TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Advancements in machine learning for machine learning

323 点作者 atg_abhishek超过 1 年前

8 条评论

owlbite超过 1 年前
These ML-compilers are being overhyped. It&#x27;s all the same trade-off as a traditional compiler: you get a lot more throughput than hiring a specialist performance programmer, but the latter will typically outperform, possibly by orders of magnitude.<p>These things are inferior at many levels: - Algorithmic: These things aren&#x27;t feeding back to their human masters tips and tricks on how to modify the network to go faster beyond some very basic signals. - Loss of intent: ML network designers are specifying architecture in python, and by the time it&#x27;s gone through many layers of lowering, you can get some complete garbage. Highly efficient garbage, but still garbage. (recent example, we caught one of these compilers doing a slice update operation by first forming the range of all possible indices to the array, slicing that to get indices to update, and then doing a scatter; we replaced it with a single memcpy call). - Inefficient kernels. Every time we see the output of these compilers go head-to-head with an expert assembly programmer, the compiler loses, often by 30%+. This always seems like the sort of thing that should be easy to solve, but given no-one seems to have cracked it in the past 50 years, it&#x27;s obviously not as simple as it sounds.
评论 #38664379 未加载
评论 #38666067 未加载
评论 #38666405 未加载
评论 #38666442 未加载
评论 #38667190 未加载
评论 #38664312 未加载
dalbasal超过 1 年前
Can anyone bring this down to earth for me?<p>What&#x27;s the actual state of these &quot;ML compilers&quot; currently, and what is rhe near term promise?
评论 #38663119 未加载
评论 #38664223 未加载
评论 #38662567 未加载
aconz2超过 1 年前
summary: improve prediction of run-time performance of a computation graph using GNN, they use an embedding dictionary for each node&#x27;s opcode along with some other node features (eg shape, bits, window size, see [1]), they released a big dataset of these graphs in [2] with varying XLA compilation configurations and their resulting perf on TPUs, they did some stuff to improve prediction on bigger graphs than before in [3] by partitioning the graph (METIS graph partition, new to me) and other training things<p>This is only about predicting performance of a given graph and not about improving&#x2F;suggesting&#x2F;editing a new equivalent graph. As in FunSearch, models which have decent predictive power could be used with evolutionary search.<p>[1] <a href="https:&#x2F;&#x2F;github.com&#x2F;google-research-datasets&#x2F;tpu_graphs#features">https:&#x2F;&#x2F;github.com&#x2F;google-research-datasets&#x2F;tpu_graphs#featu...</a><p>[2] TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphs <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2308.13490" rel="nofollow noreferrer">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2308.13490</a><p>[3] Learning Large Graph Property Prediction via Graph Segment Training <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2305.12322" rel="nofollow noreferrer">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2305.12322</a>
potac超过 1 年前
Can anyone explain how conv works in that graph. You have a tensor of shape [2,4,16] and you convolve with a kernel of shape [4,16,8] and that gives you a [2,8] tensor? How&#x27;s that possible?
评论 #38663496 未加载
GreedClarifies超过 1 年前
How’s Gemini looking?
评论 #38662030 未加载
seydor超过 1 年前
What about transformer itself, any indication that it is optimal in some way?
ikers超过 1 年前
Feels like they bury the lede with the first paragraph, but otherwise cool stuff!
RyanShook超过 1 年前
The pace that ML seems to be advancing right now is amazing. I don’t believe in the singularity but it’s changing software and then society in ways no one can predict.
评论 #38661701 未加载
评论 #38662316 未加载
评论 #38661654 未加载
评论 #38662023 未加载
评论 #38661943 未加载
评论 #38662384 未加载