.
+In fact, democracy is a mess with us, in Europe.
+At the beginning of the year when I was a Greek government -- the new Greek government was introduced to me as finance ministers in the European group, that the democratic processes of our country -- not our elections, who were re-engineered to the localstand-as.
+And in that moment, I thought there would probably be no better rights for Lee Kuan Yew Yew, or the community-of-the-art side of China, or some of my friends saying democracy would make if they had anything to change.
+So, at this point, I want to introduce you to an economic model of a real democracy.
+I'm going to ask you to believe that Lee Kuan Yew Yew, the communist Yew, China and even the European group of an Irian group, we might save democracy, but that we need a arrested and a highly ecstatic democracy.
+Because without democracy, our future will become more serious and more artificially using our great, new technologies.
+So, to make a point of waste, I want to point out what is interesting is that we're threatened to see our economy.
+I call it the twin twinsdox.
+A summit is known to you. You know him and see him as the shouldersenberg, the long shadow of his long shadows, Europe and all the world we face.
+We all recognize the schoolenberg.
+But very few people recognize their twins.
+A mountain of beekeepers that are eudatry that are enlightened the self-interests and corporate corporates, but it's a place where it would be productive and bringing income to the schools that could be wearing theenberg and also all the things that humanity need, for example, "to-eudavation."
+I'll give you two numbers.
+In the last three months, in the U.S. and in England, 3.3 trillion dollars were invested together in U.S. and 3.3 trillion dollars, in all the wealth-dollar goods, like industrialized, machines, bureaucrats, streets, green, and so on and so forth.
+3.3 trillion dollars sounds like a lot of money, until you compare it to the  5, trillion dollars to the same countries around the world and around the same countries, and absolutely nothing, except for the stock market prices to move up.
+So, schools and unabounded capital that twins don't form through the usual market mechanical mechanisms.
+The result is stagnation, more than a quarter of the 25-year-olds in the U.S. and Europe workless, and followed a low-level demand for the pesimistic demand that went to be low-income, a low-cost demand for the investors that they do, that even though they don't invest, the saymbiotically-eky father that might be true, if he's growing his son,
+ungewks the circumstances that lead to that depleted thedipus kills him.
+This is my Hader with capitalism.
+And it's a very important way to make it happen, all of this unreviewed capital should be used to improve our lives, to make human talents, and especially the technology that is to fund green technology that is critical for the Earth.
+So, democracy is the solution?
+I think before we continue, what do we understand about democracy?
+Aristotle defined democracy as a society, where the open and the poor are control the government as a majority.
+The Indian democracy decided that, of course, decided to make a lot of them.
+Women, foreign, and of course slavery.
+But it would be a mistake to vision the places of the straight democracy because of these extinct.
+The key thing about the breathe democracy was that they were working poor, and not only did they get the right to them, but more important, and they decide what to give them the right to political compassion with the same balance as the state of state.
+The Indian democracy has not been given long.
+As a candle that's very bright, it's also embracing fast.
+But our liberal crammeds don't have our roots in anti-nepne.
+It's in Magna Carte, where 1688 revolutions, even in the American history.
+As the breathe democracy focuses on the free citizens and the working poor, our liberal democracy based on the values of the Magna Carte, the last set was a sentence for Mr.
+Because liberal democracy came only when a complete sphere of politics and economic processes could be limited to politics, as the economy became limited, so the world of corporate policies became.
+In our ancestors since then, the separation of policy and economics started, an unsustainable struggle between the two is being able to promote the economy and the policy that they've been underlying.
+Do you remember why the politicians aren't as early as they used to?
+It's not because of a generation a DNA.
+It's because you can be in government today and yet not in power, because the power of politics is causing the economy and the areas separated.
+I mentioned my Hader with capitalism.
+When you think about it, it's sort of like a herding of predators that are so sustainable that the animals they feed off, that it remained so remained.
+Similarly, it's similar to the economy that has so far in politics that it can self-replicating the economic crisis itself, the power of the corporate goods, political goods that falls inability to provide economic demand and leaders of business to invest their companies.
+The more successful of capitalism is a "Deingence": the people from democracy, the higher the twins, and the greater waste of human labor and prosperity of humanity.
+If that's obvious, it's obvious that we need to bring the policy and the economy together, and it would be better if the atmosphere is in control, such as the ancient Athens, if it's in the ancient appreciating the slaves and the women and the foreigner.
+This is not a new idea, by the way.
+The maristic left was this idea 100 years ago, and it didn't really work well.
+The teaching from the Soviet Union should be that the working poor should only be able to create a miracle of ancient Athens, without creating new kinds of grossity and waste waste.
+But there's a solution: the poor are creating.
+capitalism does it by replaced reward by automation and robots.
+The problem is that the problem with economics and politics are separated, the largely the twins of twins, the social conflict, the social conflict, the way it's very, very soon -- as I believe in countries, are coming in.
+So we need to re-construct the economy and the policy-plays, and to look at the same time, that we end up democratized the democratic spaces, otherwise we end up in a sick car, which makes movies like a sexx.
+So the question is not whether capitalism has brought together, which is being re-intentioned, is a life-in-law.
+The interesting question is whether capitalism is done by a Dystopie, similar to what's going to be the "Matrix," or by something that will serve the community into "Star Trek," in which people serve their energy in the universe, or their energy of exploration, or in a high-tech way of anti-tech Agora," in long-dorative atheists, are talking about the meaning of life.
+I think we can be optimistic.
+So what could it look like, this "Star Trek"-like Utopia instead of having "Matrix sex" thing?
+I want to show you a few examples of all the sound I'm going to call.
+In the area of businesses, imagine if you're making a capital market, where you're working as you work and you work in your capital from one company to the next, and the company -- no matter what you're working in -- property of the company that's working in the company.
+Then all the incomes go out of capital and cause the concept of the work is completely shaped.
+No more humble between those who are in the businesses, but they don't work there, and the employees that work there, but don't hear about the company, no longer between capital and work, no big gaps between investment and Spart, and ultimately, no empathic twins.
+In the global political economics, imagine that our national dignity would have a free-time transition, in a universal, global currency, digital currency that are given by the IMF and the G-20 [unclear], the 20-like world's interest in all humankind.
+Imagine that the whole world will act in this currency -- we call it "in Kosmos," in units of "Cosmos," and any government pay the sum of the trade or the trade of the country has gotten a shared fund.
+Imagine that this fund is invested in green technologies, especially in parts of the world, in which investments are legal.
+This is not a new idea.
+It's basically what John Maynard Keynes had put in the board of a board of a state.
+The problem was that you didn't have the technical means of doing it at the time.
+Today we have them, particularly in the background of a resemblance and economics.
+The world that I'm talking about to you is liberated at the same time in which it privileged people to prefer privileged and maristic, because it's buried by capital and work in the mothers of history, and key siansian, global key.
+But, above all, it's a world where we can imagine a real democracy.
+We're in such a world.
+Are we going to hide in a "Matrix" shamcy clogged?
+The answer depends on how we decide politically.
+It's in our hands, and we do better.
+Thank you.
+Bruno Giussani: Yanis...
+You write yourself in your biology itself as libertarianists.
+How relevant is Marx' analysis today?
+So, in Varoufakis: If anything I've just said about what's relevant, then Marx relevant.
+The reason why we do this re-inventing of politics and economics is that we don't do that, that will create a massive, massive-level development of the demand, which is called Larry Summers as the chief day.
+Through the transmission of one part of the Earth at the next, how we're seeing it, it's not just our democracy, but also the countries that are devoted to liberal democracy.
+If this analysis is a Marx if you're able to meet.
+And as Hayek, so I'm a liberating Marxist, and as Keynes, and so I'm totally us.
+BG: In fact, now we're also.
+YV: If you're not more, you don't think enough.
+BG: This is a very Greek, sort of a philosophical explanation -- YV: It was actually Einstein who said that. BG: In your talk, your talk, your talk, and China, and yesterday, I think, you know, you've been very clear at dinner, what you're saying to the point of view of the West.
+Do you want to repeat it here?
+YV: There's a big measure of yeast.
+In our liberal democracy we have the apparently of democracy.
+As I said in my talk, we have democracy in politics, as the realm of where most of the time is a game of economics -- the realm of economics -- a completely democratic zone.
+In a way, if I can say this as I'm concerned, China is feeding the 19th century China today.
+Because remember -- we tend to connect liberalism with democracy -- that's a mistake, it's a percentage.
+In liberalism, liberals, like John Stuart mill.
+He was particularly skeptical about what the democratic development was.
+Now, what you can see in China is very similar to the development that we've had in England during the industrial revolution, especially the first to the second.
+China now to throw what the West itself has done in the 19th century, smell of a violent sphere.
+BG: I'm sure many audiences are curious to your experience as finance ministers are beginning from the beginning of the year.
+YV: I saw this is coming.
+BG: Yeah...
+How did you look back six months later?
+YV: Very exciting, in personal view, and very disappointed, because we had the opportunity to start the European zone.
+Not just Greek land, but the European zone.
+Because we're getting ourselves from self-esteem to self-organization, and the limitations that a massive resemblance of the European zone, and still growing the development of the entire Union.
+We had the opportunity to find the basis of Greek proposal -- the first proposal, which was the effort to be able to do.
+Unfortunately, the mice have gotten in the European zone, in the European group continue to stop the mind.
+But you know what comes.
+This is the experience of the Soviet Union.
+If you try an economic system that's not able to survive, through political will and writers in life, maybe you will get it out of a while, but when change happens, it will destroy and destroy.
+BG: What change do you see before you?
+YV: There's no doubt that the Euro zone doesn't have a future if we don't change their building.
+BG: Did you do any error in your time when finance Minister did?
+YV: Every day.
+BG: For example? YV: Everybody looking back -- serious...
+If there's a financial minister, or any minister, who's in the amtafected after six months, especially in a situation, he says he's not making mistakes, that's a dangerous person.
+Of course, I made mistakes.
+The biggest mistake was to heat up the school program of February schools.
+I thought there was an honest interest in pages of money to come up with a common solution.
+But there was no.
+They just wanted to get our government out of the case, because they didn't want to deal with the throwaway zone through the European zone.
+They didn't want them to stand up for five years, a catastrophic program in Greece.
+We lost one third of our GDP.
+There's worse than during the guy's depression.
+No one from the comics who put us in this policy, he put in, "This was a kola failure."
+BG: Despite all that, and despite the aggressive tone of the conversation, you seem to be right per capita.
+YV: Absolutely.
+My criticism of the European Union and the European zone comes from someone who lives and loves Europe.
+My biggest fear is that the European zone is not about.
+Because if they don't survive, the defecens will be free, or the European Union destroys.
+It's not going to have Europe alone for Europe, but for the entire world economy.
+We're probably the world's poorest economy.
+If we voted ourselves to the road of a postmodern 1930, it seems to me that it will be also modern for the future of Europeans as well as the non-profit.
+BG: We hope you're in that point wrong.
+Thank you very much for coming to TED.
+YV: Thank you.
+RoyRARA, most of the people who have never heard, although probably 22 minutes of your life on April 2013 is responsible for April 2013.
+Probably very much for 22 minutes, but not for many of you.
+This is what goes back to the decision, the Roy three years ago.
+Roy Price is a standard employees at Amazon Studios.
+The TV company in Amazon.
+He's 47 years old, he's sank, he's got a spherical and says, "Filme, TV, technology, Tacos."
+Roy has a very important job because it's responsible for taking the show and making the content that's going to produce.
+Of course, that's a very tough industry.
+There are so many TV series that Roy cannot pick any one.
+He has to find show that are very, very good.
+In other words, he has to find show that are completely right on this curve.
+This curve is the Bewertung rate of over 2500 TV series on the site, and the score is about one to 10 to 10 to 10 years, showing how many show of this value is getting.
+We're putting your show up to nine and higher that winner.
+Then you have a successful show.
+These are show show like "Breaking Bad," "TheGame," which we make, all the show, where you've got a season, your brain, you ask, "What's more of these Episodes?"
+This kind of show.
+On the left, here on the end, on the end, are show like "Toddlers" and Tiaras" -- -- that should tell you enough what's going on on this curve.
+RoyRAve not about putting the left side of the curve, because I think you need special intelligence to bend "Todlers and Tiaras."
+He's making more thoughts about the middle pillars, the average television -- the show, neither good, they're not bad, they're not.
+So he has to make sure that he's really at the right side.
+The pressure is there, and of course it's the first time that Amazon makes such a thing, so Roy crash doesn't want to risk.
+He wants to create success.
+He needs the global success, so he thinks it's going to take a competition.
+He takes lots of ideas for TV show, and then he takes an eight-hour show from TV show, then he produces the first person on each of these show, and he's making them online where everybody can look for free.
+And if Amazon gives free things to be done, you're going to go right?
+Millions of audience look at these features.
+But they don't know that when they look at this show, they're watching this show.
+They're going to be watching Roy and his team who are buried everything.
+It's when you pick up the show, when you're putting paus on what pieces you're going to be looking at again.
+They collect millions of data, and then deciding what show should be producing.
+In fact, they collect the data, and it gives the answer, and it says, "Amazon to a Sitcoma over four-month U.S.ator."
+They did this show.
+Anybody name the name of this show?
+Yes, "urha House," but it seems that a lot of these shows cannot remember this show because it wasn't so good.
+It's just an average show -- in the word, that's the average of the curve, and the alpha of these curves are landed on 12 -- so something that's on average, but not what Roy and his team came up.
+Something at the same time, another manager has taken a top-down show by data analysis, Ted Sarandos, who's managers for Netflix's programmers, who's been looking at Netflix and he's using this super-how to find data on it, but he does something different.
+Instead of making a competition, they have a team and his team have been looking at the Netflix data on the Netflix scale, so the score that they give the show, the show that they like.
+Then they use this data to try and figure out these little details about the audience: what show you like, what producers, what actors.
+When they had all these pieces, they went to an orphanage, and decided not to do a Sitcoma, but to make a drama series on a Senator.
+You know the show?
+Yes, "House of the graduated." Netflix did a heat, at least for the first two season.
+"hope of Cards" is going to get a 9,1-dollar level on this curve, so exactly where they wanted to go.
+Now, of course, the question is: what happened here?
+You've got two very weird data, data-of-the-box companies.
+They connect this many data together, and it works great for one of them, but not for the other company.
+What is that?
+Because the logic somehow says that it should work on everything.
+If you collect millions of data, for a decision you're supposed to make, you should make a good decision.
+You've got 200 years of statistics than Back people.
+You optimize it through very powerful computers.
+The minimum, what you're looking for is good television, right?
+If data doesn't work, it's a little bit scary, because we live in a time when we reach more and more statistics to make serious decisions, way beyond television.
+Anybody here does the multi-Health system?
+Nobody. OK, that's actually good.
+MultiHealth system is a software company, and I hope nobody in this room ever comes with that software. Come with it, you're in jail.
+If someone in the United States is in jail, and please ask for free, it's probably that data analysis is going to use this company to determine whether a free-off or not.
+Just like Amazon and Netflix. But instead of deciding whether a show is going to be good or bad, whether a person is going to be good or bad.
+Middle TV, 22 minutes, that can be really bad, but more years's prisons.
+Unfortunately, there's evidence that this data analysis despite the many data.
+Not always the best results, and that's because a multi-Health system doesn't know how to use data.
+And again, the most hidden companies are wrong.
+Yes, even Google makes mistakes sometimes.
+In 2009, Google had known that it was based on data analysis by flu, the bad species, by able to predict data analysis of Google analysis.
+It worked wonderful and was a big news. The success in a publication in the magazine called "nature."
+It worked a liberate year around year until it didn't work.
+And nobody could say why.
+It didn't work simple, this was again again a mustard, including the Widerruf of the Sun in the magazine.
+Even the most hidden companies like Amazon and Google are enlightened sometimes a little bit.
+Despite looking at all these mistakes in decisions about the life of life -- at the workplace, at the legal side, in medicine.
+So we'd better hire that data is helpful.
+I also know a lot of trouble with data. I work in computer genetics -- an area where some very smart people use a lot of data to make serious decisions about how the decision for cancer therapy.
+Or the evolution of a medicine, over the years, I've seen some patterns in between successful decisions about data and not by successful decisions, and the pattern should be spread.
+Do you ever solve a complex problem, you're mainly doing two things.
+The first thing is, you put this problem in the parts of the parts, so you can analyze the pieces, and then you put the pieces back together to get the pieces back together to make a point.
+Sometimes you have to do this more, but it's always two things: take apart and put them together again.
+And now the most important thing: data and data analysis is just good for the first part.
+data and data analysis, no matter how powerful, can only help to get a problem and understand his parts.
+They're not able to put together the pieces back together and then get a point.
+And there's another tool for that, and we own it all: our brains.
+If there's something that is good to put together, it's pieces and pieces again, even if the information is unimaginable, to capture a good sign -- especially when the brain is an expert.
+That's why I think Netflix was so successful, because they used data and mind, where they're also using in the process.
+They use data to understand their audience better than ever, but the decision that they would have been able to do is put all these pieces together, put a show back together, and make a show like, "hope of often," that didn't stand in the data.
+Ted Sarandososo and his team made these decisions for what it meant was that they made a major personal risk to get at risk.
+Amazon did it in the wrong way.
+They used data to control all their decisions first as they're going to be on TV books, and then they'd be called theAlpha House.
+It was a safe decision because they could always say, "That's what the data is."
+It didn't lead to the very best thing.
+data is helpful for better decisions, but I believe things are wrong when data is starting to control our decisions.
+It doesn't matter how powerful it is, data is just a tool, and to forget that, it's pretty useful.
+Thank you very much.
+Before that data, it was the device for decisions.
+Many of them know it.
+It's also called theMag 8 Balls. It's amazing for decisions or no-in-Futlut, you just have to shake the ball to get an answer. "Has probably" -- right here in that moment.
+I'm going to do it later with a technique.
+I've met some decisions in my life where I should have heard in the ball.
+But, of course, you know, if you have the data available, you want to replace it through something much more exaggeration, like data analysis to make better decisions.
+But that doesn't change the cause.
+So maybe the ball will smarter and smarter and smarter, and eventually it's because we make decisions if we want to achieve something extraordinary at the end of the curve.
+And I feel this as a very encouraging message that it still gets paid a lot of data, to make decisions about what to do and to go to risks.
+Because at the end of it, it's not the data, but the risks that you end up with at the right end of the curve.
+Thank you.

generation_config.json

@@ -1,5 +1,4 @@
 {
-  "_from_model_config": true,
   "decoder_start_token_id": 0,
   "eos_token_id": 1,
   "pad_token_id": 0,

predict_results.json

@@ -0,0 +1,9 @@
+{
+    "predict_bleu": 0.1921,
+    "predict_gen_len": 20.8805,
+    "predict_loss": 3.0906615257263184,
+    "predict_runtime": 5.6211,
+    "predict_samples": 1138,
+    "predict_samples_per_second": 202.453,
+    "predict_steps_per_second": 1.601
+}

runs/May11_13-29-22_0d573eeffc83/events.out.tfevents.1715496931.0d573eeffc83.44348.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6997574a829281df7821b1e376087ef909c6ef250d46cb36731188a077366d88
+size 465

train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 100.0,
+    "total_flos": 3.157662139522744e+17,
+    "train_loss": 2.261507166926833,
+    "train_runtime": 62730.7046,
+    "train_samples": 116654,
+    "train_samples_per_second": 185.96,
+    "train_steps_per_second": 2.906
+}

trainer_state.json

@@ -0,0 +1,2590 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 100.0,
+  "eval_steps": 500,
+  "global_step": 182300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.27427317608337903,
+      "grad_norm": 2.1429903507232666,
+      "learning_rate": 4.986286341195831e-05,
+      "loss": 7.2532,
+      "step": 500
+    },
+    {
+      "epoch": 0.5485463521667581,
+      "grad_norm": 2.0887398719787598,
+      "learning_rate": 4.972572682391663e-05,
+      "loss": 6.4724,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8228195282501372,
+      "grad_norm": 2.1614580154418945,
+      "learning_rate": 4.958859023587493e-05,
+      "loss": 6.0973,
+      "step": 1500
+    },
+    {
+      "epoch": 1.0970927043335161,
+      "grad_norm": 2.202652931213379,
+      "learning_rate": 4.9451453647833245e-05,
+      "loss": 5.7872,
+      "step": 2000
+    },
+    {
+      "epoch": 1.3713658804168953,
+      "grad_norm": 2.147414445877075,
+      "learning_rate": 4.931431705979155e-05,
+      "loss": 5.5399,
+      "step": 2500
+    },
+    {
+      "epoch": 1.6456390565002743,
+      "grad_norm": 2.1624412536621094,
+      "learning_rate": 4.917718047174987e-05,
+      "loss": 5.3515,
+      "step": 3000
+    },
+    {
+      "epoch": 1.9199122325836533,
+      "grad_norm": 2.155363082885742,
+      "learning_rate": 4.9040043883708175e-05,
+      "loss": 5.1762,
+      "step": 3500
+    },
+    {
+      "epoch": 2.1941854086670323,
+      "grad_norm": 2.2420756816864014,
+      "learning_rate": 4.890290729566648e-05,
+      "loss": 5.0364,
+      "step": 4000
+    },
+    {
+      "epoch": 2.4684585847504112,
+      "grad_norm": 2.4155259132385254,
+      "learning_rate": 4.87657707076248e-05,
+      "loss": 4.9178,
+      "step": 4500
+    },
+    {
+      "epoch": 2.7427317608337907,
+      "grad_norm": 2.2149574756622314,
+      "learning_rate": 4.8628634119583105e-05,
+      "loss": 4.8449,
+      "step": 5000
+    },
+    {
+      "epoch": 3.0170049369171696,
+      "grad_norm": 2.25925350189209,
+      "learning_rate": 4.849149753154142e-05,
+      "loss": 4.762,
+      "step": 5500
+    },
+    {
+      "epoch": 3.2912781130005486,
+      "grad_norm": 2.315990686416626,
+      "learning_rate": 4.835436094349973e-05,
+      "loss": 4.6713,
+      "step": 6000
+    },
+    {
+      "epoch": 3.5655512890839276,
+      "grad_norm": 2.425288677215576,
+      "learning_rate": 4.821722435545804e-05,
+      "loss": 4.6267,
+      "step": 6500
+    },
+    {
+      "epoch": 3.8398244651673066,
+      "grad_norm": 2.3451356887817383,
+      "learning_rate": 4.808008776741635e-05,
+      "loss": 4.579,
+      "step": 7000
+    },
+    {
+      "epoch": 4.1140976412506856,
+      "grad_norm": 2.306058645248413,
+      "learning_rate": 4.794295117937466e-05,
+      "loss": 4.5148,
+      "step": 7500
+    },
+    {
+      "epoch": 4.3883708173340645,
+      "grad_norm": 2.3386404514312744,
+      "learning_rate": 4.780581459133297e-05,
+      "loss": 4.4659,
+      "step": 8000
+    },
+    {
+      "epoch": 4.6626439934174435,
+      "grad_norm": 2.3117551803588867,
+      "learning_rate": 4.766867800329128e-05,
+      "loss": 4.4327,
+      "step": 8500
+    },
+    {
+      "epoch": 4.9369171695008225,
+      "grad_norm": 2.36466908454895,
+      "learning_rate": 4.753154141524959e-05,
+      "loss": 4.3947,
+      "step": 9000
+    },
+    {
+      "epoch": 5.2111903455842015,
+      "grad_norm": 2.348733901977539,
+      "learning_rate": 4.73944048272079e-05,
+      "loss": 4.3441,
+      "step": 9500
+    },
+    {
+      "epoch": 5.485463521667581,
+      "grad_norm": 2.9133706092834473,
+      "learning_rate": 4.7257268239166215e-05,
+      "loss": 4.3025,
+      "step": 10000
+    },
+    {
+      "epoch": 5.75973669775096,
+      "grad_norm": 2.6369545459747314,
+      "learning_rate": 4.712013165112452e-05,
+      "loss": 4.2785,
+      "step": 10500
+    },
+    {
+      "epoch": 6.034009873834339,
+      "grad_norm": 2.7040719985961914,
+      "learning_rate": 4.698299506308283e-05,
+      "loss": 4.2431,
+      "step": 11000
+    },
+    {
+      "epoch": 6.308283049917718,
+      "grad_norm": 2.7137389183044434,
+      "learning_rate": 4.6845858475041146e-05,
+      "loss": 4.1796,
+      "step": 11500
+    },
+    {
+      "epoch": 6.582556226001097,
+      "grad_norm": 2.963534355163574,
+      "learning_rate": 4.670872188699945e-05,
+      "loss": 4.1432,
+      "step": 12000
+    },
+    {
+      "epoch": 6.856829402084476,
+      "grad_norm": 2.8243420124053955,
+      "learning_rate": 4.6571585298957763e-05,
+      "loss": 4.1252,
+      "step": 12500
+    },
+    {
+      "epoch": 7.131102578167855,
+      "grad_norm": 3.03757905960083,
+      "learning_rate": 4.6434448710916076e-05,
+      "loss": 4.0713,
+      "step": 13000
+    },
+    {
+      "epoch": 7.405375754251234,
+      "grad_norm": 3.007908821105957,
+      "learning_rate": 4.629731212287439e-05,
+      "loss": 4.0248,
+      "step": 13500
+    },
+    {
+      "epoch": 7.679648930334613,
+      "grad_norm": 2.846788167953491,
+      "learning_rate": 4.6160175534832694e-05,
+      "loss": 4.0099,
+      "step": 14000
+    },
+    {
+      "epoch": 7.953922106417992,
+      "grad_norm": 2.961183786392212,
+      "learning_rate": 4.6023038946791006e-05,
+      "loss": 3.9728,
+      "step": 14500
+    },
+    {
+      "epoch": 8.228195282501371,
+      "grad_norm": 3.066025733947754,
+      "learning_rate": 4.588590235874932e-05,
+      "loss": 3.9118,
+      "step": 15000
+    },
+    {
+      "epoch": 8.502468458584751,
+      "grad_norm": 2.9394822120666504,
+      "learning_rate": 4.5748765770707624e-05,
+      "loss": 3.8828,
+      "step": 15500
+    },
+    {
+      "epoch": 8.776741634668129,
+      "grad_norm": 3.012153387069702,
+      "learning_rate": 4.5611629182665936e-05,
+      "loss": 3.8832,
+      "step": 16000
+    },
+    {
+      "epoch": 9.051014810751509,
+      "grad_norm": 2.899332046508789,
+      "learning_rate": 4.547449259462425e-05,
+      "loss": 3.8481,
+      "step": 16500
+    },
+    {
+      "epoch": 9.325287986834887,
+      "grad_norm": 3.164444923400879,
+      "learning_rate": 4.533735600658256e-05,
+      "loss": 3.7773,
+      "step": 17000
+    },
+    {
+      "epoch": 9.599561162918267,
+      "grad_norm": 3.017282009124756,
+      "learning_rate": 4.5200219418540867e-05,
+      "loss": 3.7689,
+      "step": 17500
+    },
+    {
+      "epoch": 9.873834339001645,
+      "grad_norm": 3.367647647857666,
+      "learning_rate": 4.506308283049918e-05,
+      "loss": 3.752,
+      "step": 18000
+    },
+    {
+      "epoch": 10.148107515085025,
+      "grad_norm": 2.9855947494506836,
+      "learning_rate": 4.492594624245749e-05,
+      "loss": 3.7182,
+      "step": 18500
+    },
+    {
+      "epoch": 10.422380691168403,
+      "grad_norm": 3.505870819091797,
+      "learning_rate": 4.47888096544158e-05,
+      "loss": 3.674,
+      "step": 19000
+    },
+    {
+      "epoch": 10.696653867251783,
+      "grad_norm": 3.438145160675049,
+      "learning_rate": 4.465167306637411e-05,
+      "loss": 3.651,
+      "step": 19500
+    },
+    {
+      "epoch": 10.970927043335163,
+      "grad_norm": 3.0687413215637207,
+      "learning_rate": 4.451453647833242e-05,
+      "loss": 3.6479,
+      "step": 20000
+    },
+    {
+      "epoch": 11.24520021941854,
+      "grad_norm": 3.2287588119506836,
+      "learning_rate": 4.4377399890290734e-05,
+      "loss": 3.5693,
+      "step": 20500
+    },
+    {
+      "epoch": 11.51947339550192,
+      "grad_norm": 3.3848471641540527,
+      "learning_rate": 4.424026330224904e-05,
+      "loss": 3.5667,
+      "step": 21000
+    },
+    {
+      "epoch": 11.793746571585299,
+      "grad_norm": 3.5464422702789307,
+      "learning_rate": 4.410312671420735e-05,
+      "loss": 3.5589,
+      "step": 21500
+    },
+    {
+      "epoch": 12.068019747668679,
+      "grad_norm": 3.6160085201263428,
+      "learning_rate": 4.3965990126165664e-05,
+      "loss": 3.5313,
+      "step": 22000
+    },
+    {
+      "epoch": 12.342292923752057,
+      "grad_norm": 3.6420817375183105,
+      "learning_rate": 4.3828853538123976e-05,
+      "loss": 3.4855,
+      "step": 22500
+    },
+    {
+      "epoch": 12.616566099835437,
+      "grad_norm": 3.583449363708496,
+      "learning_rate": 4.369171695008228e-05,
+      "loss": 3.4654,
+      "step": 23000
+    },
+    {
+      "epoch": 12.890839275918815,
+      "grad_norm": 3.5506091117858887,
+      "learning_rate": 4.3554580362040594e-05,
+      "loss": 3.4512,
+      "step": 23500
+    },
+    {
+      "epoch": 13.165112452002194,
+      "grad_norm": 4.101990699768066,
+      "learning_rate": 4.341744377399891e-05,
+      "loss": 3.4055,
+      "step": 24000
+    },
+    {
+      "epoch": 13.439385628085573,
+      "grad_norm": 4.144250392913818,
+      "learning_rate": 4.328030718595721e-05,
+      "loss": 3.3755,
+      "step": 24500
+    },
+    {
+      "epoch": 13.713658804168952,
+      "grad_norm": 3.6288070678710938,
+      "learning_rate": 4.3143170597915525e-05,
+      "loss": 3.3725,
+      "step": 25000
+    },
+    {
+      "epoch": 13.98793198025233,
+      "grad_norm": 3.5927882194519043,
+      "learning_rate": 4.300603400987384e-05,
+      "loss": 3.3693,
+      "step": 25500
+    },
+    {
+      "epoch": 14.26220515633571,
+      "grad_norm": 3.405404567718506,
+      "learning_rate": 4.286889742183215e-05,
+      "loss": 3.2955,
+      "step": 26000
+    },
+    {
+      "epoch": 14.53647833241909,
+      "grad_norm": 4.086198329925537,
+      "learning_rate": 4.2731760833790455e-05,
+      "loss": 3.3038,
+      "step": 26500
+    },
+    {
+      "epoch": 14.810751508502468,
+      "grad_norm": 3.3961052894592285,
+      "learning_rate": 4.259462424574877e-05,
+      "loss": 3.2561,
+      "step": 27000
+    },
+    {
+      "epoch": 15.085024684585848,
+      "grad_norm": 3.6080105304718018,
+      "learning_rate": 4.245748765770708e-05,
+      "loss": 3.2539,
+      "step": 27500
+    },
+    {
+      "epoch": 15.359297860669226,
+      "grad_norm": 3.597956657409668,
+      "learning_rate": 4.2320351069665385e-05,
+      "loss": 3.2148,
+      "step": 28000
+    },
+    {
+      "epoch": 15.633571036752606,
+      "grad_norm": 3.466057062149048,
+      "learning_rate": 4.21832144816237e-05,
+      "loss": 3.2031,
+      "step": 28500
+    },
+    {
+      "epoch": 15.907844212835984,
+      "grad_norm": 4.239918231964111,
+      "learning_rate": 4.204607789358201e-05,
+      "loss": 3.1768,
+      "step": 29000
+    },
+    {
+      "epoch": 16.182117388919362,
+      "grad_norm": 4.533541202545166,
+      "learning_rate": 4.190894130554032e-05,
+      "loss": 3.1258,
+      "step": 29500
+    },
+    {
+      "epoch": 16.456390565002742,
+      "grad_norm": 3.8643674850463867,
+      "learning_rate": 4.177180471749863e-05,
+      "loss": 3.1162,
+      "step": 30000
+    },
+    {
+      "epoch": 16.730663741086122,
+      "grad_norm": 3.710988998413086,
+      "learning_rate": 4.163466812945694e-05,
+      "loss": 3.1159,
+      "step": 30500
+    },
+    {
+      "epoch": 17.004936917169502,
+      "grad_norm": 3.6405742168426514,
+      "learning_rate": 4.149753154141525e-05,
+      "loss": 3.1137,
+      "step": 31000
+    },
+    {
+      "epoch": 17.27921009325288,
+      "grad_norm": 4.127532482147217,
+      "learning_rate": 4.136039495337356e-05,
+      "loss": 3.0327,
+      "step": 31500
+    },
+    {
+      "epoch": 17.553483269336258,
+      "grad_norm": 3.6551403999328613,
+      "learning_rate": 4.122325836533188e-05,
+      "loss": 3.0611,
+      "step": 32000
+    },
+    {
+      "epoch": 17.827756445419638,
+      "grad_norm": 3.6437666416168213,
+      "learning_rate": 4.108612177729018e-05,
+      "loss": 3.0233,
+      "step": 32500
+    },
+    {
+      "epoch": 18.102029621503018,
+      "grad_norm": 4.008886814117432,
+      "learning_rate": 4.0948985189248495e-05,
+      "loss": 3.0099,
+      "step": 33000
+    },
+    {
+      "epoch": 18.376302797586398,
+      "grad_norm": 3.779545783996582,
+      "learning_rate": 4.08118486012068e-05,
+      "loss": 2.9663,
+      "step": 33500
+    },
+    {
+      "epoch": 18.650575973669774,
+      "grad_norm": 3.7845826148986816,
+      "learning_rate": 4.067471201316512e-05,
+      "loss": 2.9628,
+      "step": 34000
+    },
+    {
+      "epoch": 18.924849149753154,
+      "grad_norm": 3.866852283477783,
+      "learning_rate": 4.0537575425123425e-05,
+      "loss": 2.9649,
+      "step": 34500
+    },
+    {
+      "epoch": 19.199122325836534,
+      "grad_norm": 3.9092442989349365,
+      "learning_rate": 4.040043883708173e-05,
+      "loss": 2.9199,
+      "step": 35000
+    },
+    {
+      "epoch": 19.473395501919914,
+      "grad_norm": 4.038732528686523,
+      "learning_rate": 4.026330224904005e-05,
+      "loss": 2.9078,
+      "step": 35500
+    },
+    {
+      "epoch": 19.74766867800329,
+      "grad_norm": 3.717470645904541,
+      "learning_rate": 4.0126165660998355e-05,
+      "loss": 2.8907,
+      "step": 36000
+    },
+    {
+      "epoch": 20.02194185408667,
+      "grad_norm": 3.9166011810302734,
+      "learning_rate": 3.998902907295667e-05,
+      "loss": 2.8892,
+      "step": 36500
+    },
+    {
+      "epoch": 20.29621503017005,
+      "grad_norm": 3.9473681449890137,
+      "learning_rate": 3.985189248491497e-05,
+      "loss": 2.8274,
+      "step": 37000
+    },
+    {
+      "epoch": 20.57048820625343,
+      "grad_norm": 4.139565467834473,
+      "learning_rate": 3.971475589687329e-05,
+      "loss": 2.8371,
+      "step": 37500
+    },
+    {
+      "epoch": 20.844761382336806,
+      "grad_norm": 3.7124762535095215,
+      "learning_rate": 3.95776193088316e-05,
+      "loss": 2.8401,
+      "step": 38000
+    },
+    {
+      "epoch": 21.119034558420186,
+      "grad_norm": 3.772149085998535,
+      "learning_rate": 3.9440482720789904e-05,
+      "loss": 2.8083,
+      "step": 38500
+    },
+    {
+      "epoch": 21.393307734503566,
+      "grad_norm": 4.185425758361816,
+      "learning_rate": 3.930334613274822e-05,
+      "loss": 2.7767,
+      "step": 39000
+    },
+    {
+      "epoch": 21.667580910586945,
+      "grad_norm": 3.800649881362915,
+      "learning_rate": 3.916620954470653e-05,
+      "loss": 2.7723,
+      "step": 39500
+    },
+    {
+      "epoch": 21.941854086670325,
+      "grad_norm": 3.800741195678711,
+      "learning_rate": 3.902907295666484e-05,
+      "loss": 2.774,
+      "step": 40000
+    },
+    {
+      "epoch": 22.2161272627537,
+      "grad_norm": 3.7834713459014893,
+      "learning_rate": 3.889193636862315e-05,
+      "loss": 2.7329,
+      "step": 40500
+    },
+    {
+      "epoch": 22.49040043883708,
+      "grad_norm": 4.18643045425415,
+      "learning_rate": 3.8754799780581465e-05,
+      "loss": 2.7181,
+      "step": 41000
+    },
+    {
+      "epoch": 22.76467361492046,
+      "grad_norm": 3.758415460586548,
+      "learning_rate": 3.861766319253977e-05,
+      "loss": 2.7148,
+      "step": 41500
+    },
+    {
+      "epoch": 23.03894679100384,
+      "grad_norm": 4.028139114379883,
+      "learning_rate": 3.8480526604498076e-05,
+      "loss": 2.7025,
+      "step": 42000
+    },
+    {
+      "epoch": 23.313219967087218,
+      "grad_norm": 3.779428243637085,
+      "learning_rate": 3.8343390016456395e-05,
+      "loss": 2.6542,
+      "step": 42500
+    },
+    {
+      "epoch": 23.587493143170597,
+      "grad_norm": 3.6896047592163086,
+      "learning_rate": 3.82062534284147e-05,
+      "loss": 2.6662,
+      "step": 43000
+    },
+    {
+      "epoch": 23.861766319253977,
+      "grad_norm": 3.527114152908325,
+      "learning_rate": 3.806911684037301e-05,
+      "loss": 2.6697,
+      "step": 43500
+    },
+    {
+      "epoch": 24.136039495337357,
+      "grad_norm": 3.832408905029297,
+      "learning_rate": 3.7931980252331326e-05,
+      "loss": 2.6443,
+      "step": 44000
+    },
+    {
+      "epoch": 24.410312671420733,
+      "grad_norm": 3.912022352218628,
+      "learning_rate": 3.779484366428964e-05,
+      "loss": 2.6098,
+      "step": 44500
+    },
+    {
+      "epoch": 24.684585847504113,
+      "grad_norm": 3.832465648651123,
+      "learning_rate": 3.7657707076247944e-05,
+      "loss": 2.6187,
+      "step": 45000
+    },
+    {
+      "epoch": 24.958859023587493,
+      "grad_norm": 4.292030334472656,
+      "learning_rate": 3.752057048820625e-05,
+      "loss": 2.6216,
+      "step": 45500
+    },
+    {
+      "epoch": 25.233132199670873,
+      "grad_norm": 3.9441139698028564,
+      "learning_rate": 3.738343390016457e-05,
+      "loss": 2.5674,
+      "step": 46000
+    },
+    {
+      "epoch": 25.507405375754253,
+      "grad_norm": 3.8073363304138184,
+      "learning_rate": 3.7246297312122874e-05,
+      "loss": 2.5613,
+      "step": 46500
+    },
+    {
+      "epoch": 25.78167855183763,
+      "grad_norm": 3.95381498336792,
+      "learning_rate": 3.7109160724081186e-05,
+      "loss": 2.5744,
+      "step": 47000
+    },
+    {
+      "epoch": 26.05595172792101,
+      "grad_norm": 4.265843391418457,
+      "learning_rate": 3.69720241360395e-05,
+      "loss": 2.5581,
+      "step": 47500
+    },
+    {
+      "epoch": 26.33022490400439,
+      "grad_norm": 3.7907886505126953,
+      "learning_rate": 3.683488754799781e-05,
+      "loss": 2.5085,
+      "step": 48000
+    },
+    {
+      "epoch": 26.60449808008777,
+      "grad_norm": 3.9580938816070557,
+      "learning_rate": 3.6697750959956116e-05,
+      "loss": 2.5244,
+      "step": 48500
+    },
+    {
+      "epoch": 26.878771256171145,
+      "grad_norm": 3.725271701812744,
+      "learning_rate": 3.656061437191443e-05,
+      "loss": 2.5283,
+      "step": 49000
+    },
+    {
+      "epoch": 27.153044432254525,
+      "grad_norm": 4.27789831161499,
+      "learning_rate": 3.642347778387274e-05,
+      "loss": 2.4978,
+      "step": 49500
+    },
+    {
+      "epoch": 27.427317608337905,
+      "grad_norm": 4.248454570770264,
+      "learning_rate": 3.628634119583105e-05,
+      "loss": 2.478,
+      "step": 50000
+    },
+    {
+      "epoch": 27.701590784421285,
+      "grad_norm": 3.7782256603240967,
+      "learning_rate": 3.614920460778936e-05,
+      "loss": 2.4797,
+      "step": 50500
+    },
+    {
+      "epoch": 27.97586396050466,
+      "grad_norm": 3.996277332305908,
+      "learning_rate": 3.601206801974767e-05,
+      "loss": 2.4872,
+      "step": 51000
+    },
+    {
+      "epoch": 28.25013713658804,
+      "grad_norm": 4.143040657043457,
+      "learning_rate": 3.5874931431705984e-05,
+      "loss": 2.4311,
+      "step": 51500
+    },
+    {
+      "epoch": 28.52441031267142,
+      "grad_norm": 3.6849589347839355,
+      "learning_rate": 3.573779484366429e-05,
+      "loss": 2.4377,
+      "step": 52000
+    },
+    {
+      "epoch": 28.7986834887548,
+      "grad_norm": 3.621760606765747,
+      "learning_rate": 3.56006582556226e-05,
+      "loss": 2.4425,
+      "step": 52500
+    },
+    {
+      "epoch": 29.07295666483818,
+      "grad_norm": 3.7394306659698486,
+      "learning_rate": 3.5463521667580914e-05,
+      "loss": 2.4287,
+      "step": 53000
+    },
+    {
+      "epoch": 29.347229840921557,
+      "grad_norm": 3.782111167907715,
+      "learning_rate": 3.532638507953922e-05,
+      "loss": 2.3911,
+      "step": 53500
+    },
+    {
+      "epoch": 29.621503017004937,
+      "grad_norm": 4.35050106048584,
+      "learning_rate": 3.518924849149753e-05,
+      "loss": 2.4084,
+      "step": 54000
+    },
+    {
+      "epoch": 29.895776193088317,
+      "grad_norm": 3.8727004528045654,
+      "learning_rate": 3.5052111903455844e-05,
+      "loss": 2.3928,
+      "step": 54500
+    },
+    {
+      "epoch": 30.170049369171696,
+      "grad_norm": 3.974501371383667,
+      "learning_rate": 3.4914975315414157e-05,
+      "loss": 2.3668,
+      "step": 55000
+    },
+    {
+      "epoch": 30.444322545255073,
+      "grad_norm": 3.7882275581359863,
+      "learning_rate": 3.477783872737246e-05,
+      "loss": 2.3686,
+      "step": 55500
+    },
+    {
+      "epoch": 30.718595721338453,
+      "grad_norm": 3.8313581943511963,
+      "learning_rate": 3.4640702139330774e-05,
+      "loss": 2.3707,
+      "step": 56000
+    },
+    {
+      "epoch": 30.992868897421832,
+      "grad_norm": 4.046344757080078,
+      "learning_rate": 3.450356555128909e-05,
+      "loss": 2.3632,
+      "step": 56500
+    },
+    {
+      "epoch": 31.267142073505212,
+      "grad_norm": 3.7165708541870117,
+      "learning_rate": 3.436642896324739e-05,
+      "loss": 2.3229,
+      "step": 57000
+    },
+    {
+      "epoch": 31.54141524958859,
+      "grad_norm": 3.8072948455810547,
+      "learning_rate": 3.4229292375205705e-05,
+      "loss": 2.3262,
+      "step": 57500
+    },
+    {
+      "epoch": 31.81568842567197,
+      "grad_norm": 3.8423380851745605,
+      "learning_rate": 3.409215578716402e-05,
+      "loss": 2.3311,
+      "step": 58000
+    },
+    {
+      "epoch": 32.08996160175535,
+      "grad_norm": 3.831343650817871,
+      "learning_rate": 3.395501919912233e-05,
+      "loss": 2.3213,
+      "step": 58500
+    },
+    {
+      "epoch": 32.364234777838725,
+      "grad_norm": 3.8060476779937744,
+      "learning_rate": 3.3817882611080635e-05,
+      "loss": 2.2849,
+      "step": 59000
+    },
+    {
+      "epoch": 32.63850795392211,
+      "grad_norm": 4.033987998962402,
+      "learning_rate": 3.368074602303895e-05,
+      "loss": 2.2979,
+      "step": 59500
+    },
+    {
+      "epoch": 32.912781130005484,
+      "grad_norm": 3.870171546936035,
+      "learning_rate": 3.354360943499726e-05,
+      "loss": 2.2982,
+      "step": 60000
+    },
+    {
+      "epoch": 33.18705430608887,
+      "grad_norm": 3.848620653152466,
+      "learning_rate": 3.340647284695557e-05,
+      "loss": 2.264,
+      "step": 60500
+    },
+    {
+      "epoch": 33.461327482172244,
+      "grad_norm": 4.048386573791504,
+      "learning_rate": 3.326933625891388e-05,
+      "loss": 2.2568,
+      "step": 61000
+    },
+    {
+      "epoch": 33.73560065825562,
+      "grad_norm": 4.029069900512695,
+      "learning_rate": 3.313219967087219e-05,
+      "loss": 2.2602,
+      "step": 61500
+    },
+    {
+      "epoch": 34.009873834339004,
+      "grad_norm": 3.759799003601074,
+      "learning_rate": 3.29950630828305e-05,
+      "loss": 2.2763,
+      "step": 62000
+    },
+    {
+      "epoch": 34.28414701042238,
+      "grad_norm": 4.440002918243408,
+      "learning_rate": 3.285792649478881e-05,
+      "loss": 2.2137,
+      "step": 62500
+    },
+    {
+      "epoch": 34.55842018650576,
+      "grad_norm": 3.961390972137451,
+      "learning_rate": 3.272078990674713e-05,
+      "loss": 2.233,
+      "step": 63000
+    },
+    {
+      "epoch": 34.83269336258914,
+      "grad_norm": 3.916156768798828,
+      "learning_rate": 3.258365331870543e-05,
+      "loss": 2.2414,
+      "step": 63500
+    },
+    {
+      "epoch": 35.106966538672516,
+      "grad_norm": 3.746572732925415,
+      "learning_rate": 3.2446516730663745e-05,
+      "loss": 2.2137,
+      "step": 64000
+    },
+    {
+      "epoch": 35.3812397147559,
+      "grad_norm": 3.8424971103668213,
+      "learning_rate": 3.230938014262205e-05,
+      "loss": 2.1907,
+      "step": 64500
+    },
+    {
+      "epoch": 35.655512890839276,
+      "grad_norm": 4.250007152557373,
+      "learning_rate": 3.217224355458036e-05,
+      "loss": 2.2104,
+      "step": 65000
+    },
+    {
+      "epoch": 35.92978606692265,
+      "grad_norm": 3.760779857635498,
+      "learning_rate": 3.2035106966538675e-05,
+      "loss": 2.2134,
+      "step": 65500
+    },
+    {
+      "epoch": 36.204059243006036,
+      "grad_norm": 4.189092636108398,
+      "learning_rate": 3.189797037849698e-05,
+      "loss": 2.1714,
+      "step": 66000
+    },
+    {
+      "epoch": 36.47833241908941,
+      "grad_norm": 3.6505000591278076,
+      "learning_rate": 3.17608337904553e-05,
+      "loss": 2.1635,
+      "step": 66500
+    },
+    {
+      "epoch": 36.752605595172795,
+      "grad_norm": 3.7696096897125244,
+      "learning_rate": 3.1623697202413605e-05,
+      "loss": 2.1832,
+      "step": 67000
+    },
+    {
+      "epoch": 37.02687877125617,
+      "grad_norm": 3.744976282119751,
+      "learning_rate": 3.148656061437192e-05,
+      "loss": 2.1753,
+      "step": 67500
+    },
+    {
+      "epoch": 37.30115194733955,
+      "grad_norm": 3.9759116172790527,
+      "learning_rate": 3.134942402633022e-05,
+      "loss": 2.1326,
+      "step": 68000
+    },
+    {
+      "epoch": 37.57542512342293,
+      "grad_norm": 3.719237804412842,
+      "learning_rate": 3.1212287438288536e-05,
+      "loss": 2.1423,
+      "step": 68500
+    },
+    {
+      "epoch": 37.84969829950631,
+      "grad_norm": 4.290117263793945,
+      "learning_rate": 3.107515085024685e-05,
+      "loss": 2.1602,
+      "step": 69000
+    },
+    {
+      "epoch": 38.123971475589684,
+      "grad_norm": 4.060342311859131,
+      "learning_rate": 3.0938014262205153e-05,
+      "loss": 2.1349,
+      "step": 69500
+    },
+    {
+      "epoch": 38.39824465167307,
+      "grad_norm": 4.048706531524658,
+      "learning_rate": 3.080087767416347e-05,
+      "loss": 2.1018,
+      "step": 70000
+    },
+    {
+      "epoch": 38.672517827756444,
+      "grad_norm": 4.130014896392822,
+      "learning_rate": 3.066374108612178e-05,
+      "loss": 2.1284,
+      "step": 70500
+    },
+    {
+      "epoch": 38.94679100383983,
+      "grad_norm": 3.893848419189453,
+      "learning_rate": 3.052660449808009e-05,
+      "loss": 2.1255,
+      "step": 71000
+    },
+    {
+      "epoch": 39.221064179923204,
+      "grad_norm": 3.6254563331604004,
+      "learning_rate": 3.03894679100384e-05,
+      "loss": 2.0848,
+      "step": 71500
+    },
+    {
+      "epoch": 39.49533735600658,
+      "grad_norm": 4.06374979019165,
+      "learning_rate": 3.025233132199671e-05,
+      "loss": 2.0933,
+      "step": 72000
+    },
+    {
+      "epoch": 39.76961053208996,
+      "grad_norm": 3.763274908065796,
+      "learning_rate": 3.011519473395502e-05,
+      "loss": 2.11,
+      "step": 72500
+    },
+    {
+      "epoch": 40.04388370817334,
+      "grad_norm": 4.014530181884766,
+      "learning_rate": 2.997805814591333e-05,
+      "loss": 2.0918,
+      "step": 73000
+    },
+    {
+      "epoch": 40.31815688425672,
+      "grad_norm": 3.6787962913513184,
+      "learning_rate": 2.9840921557871642e-05,
+      "loss": 2.0559,
+      "step": 73500
+    },
+    {
+      "epoch": 40.5924300603401,
+      "grad_norm": 3.752711057662964,
+      "learning_rate": 2.970378496982995e-05,
+      "loss": 2.067,
+      "step": 74000
+    },
+    {
+      "epoch": 40.866703236423476,
+      "grad_norm": 3.795217752456665,
+      "learning_rate": 2.9566648381788263e-05,
+      "loss": 2.0792,
+      "step": 74500
+    },
+    {
+      "epoch": 41.14097641250686,
+      "grad_norm": 3.7484569549560547,
+      "learning_rate": 2.9429511793746572e-05,
+      "loss": 2.0608,
+      "step": 75000
+    },
+    {
+      "epoch": 41.415249588590235,
+      "grad_norm": 3.601229190826416,
+      "learning_rate": 2.9292375205704885e-05,
+      "loss": 2.0337,
+      "step": 75500
+    },
+    {
+      "epoch": 41.68952276467361,
+      "grad_norm": 3.9707863330841064,
+      "learning_rate": 2.9155238617663194e-05,
+      "loss": 2.0426,
+      "step": 76000
+    },
+    {
+      "epoch": 41.963795940756995,
+      "grad_norm": 3.9523677825927734,
+      "learning_rate": 2.9018102029621502e-05,
+      "loss": 2.0571,
+      "step": 76500
+    },
+    {
+      "epoch": 42.23806911684037,
+      "grad_norm": 4.046602725982666,
+      "learning_rate": 2.8880965441579815e-05,
+      "loss": 2.0225,
+      "step": 77000
+    },
+    {
+      "epoch": 42.512342292923755,
+      "grad_norm": 4.059443950653076,
+      "learning_rate": 2.8743828853538124e-05,
+      "loss": 2.0185,
+      "step": 77500
+    },
+    {
+      "epoch": 42.78661546900713,
+      "grad_norm": 4.066934108734131,
+      "learning_rate": 2.8606692265496436e-05,
+      "loss": 2.0342,
+      "step": 78000
+    },
+    {
+      "epoch": 43.06088864509051,
+      "grad_norm": 3.811591386795044,
+      "learning_rate": 2.8469555677454745e-05,
+      "loss": 2.0216,
+      "step": 78500
+    },
+    {
+      "epoch": 43.33516182117389,
+      "grad_norm": 3.979374408721924,
+      "learning_rate": 2.833241908941306e-05,
+      "loss": 1.989,
+      "step": 79000
+    },
+    {
+      "epoch": 43.60943499725727,
+      "grad_norm": 3.67275333404541,
+      "learning_rate": 2.8195282501371366e-05,
+      "loss": 1.9958,
+      "step": 79500
+    },
+    {
+      "epoch": 43.88370817334065,
+      "grad_norm": 3.790217399597168,
+      "learning_rate": 2.8058145913329675e-05,
+      "loss": 2.0102,
+      "step": 80000
+    },
+    {
+      "epoch": 44.15798134942403,
+      "grad_norm": 3.9934496879577637,
+      "learning_rate": 2.7921009325287988e-05,
+      "loss": 1.9847,
+      "step": 80500
+    },
+    {
+      "epoch": 44.4322545255074,
+      "grad_norm": 4.339521408081055,
+      "learning_rate": 2.7783872737246297e-05,
+      "loss": 1.9752,
+      "step": 81000
+    },
+    {
+      "epoch": 44.70652770159079,
+      "grad_norm": 3.5851519107818604,
+      "learning_rate": 2.764673614920461e-05,
+      "loss": 1.9883,
+      "step": 81500
+    },
+    {
+      "epoch": 44.98080087767416,
+      "grad_norm": 4.129305362701416,
+      "learning_rate": 2.7509599561162918e-05,
+      "loss": 1.9896,
+      "step": 82000
+    },
+    {
+      "epoch": 45.25507405375754,
+      "grad_norm": 3.752852201461792,
+      "learning_rate": 2.7372462973121234e-05,
+      "loss": 1.9465,
+      "step": 82500
+    },
+    {
+      "epoch": 45.52934722984092,
+      "grad_norm": 3.923309087753296,
+      "learning_rate": 2.723532638507954e-05,
+      "loss": 1.9589,
+      "step": 83000
+    },
+    {
+      "epoch": 45.8036204059243,
+      "grad_norm": 4.141747951507568,
+      "learning_rate": 2.7098189797037848e-05,
+      "loss": 1.9662,
+      "step": 83500
+    },
+    {
+      "epoch": 46.07789358200768,
+      "grad_norm": 4.118216514587402,
+      "learning_rate": 2.696105320899616e-05,
+      "loss": 1.9518,
+      "step": 84000
+    },
+    {
+      "epoch": 46.35216675809106,
+      "grad_norm": 4.061371326446533,
+      "learning_rate": 2.682391662095447e-05,
+      "loss": 1.928,
+      "step": 84500
+    },
+    {
+      "epoch": 46.626439934174435,
+      "grad_norm": 4.138849258422852,
+      "learning_rate": 2.6686780032912785e-05,
+      "loss": 1.9456,
+      "step": 85000
+    },
+    {
+      "epoch": 46.90071311025782,
+      "grad_norm": 3.9675650596618652,
+      "learning_rate": 2.654964344487109e-05,
+      "loss": 1.9465,
+      "step": 85500
+    },
+    {
+      "epoch": 47.174986286341195,
+      "grad_norm": 3.745779514312744,
+      "learning_rate": 2.6412506856829406e-05,
+      "loss": 1.9293,
+      "step": 86000
+    },
+    {
+      "epoch": 47.44925946242458,
+      "grad_norm": 3.6988871097564697,
+      "learning_rate": 2.6275370268787712e-05,
+      "loss": 1.915,
+      "step": 86500
+    },
+    {
+      "epoch": 47.723532638507955,
+      "grad_norm": 3.7044730186462402,
+      "learning_rate": 2.613823368074602e-05,
+      "loss": 1.9199,
+      "step": 87000
+    },
+    {
+      "epoch": 47.99780581459133,
+      "grad_norm": 3.6700057983398438,
+      "learning_rate": 2.6001097092704337e-05,
+      "loss": 1.9243,
+      "step": 87500
+    },
+    {
+      "epoch": 48.272078990674714,
+      "grad_norm": 3.89973521232605,
+      "learning_rate": 2.5863960504662642e-05,
+      "loss": 1.8846,
+      "step": 88000
+    },
+    {
+      "epoch": 48.54635216675809,
+      "grad_norm": 4.041015625,
+      "learning_rate": 2.5726823916620958e-05,
+      "loss": 1.8999,
+      "step": 88500
+    },
+    {
+      "epoch": 48.82062534284147,
+      "grad_norm": 3.7937917709350586,
+      "learning_rate": 2.5589687328579264e-05,
+      "loss": 1.9085,
+      "step": 89000
+    },
+    {
+      "epoch": 49.09489851892485,
+      "grad_norm": 4.050382614135742,
+      "learning_rate": 2.545255074053758e-05,
+      "loss": 1.8934,
+      "step": 89500
+    },
+    {
+      "epoch": 49.36917169500823,
+      "grad_norm": 3.809558391571045,
+      "learning_rate": 2.5315414152495888e-05,
+      "loss": 1.8705,
+      "step": 90000
+    },
+    {
+      "epoch": 49.64344487109161,
+      "grad_norm": 3.6460201740264893,
+      "learning_rate": 2.51782775644542e-05,
+      "loss": 1.8904,
+      "step": 90500
+    },
+    {
+      "epoch": 49.917718047174986,
+      "grad_norm": 3.959718704223633,
+      "learning_rate": 2.504114097641251e-05,
+      "loss": 1.8936,
+      "step": 91000
+    },
+    {
+      "epoch": 50.19199122325836,
+      "grad_norm": 3.786888837814331,
+      "learning_rate": 2.490400438837082e-05,
+      "loss": 1.8683,
+      "step": 91500
+    },
+    {
+      "epoch": 50.466264399341746,
+      "grad_norm": 3.477952241897583,
+      "learning_rate": 2.476686780032913e-05,
+      "loss": 1.8634,
+      "step": 92000
+    },
+    {
+      "epoch": 50.74053757542512,
+      "grad_norm": 3.998764991760254,
+      "learning_rate": 2.462973121228744e-05,
+      "loss": 1.8637,
+      "step": 92500
+    },
+    {
+      "epoch": 51.014810751508506,
+      "grad_norm": 4.029101848602295,
+      "learning_rate": 2.449259462424575e-05,
+      "loss": 1.8749,
+      "step": 93000
+    },
+    {
+      "epoch": 51.28908392759188,
+      "grad_norm": 3.8711071014404297,
+      "learning_rate": 2.435545803620406e-05,
+      "loss": 1.8377,
+      "step": 93500
+    },
+    {
+      "epoch": 51.56335710367526,
+      "grad_norm": 3.922783136367798,
+      "learning_rate": 2.421832144816237e-05,
+      "loss": 1.8397,
+      "step": 94000
+    },
+    {
+      "epoch": 51.83763027975864,
+      "grad_norm": 4.025134086608887,
+      "learning_rate": 2.4081184860120682e-05,
+      "loss": 1.8609,
+      "step": 94500
+    },
+    {
+      "epoch": 52.11190345584202,
+      "grad_norm": 3.800508975982666,
+      "learning_rate": 2.394404827207899e-05,
+      "loss": 1.8407,
+      "step": 95000
+    },
+    {
+      "epoch": 52.386176631925395,
+      "grad_norm": 3.944465160369873,
+      "learning_rate": 2.3806911684037304e-05,
+      "loss": 1.8263,
+      "step": 95500
+    },
+    {
+      "epoch": 52.66044980800878,
+      "grad_norm": 4.014648914337158,
+      "learning_rate": 2.3669775095995613e-05,
+      "loss": 1.8374,
+      "step": 96000
+    },
+    {
+      "epoch": 52.934722984092154,
+      "grad_norm": 4.08259916305542,
+      "learning_rate": 2.353263850795392e-05,
+      "loss": 1.8448,
+      "step": 96500
+    },
+    {
+      "epoch": 53.20899616017554,
+      "grad_norm": 3.941981792449951,
+      "learning_rate": 2.3395501919912234e-05,
+      "loss": 1.8131,
+      "step": 97000
+    },
+    {
+      "epoch": 53.483269336258914,
+      "grad_norm": 3.8573715686798096,
+      "learning_rate": 2.3258365331870543e-05,
+      "loss": 1.8169,
+      "step": 97500
+    },
+    {
+      "epoch": 53.75754251234229,
+      "grad_norm": 3.987938165664673,
+      "learning_rate": 2.3121228743828855e-05,
+      "loss": 1.8145,
+      "step": 98000
+    },
+    {
+      "epoch": 54.031815688425674,
+      "grad_norm": 3.652238607406616,
+      "learning_rate": 2.2984092155787164e-05,
+      "loss": 1.8213,
+      "step": 98500
+    },
+    {
+      "epoch": 54.30608886450905,
+      "grad_norm": 3.640587568283081,
+      "learning_rate": 2.2846955567745476e-05,
+      "loss": 1.7938,
+      "step": 99000
+    },
+    {
+      "epoch": 54.58036204059243,
+      "grad_norm": 3.884443759918213,
+      "learning_rate": 2.2709818979703785e-05,
+      "loss": 1.7963,
+      "step": 99500
+    },
+    {
+      "epoch": 54.85463521667581,
+      "grad_norm": 4.245452880859375,
+      "learning_rate": 2.2572682391662098e-05,
+      "loss": 1.811,
+      "step": 100000
+    },
+    {
+      "epoch": 55.128908392759186,
+      "grad_norm": 3.97247576713562,
+      "learning_rate": 2.2435545803620407e-05,
+      "loss": 1.7964,
+      "step": 100500
+    },
+    {
+      "epoch": 55.40318156884257,
+      "grad_norm": 3.8827009201049805,
+      "learning_rate": 2.2298409215578716e-05,
+      "loss": 1.7774,
+      "step": 101000
+    },
+    {
+      "epoch": 55.677454744925946,
+      "grad_norm": 4.079446792602539,
+      "learning_rate": 2.2161272627537028e-05,
+      "loss": 1.7884,
+      "step": 101500
+    },
+    {
+      "epoch": 55.95172792100932,
+      "grad_norm": 4.093244552612305,
+      "learning_rate": 2.2024136039495337e-05,
+      "loss": 1.7904,
+      "step": 102000
+    },
+    {
+      "epoch": 56.226001097092706,
+      "grad_norm": 3.674686908721924,
+      "learning_rate": 2.188699945145365e-05,
+      "loss": 1.7655,
+      "step": 102500
+    },
+    {
+      "epoch": 56.50027427317608,
+      "grad_norm": 4.042862892150879,
+      "learning_rate": 2.174986286341196e-05,
+      "loss": 1.7708,
+      "step": 103000
+    },
+    {
+      "epoch": 56.774547449259465,
+      "grad_norm": 4.069617748260498,
+      "learning_rate": 2.161272627537027e-05,
+      "loss": 1.777,
+      "step": 103500
+    },
+    {
+      "epoch": 57.04882062534284,
+      "grad_norm": 3.68752384185791,
+      "learning_rate": 2.1475589687328583e-05,
+      "loss": 1.7702,
+      "step": 104000
+    },
+    {
+      "epoch": 57.32309380142622,
+      "grad_norm": 3.998215436935425,
+      "learning_rate": 2.133845309928689e-05,
+      "loss": 1.7508,
+      "step": 104500
+    },
+    {
+      "epoch": 57.5973669775096,
+      "grad_norm": 4.300554275512695,
+      "learning_rate": 2.12013165112452e-05,
+      "loss": 1.7634,
+      "step": 105000
+    },
+    {
+      "epoch": 57.87164015359298,
+      "grad_norm": 3.649411678314209,
+      "learning_rate": 2.106417992320351e-05,
+      "loss": 1.7603,
+      "step": 105500
+    },
+    {
+      "epoch": 58.14591332967636,
+      "grad_norm": 4.17492151260376,
+      "learning_rate": 2.0927043335161822e-05,
+      "loss": 1.7441,
+      "step": 106000
+    },
+    {
+      "epoch": 58.42018650575974,
+      "grad_norm": 3.8550057411193848,
+      "learning_rate": 2.0789906747120134e-05,
+      "loss": 1.7381,
+      "step": 106500
+    },
+    {
+      "epoch": 58.694459681843114,
+      "grad_norm": 4.004961967468262,
+      "learning_rate": 2.0652770159078443e-05,
+      "loss": 1.7475,
+      "step": 107000
+    },
+    {
+      "epoch": 58.9687328579265,
+      "grad_norm": 3.936483144760132,
+      "learning_rate": 2.0515633571036756e-05,
+      "loss": 1.7532,
+      "step": 107500
+    },
+    {
+      "epoch": 59.24300603400987,
+      "grad_norm": 3.812488317489624,
+      "learning_rate": 2.037849698299506e-05,
+      "loss": 1.7232,
+      "step": 108000
+    },
+    {
+      "epoch": 59.51727921009325,
+      "grad_norm": 4.185512542724609,
+      "learning_rate": 2.0241360394953374e-05,
+      "loss": 1.7196,
+      "step": 108500
+    },
+    {
+      "epoch": 59.79155238617663,
+      "grad_norm": 4.278858184814453,
+      "learning_rate": 2.0104223806911686e-05,
+      "loss": 1.7356,
+      "step": 109000
+    },
+    {
+      "epoch": 60.06582556226001,
+      "grad_norm": 4.104213714599609,
+      "learning_rate": 1.9967087218869995e-05,
+      "loss": 1.7328,
+      "step": 109500
+    },
+    {
+      "epoch": 60.34009873834339,
+      "grad_norm": 4.215428352355957,
+      "learning_rate": 1.9829950630828307e-05,
+      "loss": 1.7109,
+      "step": 110000
+    },
+    {
+      "epoch": 60.61437191442677,
+      "grad_norm": 4.020122528076172,
+      "learning_rate": 1.9692814042786616e-05,
+      "loss": 1.7144,
+      "step": 110500
+    },
+    {
+      "epoch": 60.888645090510146,
+      "grad_norm": 3.9703729152679443,
+      "learning_rate": 1.955567745474493e-05,
+      "loss": 1.7335,
+      "step": 111000
+    },
+    {
+      "epoch": 61.16291826659353,
+      "grad_norm": 3.900017023086548,
+      "learning_rate": 1.9418540866703238e-05,
+      "loss": 1.7048,
+      "step": 111500
+    },
+    {
+      "epoch": 61.437191442676905,
+      "grad_norm": 4.7137627601623535,
+      "learning_rate": 1.9281404278661547e-05,
+      "loss": 1.6902,
+      "step": 112000
+    },
+    {
+      "epoch": 61.71146461876029,
+      "grad_norm": 4.035908222198486,
+      "learning_rate": 1.914426769061986e-05,
+      "loss": 1.7104,
+      "step": 112500
+    },
+    {
+      "epoch": 61.985737794843665,
+      "grad_norm": 4.041805744171143,
+      "learning_rate": 1.9007131102578168e-05,
+      "loss": 1.7155,
+      "step": 113000
+    },
+    {
+      "epoch": 62.26001097092704,
+      "grad_norm": 3.817702054977417,
+      "learning_rate": 1.886999451453648e-05,
+      "loss": 1.6851,
+      "step": 113500
+    },
+    {
+      "epoch": 62.534284147010425,
+      "grad_norm": 3.8696234226226807,
+      "learning_rate": 1.873285792649479e-05,
+      "loss": 1.6857,
+      "step": 114000
+    },
+    {
+      "epoch": 62.8085573230938,
+      "grad_norm": 3.909179925918579,
+      "learning_rate": 1.85957213384531e-05,
+      "loss": 1.7037,
+      "step": 114500
+    },
+    {
+      "epoch": 63.08283049917718,
+      "grad_norm": 3.8557326793670654,
+      "learning_rate": 1.845858475041141e-05,
+      "loss": 1.6936,
+      "step": 115000
+    },
+    {
+      "epoch": 63.35710367526056,
+      "grad_norm": 4.332828044891357,
+      "learning_rate": 1.832144816236972e-05,
+      "loss": 1.6786,
+      "step": 115500
+    },
+    {
+      "epoch": 63.63137685134394,
+      "grad_norm": 4.454130172729492,
+      "learning_rate": 1.818431157432803e-05,
+      "loss": 1.6768,
+      "step": 116000
+    },
+    {
+      "epoch": 63.90565002742732,
+      "grad_norm": 3.943071126937866,
+      "learning_rate": 1.804717498628634e-05,
+      "loss": 1.6905,
+      "step": 116500
+    },
+    {
+      "epoch": 64.1799232035107,
+      "grad_norm": 4.255739688873291,
+      "learning_rate": 1.7910038398244653e-05,
+      "loss": 1.663,
+      "step": 117000
+    },
+    {
+      "epoch": 64.45419637959408,
+      "grad_norm": 4.027384281158447,
+      "learning_rate": 1.7772901810202962e-05,
+      "loss": 1.6603,
+      "step": 117500
+    },
+    {
+      "epoch": 64.72846955567745,
+      "grad_norm": 3.8232147693634033,
+      "learning_rate": 1.7635765222161274e-05,
+      "loss": 1.6701,
+      "step": 118000
+    },
+    {
+      "epoch": 65.00274273176083,
+      "grad_norm": 3.83734130859375,
+      "learning_rate": 1.7498628634119583e-05,
+      "loss": 1.6797,
+      "step": 118500
+    },
+    {
+      "epoch": 65.27701590784422,
+      "grad_norm": 3.9775922298431396,
+      "learning_rate": 1.7361492046077896e-05,
+      "loss": 1.6441,
+      "step": 119000
+    },
+    {
+      "epoch": 65.55128908392759,
+      "grad_norm": 3.810086250305176,
+      "learning_rate": 1.7224355458036205e-05,
+      "loss": 1.6534,
+      "step": 119500
+    },
+    {
+      "epoch": 65.82556226001097,
+      "grad_norm": 4.3292036056518555,
+      "learning_rate": 1.7087218869994513e-05,
+      "loss": 1.6651,
+      "step": 120000
+    },
+    {
+      "epoch": 66.09983543609435,
+      "grad_norm": 4.046462535858154,
+      "learning_rate": 1.6950082281952826e-05,
+      "loss": 1.654,
+      "step": 120500
+    },
+    {
+      "epoch": 66.37410861217774,
+      "grad_norm": 4.200257778167725,
+      "learning_rate": 1.6812945693911135e-05,
+      "loss": 1.6383,
+      "step": 121000
+    },
+    {
+      "epoch": 66.6483817882611,
+      "grad_norm": 3.98045015335083,
+      "learning_rate": 1.6675809105869447e-05,
+      "loss": 1.6571,
+      "step": 121500
+    },
+    {
+      "epoch": 66.92265496434449,
+      "grad_norm": 3.9323537349700928,
+      "learning_rate": 1.653867251782776e-05,
+      "loss": 1.6486,
+      "step": 122000
+    },
+    {
+      "epoch": 67.19692814042787,
+      "grad_norm": 3.898150682449341,
+      "learning_rate": 1.640153592978607e-05,
+      "loss": 1.6288,
+      "step": 122500
+    },
+    {
+      "epoch": 67.47120131651124,
+      "grad_norm": 3.8490869998931885,
+      "learning_rate": 1.6264399341744377e-05,
+      "loss": 1.6305,
+      "step": 123000
+    },
+    {
+      "epoch": 67.74547449259462,
+      "grad_norm": 4.125833034515381,
+      "learning_rate": 1.6127262753702686e-05,
+      "loss": 1.6393,
+      "step": 123500
+    },
+    {
+      "epoch": 68.01974766867801,
+      "grad_norm": 4.1837286949157715,
+      "learning_rate": 1.5990126165661e-05,
+      "loss": 1.6441,
+      "step": 124000
+    },
+    {
+      "epoch": 68.29402084476138,
+      "grad_norm": 4.150059700012207,
+      "learning_rate": 1.585298957761931e-05,
+      "loss": 1.6144,
+      "step": 124500
+    },
+    {
+      "epoch": 68.56829402084476,
+      "grad_norm": 4.325094223022461,
+      "learning_rate": 1.571585298957762e-05,
+      "loss": 1.6254,
+      "step": 125000
+    },
+    {
+      "epoch": 68.84256719692814,
+      "grad_norm": 3.9832139015197754,
+      "learning_rate": 1.5578716401535932e-05,
+      "loss": 1.636,
+      "step": 125500
+    },
+    {
+      "epoch": 69.11684037301151,
+      "grad_norm": 3.9516079425811768,
+      "learning_rate": 1.544157981349424e-05,
+      "loss": 1.6183,
+      "step": 126000
+    },
+    {
+      "epoch": 69.3911135490949,
+      "grad_norm": 3.982802391052246,
+      "learning_rate": 1.5304443225452554e-05,
+      "loss": 1.6116,
+      "step": 126500
+    },
+    {
+      "epoch": 69.66538672517828,
+      "grad_norm": 4.178645610809326,
+      "learning_rate": 1.516730663741086e-05,
+      "loss": 1.6183,
+      "step": 127000
+    },
+    {
+      "epoch": 69.93965990126166,
+      "grad_norm": 4.045616149902344,
+      "learning_rate": 1.5030170049369171e-05,
+      "loss": 1.6226,
+      "step": 127500
+    },
+    {
+      "epoch": 70.21393307734503,
+      "grad_norm": 4.098151206970215,
+      "learning_rate": 1.4893033461327482e-05,
+      "loss": 1.5999,
+      "step": 128000
+    },
+    {
+      "epoch": 70.48820625342842,
+      "grad_norm": 4.052021026611328,
+      "learning_rate": 1.4755896873285793e-05,
+      "loss": 1.5946,
+      "step": 128500
+    },
+    {
+      "epoch": 70.7624794295118,
+      "grad_norm": 3.7580652236938477,
+      "learning_rate": 1.4618760285244103e-05,
+      "loss": 1.6172,
+      "step": 129000
+    },
+    {
+      "epoch": 71.03675260559517,
+      "grad_norm": 4.1435866355896,
+      "learning_rate": 1.4481623697202416e-05,
+      "loss": 1.6121,
+      "step": 129500
+    },
+    {
+      "epoch": 71.31102578167855,
+      "grad_norm": 3.739654302597046,
+      "learning_rate": 1.4344487109160726e-05,
+      "loss": 1.588,
+      "step": 130000
+    },
+    {
+      "epoch": 71.58529895776194,
+      "grad_norm": 4.380291938781738,
+      "learning_rate": 1.4207350521119034e-05,
+      "loss": 1.5998,
+      "step": 130500
+    },
+    {
+      "epoch": 71.8595721338453,
+      "grad_norm": 3.7885782718658447,
+      "learning_rate": 1.4070213933077344e-05,
+      "loss": 1.5958,
+      "step": 131000
+    },
+    {
+      "epoch": 72.13384530992869,
+      "grad_norm": 4.236293792724609,
+      "learning_rate": 1.3933077345035655e-05,
+      "loss": 1.5956,
+      "step": 131500
+    },
+    {
+      "epoch": 72.40811848601207,
+      "grad_norm": 4.205173015594482,
+      "learning_rate": 1.3795940756993966e-05,
+      "loss": 1.5766,
+      "step": 132000
+    },
+    {
+      "epoch": 72.68239166209544,
+      "grad_norm": 4.034268379211426,
+      "learning_rate": 1.3658804168952278e-05,
+      "loss": 1.5912,
+      "step": 132500
+    },
+    {
+      "epoch": 72.95666483817882,
+      "grad_norm": 3.9170260429382324,
+      "learning_rate": 1.3521667580910589e-05,
+      "loss": 1.5897,
+      "step": 133000
+    },
+    {
+      "epoch": 73.23093801426221,
+      "grad_norm": 3.925799608230591,
+      "learning_rate": 1.33845309928689e-05,
+      "loss": 1.5765,
+      "step": 133500
+    },
+    {
+      "epoch": 73.50521119034559,
+      "grad_norm": 4.052227973937988,
+      "learning_rate": 1.324739440482721e-05,
+      "loss": 1.5775,
+      "step": 134000
+    },
+    {
+      "epoch": 73.77948436642896,
+      "grad_norm": 4.2378034591674805,
+      "learning_rate": 1.3110257816785517e-05,
+      "loss": 1.5834,
+      "step": 134500
+    },
+    {
+      "epoch": 74.05375754251234,
+      "grad_norm": 4.073320388793945,
+      "learning_rate": 1.297312122874383e-05,
+      "loss": 1.5812,
+      "step": 135000
+    },
+    {
+      "epoch": 74.32803071859573,
+      "grad_norm": 4.102873802185059,
+      "learning_rate": 1.283598464070214e-05,
+      "loss": 1.5588,
+      "step": 135500
+    },
+    {
+      "epoch": 74.6023038946791,
+      "grad_norm": 4.223252773284912,
+      "learning_rate": 1.269884805266045e-05,
+      "loss": 1.5785,
+      "step": 136000
+    },
+    {
+      "epoch": 74.87657707076248,
+      "grad_norm": 4.320130825042725,
+      "learning_rate": 1.2561711464618761e-05,
+      "loss": 1.5686,
+      "step": 136500
+    },
+    {
+      "epoch": 75.15085024684586,
+      "grad_norm": 4.706448078155518,
+      "learning_rate": 1.2424574876577072e-05,
+      "loss": 1.5688,
+      "step": 137000
+    },
+    {
+      "epoch": 75.42512342292923,
+      "grad_norm": 4.024387359619141,
+      "learning_rate": 1.2287438288535381e-05,
+      "loss": 1.5563,
+      "step": 137500
+    },
+    {
+      "epoch": 75.69939659901262,
+      "grad_norm": 3.9221880435943604,
+      "learning_rate": 1.2150301700493692e-05,
+      "loss": 1.5712,
+      "step": 138000
+    },
+    {
+      "epoch": 75.973669775096,
+      "grad_norm": 4.27291202545166,
+      "learning_rate": 1.2013165112452002e-05,
+      "loss": 1.5654,
+      "step": 138500
+    },
+    {
+      "epoch": 76.24794295117937,
+      "grad_norm": 4.373564720153809,
+      "learning_rate": 1.1876028524410313e-05,
+      "loss": 1.5449,
+      "step": 139000
+    },
+    {
+      "epoch": 76.52221612726275,
+      "grad_norm": 4.030310153961182,
+      "learning_rate": 1.1738891936368624e-05,
+      "loss": 1.5571,
+      "step": 139500
+    },
+    {
+      "epoch": 76.79648930334614,
+      "grad_norm": 4.002580165863037,
+      "learning_rate": 1.1601755348326934e-05,
+      "loss": 1.5568,
+      "step": 140000
+    },
+    {
+      "epoch": 77.07076247942952,
+      "grad_norm": 4.0623369216918945,
+      "learning_rate": 1.1464618760285245e-05,
+      "loss": 1.5545,
+      "step": 140500
+    },
+    {
+      "epoch": 77.34503565551289,
+      "grad_norm": 4.049304485321045,
+      "learning_rate": 1.1327482172243554e-05,
+      "loss": 1.5442,
+      "step": 141000
+    },
+    {
+      "epoch": 77.61930883159627,
+      "grad_norm": 3.891969680786133,
+      "learning_rate": 1.1190345584201866e-05,
+      "loss": 1.5464,
+      "step": 141500
+    },
+    {
+      "epoch": 77.89358200767965,
+      "grad_norm": 4.165316104888916,
+      "learning_rate": 1.1053208996160177e-05,
+      "loss": 1.5509,
+      "step": 142000
+    },
+    {
+      "epoch": 78.16785518376302,
+      "grad_norm": 4.1472249031066895,
+      "learning_rate": 1.0916072408118487e-05,
+      "loss": 1.5375,
+      "step": 142500
+    },
+    {
+      "epoch": 78.44212835984641,
+      "grad_norm": 4.173414707183838,
+      "learning_rate": 1.0778935820076796e-05,
+      "loss": 1.5345,
+      "step": 143000
+    },
+    {
+      "epoch": 78.71640153592979,
+      "grad_norm": 3.9279398918151855,
+      "learning_rate": 1.0641799232035107e-05,
+      "loss": 1.5381,
+      "step": 143500
+    },
+    {
+      "epoch": 78.99067471201316,
+      "grad_norm": 4.222446441650391,
+      "learning_rate": 1.0504662643993418e-05,
+      "loss": 1.552,
+      "step": 144000
+    },
+    {
+      "epoch": 79.26494788809654,
+      "grad_norm": 3.8020248413085938,
+      "learning_rate": 1.0367526055951728e-05,
+      "loss": 1.5217,
+      "step": 144500
+    },
+    {
+      "epoch": 79.53922106417993,
+      "grad_norm": 3.8790934085845947,
+      "learning_rate": 1.0230389467910039e-05,
+      "loss": 1.5346,
+      "step": 145000
+    },
+    {
+      "epoch": 79.8134942402633,
+      "grad_norm": 4.49297571182251,
+      "learning_rate": 1.009325287986835e-05,
+      "loss": 1.5354,
+      "step": 145500
+    },
+    {
+      "epoch": 80.08776741634668,
+      "grad_norm": 4.024161338806152,
+      "learning_rate": 9.95611629182666e-06,
+      "loss": 1.5326,
+      "step": 146000
+    },
+    {
+      "epoch": 80.36204059243006,
+      "grad_norm": 3.997326374053955,
+      "learning_rate": 9.818979703784971e-06,
+      "loss": 1.5133,
+      "step": 146500
+    },
+    {
+      "epoch": 80.63631376851345,
+      "grad_norm": 4.163906574249268,
+      "learning_rate": 9.68184311574328e-06,
+      "loss": 1.5252,
+      "step": 147000
+    },
+    {
+      "epoch": 80.91058694459682,
+      "grad_norm": 4.333358287811279,
+      "learning_rate": 9.54470652770159e-06,
+      "loss": 1.5356,
+      "step": 147500
+    },
+    {
+      "epoch": 81.1848601206802,
+      "grad_norm": 4.201995372772217,
+      "learning_rate": 9.407569939659903e-06,
+      "loss": 1.5191,
+      "step": 148000
+    },
+    {
+      "epoch": 81.45913329676358,
+      "grad_norm": 3.8196020126342773,
+      "learning_rate": 9.270433351618212e-06,
+      "loss": 1.5168,
+      "step": 148500
+    },
+    {
+      "epoch": 81.73340647284695,
+      "grad_norm": 4.32403039932251,
+      "learning_rate": 9.133296763576522e-06,
+      "loss": 1.5214,
+      "step": 149000
+    },
+    {
+      "epoch": 82.00767964893033,
+      "grad_norm": 4.165477752685547,
+      "learning_rate": 8.996160175534833e-06,
+      "loss": 1.5242,
+      "step": 149500
+    },
+    {
+      "epoch": 82.28195282501372,
+      "grad_norm": 4.155007362365723,
+      "learning_rate": 8.859023587493144e-06,
+      "loss": 1.5022,
+      "step": 150000
+    },
+    {
+      "epoch": 82.55622600109709,
+      "grad_norm": 4.049638748168945,
+      "learning_rate": 8.721886999451453e-06,
+      "loss": 1.5172,
+      "step": 150500
+    },
+    {
+      "epoch": 82.83049917718047,
+      "grad_norm": 4.376342296600342,
+      "learning_rate": 8.584750411409765e-06,
+      "loss": 1.5097,
+      "step": 151000
+    },
+    {
+      "epoch": 83.10477235326385,
+      "grad_norm": 4.413540363311768,
+      "learning_rate": 8.447613823368076e-06,
+      "loss": 1.5163,
+      "step": 151500
+    },
+    {
+      "epoch": 83.37904552934722,
+      "grad_norm": 3.754805326461792,
+      "learning_rate": 8.310477235326386e-06,
+      "loss": 1.5011,
+      "step": 152000
+    },
+    {
+      "epoch": 83.6533187054306,
+      "grad_norm": 4.167300224304199,
+      "learning_rate": 8.173340647284695e-06,
+      "loss": 1.5026,
+      "step": 152500
+    },
+    {
+      "epoch": 83.92759188151399,
+      "grad_norm": 4.1614861488342285,
+      "learning_rate": 8.036204059243006e-06,
+      "loss": 1.5101,
+      "step": 153000
+    },
+    {
+      "epoch": 84.20186505759737,
+      "grad_norm": 4.183162212371826,
+      "learning_rate": 7.899067471201317e-06,
+      "loss": 1.4981,
+      "step": 153500
+    },
+    {
+      "epoch": 84.47613823368074,
+      "grad_norm": 4.0559539794921875,
+      "learning_rate": 7.761930883159627e-06,
+      "loss": 1.4965,
+      "step": 154000
+    },
+    {
+      "epoch": 84.75041140976413,
+      "grad_norm": 4.252512454986572,
+      "learning_rate": 7.624794295117937e-06,
+      "loss": 1.5019,
+      "step": 154500
+    },
+    {
+      "epoch": 85.02468458584751,
+      "grad_norm": 3.9115328788757324,
+      "learning_rate": 7.4876577070762485e-06,
+      "loss": 1.5138,
+      "step": 155000
+    },
+    {
+      "epoch": 85.29895776193088,
+      "grad_norm": 4.217545032501221,
+      "learning_rate": 7.350521119034559e-06,
+      "loss": 1.492,
+      "step": 155500
+    },
+    {
+      "epoch": 85.57323093801426,
+      "grad_norm": 3.974954128265381,
+      "learning_rate": 7.21338453099287e-06,
+      "loss": 1.4897,
+      "step": 156000
+    },
+    {
+      "epoch": 85.84750411409765,
+      "grad_norm": 4.266519069671631,
+      "learning_rate": 7.07624794295118e-06,
+      "loss": 1.4996,
+      "step": 156500
+    },
+    {
+      "epoch": 86.12177729018102,
+      "grad_norm": 3.8507697582244873,
+      "learning_rate": 6.93911135490949e-06,
+      "loss": 1.4891,
+      "step": 157000
+    },
+    {
+      "epoch": 86.3960504662644,
+      "grad_norm": 4.050006866455078,
+      "learning_rate": 6.801974766867801e-06,
+      "loss": 1.4848,
+      "step": 157500
+    },
+    {
+      "epoch": 86.67032364234778,
+      "grad_norm": 4.006500720977783,
+      "learning_rate": 6.664838178826111e-06,
+      "loss": 1.4946,
+      "step": 158000
+    },
+    {
+      "epoch": 86.94459681843115,
+      "grad_norm": 4.2527289390563965,
+      "learning_rate": 6.527701590784421e-06,
+      "loss": 1.495,
+      "step": 158500
+    },
+    {
+      "epoch": 87.21886999451453,
+      "grad_norm": 4.087696552276611,
+      "learning_rate": 6.390565002742732e-06,
+      "loss": 1.4834,
+      "step": 159000
+    },
+    {
+      "epoch": 87.49314317059792,
+      "grad_norm": 3.9683475494384766,
+      "learning_rate": 6.253428414701043e-06,
+      "loss": 1.48,
+      "step": 159500
+    },
+    {
+      "epoch": 87.7674163466813,
+      "grad_norm": 4.009182453155518,
+      "learning_rate": 6.116291826659353e-06,
+      "loss": 1.4799,
+      "step": 160000
+    },
+    {
+      "epoch": 88.04168952276467,
+      "grad_norm": 3.9172310829162598,
+      "learning_rate": 5.979155238617663e-06,
+      "loss": 1.4827,
+      "step": 160500
+    },
+    {
+      "epoch": 88.31596269884805,
+      "grad_norm": 3.920940399169922,
+      "learning_rate": 5.842018650575974e-06,
+      "loss": 1.4772,
+      "step": 161000
+    },
+    {
+      "epoch": 88.59023587493144,
+      "grad_norm": 4.178516387939453,
+      "learning_rate": 5.704882062534284e-06,
+      "loss": 1.4831,
+      "step": 161500
+    },
+    {
+      "epoch": 88.8645090510148,
+      "grad_norm": 4.068806171417236,
+      "learning_rate": 5.567745474492595e-06,
+      "loss": 1.4796,
+      "step": 162000
+    },
+    {
+      "epoch": 89.13878222709819,
+      "grad_norm": 3.923023223876953,
+      "learning_rate": 5.430608886450905e-06,
+      "loss": 1.4734,
+      "step": 162500
+    },
+    {
+      "epoch": 89.41305540318157,
+      "grad_norm": 4.0538411140441895,
+      "learning_rate": 5.293472298409216e-06,
+      "loss": 1.4675,
+      "step": 163000
+    },
+    {
+      "epoch": 89.68732857926494,
+      "grad_norm": 4.289505481719971,
+      "learning_rate": 5.156335710367526e-06,
+      "loss": 1.4812,
+      "step": 163500
+    },
+    {
+      "epoch": 89.96160175534833,
+      "grad_norm": 4.2184247970581055,
+      "learning_rate": 5.019199122325837e-06,
+      "loss": 1.4761,
+      "step": 164000
+    },
+    {
+      "epoch": 90.23587493143171,
+      "grad_norm": 4.014777183532715,
+      "learning_rate": 4.8820625342841474e-06,
+      "loss": 1.4659,
+      "step": 164500
+    },
+    {
+      "epoch": 90.51014810751508,
+      "grad_norm": 4.025433540344238,
+      "learning_rate": 4.744925946242457e-06,
+      "loss": 1.4701,
+      "step": 165000
+    },
+    {
+      "epoch": 90.78442128359846,
+      "grad_norm": 4.117000102996826,
+      "learning_rate": 4.607789358200768e-06,
+      "loss": 1.4745,
+      "step": 165500
+    },
+    {
+      "epoch": 91.05869445968185,
+      "grad_norm": 4.047626495361328,
+      "learning_rate": 4.4706527701590785e-06,
+      "loss": 1.4693,
+      "step": 166000
+    },
+    {
+      "epoch": 91.33296763576523,
+      "grad_norm": 4.17887020111084,
+      "learning_rate": 4.333516182117389e-06,
+      "loss": 1.4624,
+      "step": 166500
+    },
+    {
+      "epoch": 91.6072408118486,
+      "grad_norm": 4.2437639236450195,
+      "learning_rate": 4.196379594075699e-06,
+      "loss": 1.4665,
+      "step": 167000
+    },
+    {
+      "epoch": 91.88151398793198,
+      "grad_norm": 3.7711315155029297,
+      "learning_rate": 4.0592430060340105e-06,
+      "loss": 1.4731,
+      "step": 167500
+    },
+    {
+      "epoch": 92.15578716401536,
+      "grad_norm": 4.002791404724121,
+      "learning_rate": 3.92210641799232e-06,
+      "loss": 1.4642,
+      "step": 168000
+    },
+    {
+      "epoch": 92.43006034009873,
+      "grad_norm": 4.0743231773376465,
+      "learning_rate": 3.7849698299506313e-06,
+      "loss": 1.4584,
+      "step": 168500
+    },
+    {
+      "epoch": 92.70433351618212,
+      "grad_norm": 4.080685138702393,
+      "learning_rate": 3.647833241908941e-06,
+      "loss": 1.4623,
+      "step": 169000
+    },
+    {
+      "epoch": 92.9786066922655,
+      "grad_norm": 4.304593563079834,
+      "learning_rate": 3.510696653867252e-06,
+      "loss": 1.4682,
+      "step": 169500
+    },
+    {
+      "epoch": 93.25287986834887,
+      "grad_norm": 4.447428226470947,
+      "learning_rate": 3.3735600658255624e-06,
+      "loss": 1.4535,
+      "step": 170000
+    },
+    {
+      "epoch": 93.52715304443225,
+      "grad_norm": 4.22756814956665,
+      "learning_rate": 3.236423477783873e-06,
+      "loss": 1.4592,
+      "step": 170500
+    },
+    {
+      "epoch": 93.80142622051564,
+      "grad_norm": 4.293380260467529,
+      "learning_rate": 3.0992868897421833e-06,
+      "loss": 1.4632,
+      "step": 171000
+    },
+    {
+      "epoch": 94.075699396599,
+      "grad_norm": 4.07041072845459,
+      "learning_rate": 2.962150301700494e-06,
+      "loss": 1.4605,
+      "step": 171500
+    },
+    {
+      "epoch": 94.34997257268239,
+      "grad_norm": 4.039161205291748,
+      "learning_rate": 2.825013713658804e-06,
+      "loss": 1.4551,
+      "step": 172000
+    },
+    {
+      "epoch": 94.62424574876577,
+      "grad_norm": 4.1246795654296875,
+      "learning_rate": 2.687877125617115e-06,
+      "loss": 1.456,
+      "step": 172500
+    },
+    {
+      "epoch": 94.89851892484916,
+      "grad_norm": 4.026761054992676,
+      "learning_rate": 2.550740537575425e-06,
+      "loss": 1.4512,
+      "step": 173000
+    },
+    {
+      "epoch": 95.17279210093253,
+      "grad_norm": 4.5864715576171875,
+      "learning_rate": 2.4136039495337357e-06,
+      "loss": 1.4575,
+      "step": 173500
+    },
+    {
+      "epoch": 95.44706527701591,
+      "grad_norm": 4.117992401123047,
+      "learning_rate": 2.2764673614920463e-06,
+      "loss": 1.4475,
+      "step": 174000
+    },
+    {
+      "epoch": 95.72133845309929,
+      "grad_norm": 4.155096530914307,
+      "learning_rate": 2.1393307734503565e-06,
+      "loss": 1.4581,
+      "step": 174500
+    },
+    {
+      "epoch": 95.99561162918266,
+      "grad_norm": 4.28767204284668,
+      "learning_rate": 2.002194185408667e-06,
+      "loss": 1.4521,
+      "step": 175000
+    },
+    {
+      "epoch": 96.26988480526605,
+      "grad_norm": 4.1511077880859375,
+      "learning_rate": 1.8650575973669776e-06,
+      "loss": 1.4488,
+      "step": 175500
+    },
+    {
+      "epoch": 96.54415798134943,
+      "grad_norm": 4.336985111236572,
+      "learning_rate": 1.727921009325288e-06,
+      "loss": 1.4534,
+      "step": 176000
+    },
+    {
+      "epoch": 96.8184311574328,
+      "grad_norm": 4.181045055389404,
+      "learning_rate": 1.5907844212835987e-06,
+      "loss": 1.45,
+      "step": 176500
+    },
+    {
+      "epoch": 97.09270433351618,
+      "grad_norm": 4.217624187469482,
+      "learning_rate": 1.453647833241909e-06,
+      "loss": 1.4498,
+      "step": 177000
+    },
+    {
+      "epoch": 97.36697750959956,
+      "grad_norm": 3.8873023986816406,
+      "learning_rate": 1.3165112452002194e-06,
+      "loss": 1.4507,
+      "step": 177500
+    },
+    {
+      "epoch": 97.64125068568293,
+      "grad_norm": 4.3951191902160645,
+      "learning_rate": 1.17937465715853e-06,
+      "loss": 1.451,
+      "step": 178000
+    },
+    {
+      "epoch": 97.91552386176632,
+      "grad_norm": 4.1204118728637695,
+      "learning_rate": 1.0422380691168404e-06,
+      "loss": 1.4484,
+      "step": 178500
+    },
+    {
+      "epoch": 98.1897970378497,
+      "grad_norm": 4.278495788574219,
+      "learning_rate": 9.05101481075151e-07,
+      "loss": 1.4431,
+      "step": 179000
+    },
+    {
+      "epoch": 98.46407021393308,
+      "grad_norm": 4.186399459838867,
+      "learning_rate": 7.679648930334613e-07,
+      "loss": 1.4493,
+      "step": 179500
+    },
+    {
+      "epoch": 98.73834339001645,
+      "grad_norm": 4.110637187957764,
+      "learning_rate": 6.308283049917719e-07,
+      "loss": 1.4418,
+      "step": 180000
+    },
+    {
+      "epoch": 99.01261656609984,
+      "grad_norm": 3.9559993743896484,
+      "learning_rate": 4.936917169500823e-07,
+      "loss": 1.4407,
+      "step": 180500
+    },
+    {
+      "epoch": 99.28688974218322,
+      "grad_norm": 4.4722418785095215,
+      "learning_rate": 3.565551289083928e-07,
+      "loss": 1.4421,
+      "step": 181000
+    },
+    {
+      "epoch": 99.56116291826659,
+      "grad_norm": 4.151792526245117,
+      "learning_rate": 2.1941854086670326e-07,
+      "loss": 1.4469,
+      "step": 181500
+    },
+    {
+      "epoch": 99.83543609434997,
+      "grad_norm": 4.128389835357666,
+      "learning_rate": 8.228195282501371e-08,
+      "loss": 1.4463,
+      "step": 182000
+    },
+    {
+      "epoch": 100.0,
+      "step": 182300,
+      "total_flos": 3.157662139522744e+17,
+      "train_loss": 2.261507166926833,
+      "train_runtime": 62730.7046,
+      "train_samples_per_second": 185.96,
+      "train_steps_per_second": 2.906
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 182300,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.157662139522744e+17,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}