TensorFlow Benchmarks

https://github.com/soumith/convnet-benchmarks/issues/66

by sherjilozair • 11 years ago

98 55 11 years ago

vonnik• 11 years ago

It takes 4x as long as CuDNN on AlexNet due to lack of in-place ops. What is up with that?

mtgx• 11 years ago

It's almost like Google wanted everyone to use slow obsolete software and keep the really good stuff for itself, while still making it look like they're doing a great thing for the community.

tfgg• 11 years ago

I think it makes more sense that:

a) It's Google, they can throw hardware at a problem such as out of memory

b) TensorFlow makes methods development so much easier that it's worth the loss of performance

c) It's early days and the compute graph scheduler has lots of opportunities, and is designed, for optimization, and in a more flexible fashion than other frameworks.

When I work on developing methods for scientific code, I worry more about whether the code is bug free/easy to understand and that it's giving the right answer. I don't usually worry about performance unless I'm actually not able to run things. Especially since when developing stuff you waste way more time on runs with bugs - I dread to think what my (published / unpublished) CPU hour ratio is. If the new approach allows less buggy implementations then that's a resource win.

IanCal3• 11 years ago

b) TensorFlow makes methods development so much easier that it's worth the loss of performance

Indeed, if TensorFlow means I can try out an idea with 1 day of coding and 2 days of training rather than 3 days of coding and 1 day of training then I can spend a day drinking cocktails and reading books and still be finished sooner.

Based on the tutorials it seems like I'd be able to pretty quickly build a translation pipeline, and in fact there's an implementation of that I think I'll try. If it takes a week or two to train, that's fine by me, I've got other things to be getting on with.

k__• 11 years ago

Maybe they got a pile of old software they give to their PR people.

"Hey, we would throw this away, but maybe this could be used to shine up our brand a bit"

abalos• 11 years ago

Seeing as Google has one of the most powerful computing networks in the world, I bet it's more that they can just throw resources at it until it's fast enough. This helps justify the trade off between speed and ease of use.

johanneskanybal2• 11 years ago

Like a factor of 4 matters much if your framework makes it easier to scale out horizontally. Also, I'm sure it will get optimized over time.

IanCal3• 11 years ago

Interesting benchmarks. One hopefully constructive critique, if you say things go out of memory, it'd be really useful to know what your setup is. Maybe you've got a big array of massive GPUs or you're running it on a more normal consumer GPU+box.

brianchu• 11 years ago

According to: https://github.com/soumith/convnet-benchmarks, it's an NVIDIA TitanX (12GB GPU memory), which is pretty much the top of the line GPU for training neural nets. From my limited experience, most deep learning in research - including much of the state of the art - is done on single GPUs. (My guess is that if your model doesn't fit in 12GB, your model has way too many parameters to practically train anyway).

IanCal3• 11 years ago

Ah thank you for that, I hadn't clocked on that this was opened by the person who ran the repo. I should have checked the main readme.

programnature• 11 years ago

In the announcement they called this the "reference implementation".

mark_l_watson• 11 years ago

From Google's perspective it is probably more about how TensorFlow scales out horizontally. If a researcher fires off a Borg run (or whatever they use now) and the job takes a few thousand CPUs, no problem, at least for research.

They must have better optimization a for running in production, such as in place operations.

zellyn• 11 years ago

This. Even if it takes 10,000 networked cores to replace the ultimate souped-up single machine loaded with GPUs, you can go from 10,000 to 20,000.

andmarios• 11 years ago

I wonder if this is the reason I can't get a preemptible VM from GCE to stay up for more than one hour in the last days.

argonaut3• 11 years ago

Which would be great, except for the fact that Google has not open-sourced the distributed version of TensorFlow. So this is just PR/speculation.

aljones• 11 years ago

So how horizontally scalable is it?

sanxiyn• 11 years ago

As far as I know, TensorFlow's scaling behavior has not been published.

For its predecessor DistBelief, here are some quotes from NIPS 2012 paper:

"The moderately sized speech model runs fastest on 8 machines, computing 2.2x faster than using a single machine. (Models were configured to use no more than 20 cores per machine.) Partitioning the model on more than 8 machines actually slows training, as network overhead starts to dominate in the fully-connected network structure and there is less work for each machine to perform with more partitions." ("The moderately sized" model here has 42 million model parameters. Check the paper for details.)

"In contrast, the much larger, locally-connected image models can benefit from using many more machines per model replica. The largest model, with 1.7 billion parameters benefits the most, giving a speedup of more than 12x using 81 machines. For these large models using more machines continues to increase speed, but with diminishing returns."

http://research.google.com/pubs/pub40565.html

dave_sullivan5• 11 years ago

I'm going to say something unpopular, but horizontally-scaled deep learning is overkill for most applications.

Can anyone here present a use case where they have personally needed horizontal scaling because a Titan X couldn't fit what they were trying to do? It's the thing I hear talked about the most and used the least.

The biggest misunderstanding I've heard is "I have petabytes of data so I need multi-GPU", but NNs are trained on batches of data, so it doesn't matter. It's the model that must fit, but 12gb can fit a big model.

The whole "who can build the biggest net" is like building the fastest car: it's for show, the slower ones are way more cost effective and get you there just as well.

Keras is my present deep learning library recommendation--it's like torch 7 but in python and theano. Multi-GPU is roadmap because there's a lot more going on in deep learning that acts as a bigger differentiator than horizontal scaling.

For the record, when I say scaling, I'm not talking about having multiple machines serve multiple requests (that's a good reason to purchase more machines), I'm talking about spreading your gradient computations and/or model weights across machines or GPUs.

freerobby• 11 years ago

Would image training for video recognition be a good example of this?

kastnerkyle7• 11 years ago

For big LSTMs and long-ish sequences, the intermediate gradients can take up a huge amount of memory - often more than the model parameters themselves. In my experience it is mostly big LSTMs that need the 12GB+ GPUs. You can reduce the batch size to help this a bit, or train using trucated BPTT but RNN training is already a slow, sequential business.

Of course, there are no clear wins (generally, losses) in computation by scaling RNNs horizontally on GPUs - but sometimes you really do need more than 12GB.

kajecounterhack• 11 years ago

horizontally-scaled deep learning is overkill for most applications

It's difficult to say what "most applications are" but many algos currently seem bottlenecked by how fast you can load data into your GFX card (hint: it's slow). So what you're saying about batching is correct, but it does matter.

varelse6• 11 years ago

I suspect this is a just-so-tale because the capability is so new and the GPU programming skill required to implement it efficiently is so rare. TLDR: just say no to parameter server (most of the time).

But IMO the real obstacle to horizontal scaling is the communication between servers not the usefulness of doing so. Within a server, one can pack 8 TitanX/M40 GPUs with high speed 13 GB/s P2P communication between them (and up to 16 GPUs in various unproven science project servers). That's ~50 TFLOPs and 96 GB in a box. That rocks. Just ask the guys at mindori (if they ever ship that is).

http://www.mindori.com

But between servers lies a sippy straw of 100 Gb/s Infiniband at best or worst-case, ~1 Gb/s on AWS GPU servers with freaky nearest neighbor weather. If you can't make efficient use of 8 GPUs in a single box, I agree, don't bother breaking out to the next server.

That said, frameworks like mxnet have opened the floodgates to experimenting with larger models and more distributed training algorithms. Time will tell if this pans out. But 2 years ago, Andrew Ng's group showed 12 GTX 680s distributed across 3 servers kicking the crap out of Google Brain. I expect more of this, not less, in the near future.

moultano• 11 years ago

Can anyone here present a use case where they have personally needed horizontal scaling

The embeddings for text models with a real world vocabulary can get very very large, especially if multiple languages are involved.

mtw3• 11 years ago

The rest of us don't have "a few thousand CPUs" though. Seen this way, TensorFlow is largely inaccessible and makes more sense to focus on existing librairies optimized for single cpus

johanneskanybal2• 11 years ago

I don't think you understand what this library is for. You can pick up a graphic card for $100 and voila now you got 1000's of cores. To your defense the comment you answered seems to have mixed a few random things together too.

varelse6• 11 years ago

TensorFlow is currently a single-server solution pthreads app.

But assuming you're serious and not dabbling, you can build a ~27 TFLOP server for <$7K in parts or buy it assembled for ~$9K.

http://exxactcorp.com/index.php/solution/solu_detail/225

Or if you just hate the thought of money burning a hole in your pocket, pay NVIDIA >$15K for the same thing with a pretty NVIDIA logo:

https://developer.nvidia.com/devbox

If you're dabbling, just go buy a $1000 TitanX and you're instantly a minor deep learning superpower.

iraphael• 11 years ago

I don't know how much this matters, but an issue similar to this had been raised here: https://github.com/tensorflow/tensorflow/issues/120

The responses seem to show that the way you implement things can make a big difference in runtime. Perhaps the scripts used for benchmarking can be further optimized?

That said, the lack of in-place operations might be surprising (although it has been said that they are coming)

donthateyo• 11 years ago

Until now, I've seen two responses to Google's TensorFlow from Facebook employees. Yann Le Cunn seemed to really challenge Jeff Dean about TensorFlow's scalability [1] and this benchmark puts TensorFlow down there in all the measures it tested for. I can't ignore the possibility that this criticism of TensorFlow from Facebook employees (while factually correct and constructive) might be driven by some competition and jealousy.

[1] https://www.youtube.com/watch?v=90-S1M7Ny_o&t=39m

argonaut3• 11 years ago

I think I'll give the benefit of the doubt to, you know, the pioneer of deep learning, inventor of convolutional neural networks, and (co)-inventor of the backpropagation algorithm.

discardorama• 11 years ago

Just because you invent an algorithm, it doesn't mean you know how to implement it in the most efficient way possible.

I think the key to TensorFlow is not how fast it runs on 1 machine; but how fast it runs on 10,000.

Consider map-reduce (Hadoop). Sure, you can sort 1GB data on a single machine 10x faster (using /usr/bin/sort) than using Hadoop on that machine; but make the data 1TB and add 1000 machines, now lets see how fast you can sort with /usr/bin/sort!

kastnerkyle7• 11 years ago

This benchmark basically shows that releasing TensorFlow with cudnn v2 backend support hurts - v2 is quite a bit slower than v3 (current) and v4 (upcoming). TF has announced that they will update to v4 support, which should help quite a bit - but when many hobbyists and researchers are developing on one or two GPUs performance on that scale is more important (for them) than infinite scalability.

It is not surprising that a tool developed and focused on `Google scale` work has some imperfections in a wildly different setting. The question is - will they (or some dedicated contributor) speed up a use case the business itself may not have a use for? My gut feeling is that they will, but these things usually don`t happen overnight. Torch, Theano, and Caffe have years of work put into them and have largely been focused on the one to two GPU case.

smhx2• 11 years ago

I've done an apples to apples, TensorFlow + CuDNN R2 vs Torch + CuDNN R2

varelse6• 11 years ago

You might be seeing the final gasps of Google's longstanding and now-reversed anti-GPU stance in TF's initial GPU performance.

NVIDIA support alone will make sure TF knocks it out of the park down the road. IMO it's crazy to consider the first release the final say on TF's GPU performance. Caffe, Torch, and Theano have a huge head start (and lots of pre-existing technical support from NVIDIA).

The biggest limitation I see right now is that their multi-GPU algorithms are really simple and inefficient. That will change I'm sure now that they're getting benchmarked against everyone else.

smhx2• 11 years ago

I'm the one who runs the benchmarks. It's sad that you put such a twist to the whole thing.

I've been running convnet-benchmarks forever now, and I've been running them independently on separate personal hardware. I do this as a hobby.

I've done an apples-to-apples comparison, and my benchmark review only puts the facts forward, I dont attack them.

If you read my other social media comments, I've been pretty positive about TensorFlow, and I've even put in some groundwork to write Torch bindings for it.

Please stop spewing nonsense interpolated from like 2 super-weak data points.

emcq• 11 years ago

Is this surprising? I think it's expected, when most published numbers from Google suggest that this is only 2x over DistBelief, which Adam Coates improved upon 6x (http://www.cs.stanford.edu/~acoates/papers/CoatesHuvalWangWu...) in 2013. Coming from a HPC background, DistBelief (and sometimes MR in general) has always felt like a square being forced into a round hole. Sure it "scales" but it's not how you would build a system from scratch to meet the same computational needs. Moving data around is much slower than doing extra math.

DNNs provide some interesting opportunities for domain specific optimization with lossy math because empirically it seems that the solutions dont require significant precision. Something like a general Tensor library holds you back from using some of the tricks like synchronization-free hogwild, where it wouldn't be appropriate in other scenarios where more exact computation is necessary.

One interesting step towards this is their implementation of lossy compression with a custom 16-bit half float representation. However a lot of people are dialing down the precision even further, including one of the authors of TensorFlow (http://petewarden.com/2015/05/23/why-are-eight-bits-enough-f...).

Scott Gray, one of the principle engineers at Nervana Systems, has a very wise reply in the github thread - they aren't doing much to optimize for memory locality, and wont have amazing speedups until that happens.

TensorFlow Benchmarks

Related Stories