jhj

4 ms·

1.
by jhj 6mo ago
This might include people working in lumber camps in places like Siberia, "mercenaries" in Ukraine, people in NK-managed restaurants in China, Laos etc, or similar efforts that have been reported on, where the average revenue per
2.
by jhj 8mo ago
These flops are not the same. The 2013 phone flops are fp32, the A13 flops look to be fp32 as well (not entirely sure), while the Cray numbers (like the rest of the HPC industry) are fp64 (Cray 1 predates what would become IEEE 754 binary64
3.
by jhj 1y ago
Unlike quantization, dimensionality reduction/low rank approximation, distillation etc, lossless compression is an always-correct addition to any ML system as you are computing the same thing you did before, the only question is if it
4.
by jhj 1y ago
Not really, it's just adding some data transposition (coalescing individual bytes from the data words together) and an option to use a LZ/dictionary-type compressor to compress redundant things. But an LZ-type compressor doesn&#x2
5.
by jhj 1y ago
This is just a consequence of the fact that bfloat16 has a very high dynamic range which is not all used. People like hyperparameters that look like 0.01 not 10^10, even though there is the same fractional precision available at each expone
6.
by jhj 2y ago
re #3, if your RSU windfall is substantially large, you might be eligible for the 100%/110% safe harbor that won't penalize you for tax underpayments (assuming you are a US taxpayer) e.g., you make $200K in 2024 and $5 million in
7.
by jhj 2y ago
I have some of (possibly the?) cheapest residential electric power in the US, at 5.58 cents per kWh all-in cost here in Wyoming, 90%+ hydropower. Absolute lowest cold here each year will be around -30 F / -34 C (there will be several n
8.
by jhj 2y ago
Brute-force indices are usually arithmetic bound (e.g., GEMM). Cell-probe based indices are usually memory bandwidth bound (IVF, LSH bucketing, etc). Graph-based indices are usually memory latency bound (traversing linked lists / graph
9.
by jhj 2y ago
If you have a limited number of long range ICBMs then you will likely prefer more directly military targets rather than a manufacturing facility which would likely only start to matter for a conflict months into combat, which itself is a sc
10.
by jhj 2y ago
As someone who has worked in this space (approximate compute) on both GPUs and in silicon in my research, the power consumption claims are completely bogus, as are the accuracy claims: > In this section, we show that L-Mul is more precis
11.
by jhj 2y ago
> The first thing to consider is the register pressure. Increasing the number of registers per thread to optimize for ILP can lead to register spilling when the register file is exhausted Kernels should almost never use local memory (exc
12.
by jhj 2y ago
Aiming for higher occupancy is not always a desired solution, what frequently matters more is avoiding global memory latencies by retaining more data in registers and/or shared memory. This was first noted in 2010 and is still true tod
13.
by jhj 2y ago
Remote start is accidental carbon monoxide poisoning waiting to happen if your garage is directly connected to your residence. I live in an area with brutal winters in Wyoming and just bought a new Ford Bronco, wish I could fully disable it
14.
by jhj 2y ago
The original sin here is that original 1980s designs carry over: the processor retains FP unit state, rather than each instruction indicating what subnormal flush mode (or rounding mode or whatever) one wishes to use with no retained FP uni
15.
by jhj 2y ago
A less risky use is to use the model to choose compilation flags and pass orderings many (but not all) of which (in theory) should always be correct but that's more of a problem of the compiler itself than of the model if it produces i
16.
by jhj 2y ago
While the median is much much lower, there are a couple of thousand individual contributor SWEs (non-managers) between Google, Meta and a few other big-ish tech companies who make >$1 million/year (steady state, does not depend upon
17.
by jhj 2y ago
Construction costs here (Teton County, WY) are significantly higher than CA or most places in the US due to labor constraints (we have the highest average per capita income in the US, yet an ~80 : 1 median house price : median yearly income
18.
by jhj 2y ago
It’s probably more California regulations than the wildfire risk per se? The direct backyard of my house in Wyoming is Bridger-Teton National Forest, wooded mountainous wilderness for miles with its trees abutting my property. A wildfire in
19.
by jhj 2y ago
Not just MPI over a network. We can compress floats, send them over NVLink or PCIe to another GPU in the same host, and decompress and it can be faster than sending data raw between GPUs, that's the premise behind dietgpu even (it&#x27
20.
by jhj 2y ago
People in the HPC/classical supercomputing space have done this sort of thing for a while. There's a fair amount of literature on lossless floating point compression, such as Martin Burtscher's work or stuff out of LLNL (fpzi
21.
by jhj 2y ago
I live near Yellowstone in Wyoming. The park is a lot more massive than Zion, usually involving multi-hour drives to get around, and there are multiple roads in the park, all of which don't necessarily see the same levels of traffic. T
22.
by jhj 3y ago
Finite field log/antilog lookup tables are used for efficient-ish multiplication, similar to addition/subtraction tables used for logarithmic number systems.
23.
by jhj 3y ago
> By scaling the numbers first by taking the log, multiplication becomes addition and addition becomes x + log1p(exp(y - x)). Addition/subtraction in a logarithmic number system is way more expensive than what you would spend on mul
24.
by jhj 3y ago
There is no good geometry to be exploited, and the query vectors might be (and are usually) distributed quite differently than the indexed vectors. For Euclidean (L2) distance indexes where the vectors are partitioned based on geometry (e.g
25.
by jhj 3y ago
Speaking as an author of one of the primary libraries for doing this stuff (faiss), it is not because it is still an open ended research problem on how approximate high-dimensional dense or sparse nearest neighbor should work, let alone max
26.
by jhj 3y ago
All approximate k-NN algorithms involve tradeoffs, with the main tradeoffs being usually fidelity (recall, vector reconstruction error, end-to-end application error with respect to some metric etc), speed (to query, in either serial or batc
27.
by jhj 3y ago
The implementations that I am aware of, including the one in Faiss which I wrote (described in detail in https://arxiv.org/abs/1702.08734 ), do not index the vector based on its PQ encoding (e.g., in IVFPQ). The IVF cel
28.
by jhj 3y ago
I don't know why PQ is listed as an "indexing strategy". It's a vector compression/quantization technique, not a means of partitioning the search space. You could encode vectors with PQ when using brute-force/f
29.
by jhj 3y ago
> Either CUDA will open up, if it is to survive or open API use will spread. I don't really think so, at least not anytime soon while the hardware functionality continues to evolve so much, and while they seem to be concentrating on
30.
by jhj 3y ago
The two main classes of approximate k-NN algorithms om high dimensional data would involve either (1) linear table scans (e.g., IVF / cell-probe indexes; brute-force exact indexing or the coarse quantizer for an IVF index) or (2) graph

More ›