正在加载视频...

视频加载失败

Due to overwhelming high demand, a number of orders from our first batch have unfortunately been moved to the second batch, which is now 90% ordered! Our manufacturers are hard at work to keep up with strong demand and we estimate that the second batch will now be sent...

27,821 次观看 • 2 年前 •via X (Twitter)

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Batch Normalization by hand ✍️ ~ 7 steps walkthrough below Batch normalization is common practice for improving training and achieving faster convergence. It sounds simple. But it is often misunderstood. 🤔 Does batch normalization involve trainable parameters, tunable hyper-parameters, or both? 🤔 Is batch normalization applied to inputs, features, weights, biases, or outputs? 🤔 How is batch normalization different from layer normalization? So I drew and calculated one entirely by hand. Goal: normalize a mini-batch of 4 examples to mean 0 and variance 1, then let the network scale it back. = 1. Given = A mini-batch of 4 training examples, each with 3 features. = 2. Linear layer = Let us multiply by the weights and add the biases. Batch norm sits after this, which answers the second question: what gets normalized is features, not inputs, weights or biases. = 3. ReLU = We apply the activation, and -2 becomes 0. Negative values are suppressed before any statistic is taken. = 4. Batch statistics = Let us compute the sum, mean, variance and standard deviation, one row at a time. A row is a feature and the four columns are the four examples, so every number here measures one feature against the rest of the batch. That is the "batch" in batch normalization, and it is exactly what layer normalization does not do. The statistics are rounded to whole numbers, which is what keeps the rest of the page doable in pen. = 5. Shift to mean 0 = We subtract the mean, in green. The four values in each feature now average to zero. = 6. Scale to variance 1 = Let us divide by the standard deviation, in orange. Each feature now has variance one, whatever scale it arrived at. = 7. Scale and shift = We multiply by a linear transformation and pass the result on. The diagonal and the last column are trainable, so having just forced every feature to mean 0 and variance 1, we hand the network the means to undo it. The outputs: Mean of each feature = [2, 1, 2] Std dev of each feature = [1, 1, 2] To the next layer = [2, -2, 2, 0], [-3, 3, 6, -3], [2, 0, 1, 2] The answers: 🤔 Both. The scale and shift are trainable, the statistics are not. Epsilon and the momentum on the running statistics are the hyper-parameters, and one mini-batch by hand needs neither. 🤔 Features, after the linear layer, not inputs, weights or biases. 🤔 Batch norm measures across the batch, one feature at a time. Layer norm measures across the features, one example at a time. 💾 Save this post!

Tom Yeh

20,848 次观看 • 1 个月前

Many $IREN investors are waiting for an announcement from the company, to remove any doubts around Sweetwater 2 being included in batch zero baseload. At OnlyFrans, we prefer to follow the development facts, and as you can see in the time-lapse from the past 3 weeks, the progress is undeniable. Matching the rendering, the top-view from the satellite shows that the bulk substation earthworks are showing the typical IREN proportions. Roughly 2/3 of the footprint for the 345kV network, and the 1/3 smaller footprint above for the 138kV network. Further to the left, near the top of the image, we can see that the first signs for the two 300MW Primary substations are taking shape. I expect civil works to start in Q4, and after the stormwater works at Sweetwater 1 are done, I expect the main focus of the work to be stormwater management, in parallel with building out the bulk, and primary substations. Needless to say, Sweetwater 2 is batch zero baseload. Sweetwater 2 is an early example of the market moving before the rulebook did. In March 2025, IREN committed roughly $31m to secure 600MW with AEP — ~$51.7k/MW. Nearly a year later, Texas settled on a $50k/MW financial-security benchmark (under CIAC) for large loads entering Batch Zero. This means $IREN and AEP (the interconnecting TDSP for SW2) were already operating with the kind of financial commitment Texas would later require to separate real projects from speculative load requests. In other words: IREN was positioning its projects as real/financially committed rather than speculative well before Batch Zero forced everyone to do it. Think about it, if anyone would be compliant to the batch zero baseload requirements, wouldn't it be the company that had completed CIAC (Contribution in Aid of Construction) a year before it became mandatory for everyone else? I see a site progressing as if it's approved, and a connection progress that was completed before it became common knowledge. And somehow there are people that doubt that IREN will add 600MW of power next year? Now if you are interested to find out what else there is in batch zero baseload from IREN, you know where to go. Long $IREN

Frans Bakker

102,932 次观看 • 11 天前