Pom down under
Top 20
No unfortunately, but found the Mrs
BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Afternoon DingoBorat ,Hey Esq, where's the stats for today? Or are you busy tucking into that vanilla infused brew of yours?
How big was the wall at 45 when it fell, still around 2.5 million shares?..
At the close of trade , I snuk off into the garden for an
& stretch the eyes.From what I saw at the time , Yep that line at 0.45 was wiped , actually think there was a little CXA trade which popped it to 0.4525
Good day all round .
Regards,
Esq.
The wall was 2.5 million shares. When it fell it was 4 million and sold in about 1 minute. From 54 million shares to 58 million. A further 2 million sold at 24.5,24 and 23.5 c .Hey Esq, where's the stats for today? Or are you busy tucking into that vanilla infused brew of yours?
How big was the wall at 45 when it fell, still around 2.5 million shares?..
MDhere
Top 20
Here it is DB @DingoBoratAfternoon DingoBorat ,
At the close of trade , I snuk off into the garden for an
From what I saw at the time , Yep that line at 0.45 was wiped , actually think there was a little CXA trade which popped it to 0.4525
Good day all round .
Regards,
Esq.
it got eaten at 3.51.21 volume 1,858,514 eaten up at .45c then 20 seconds later came that cxa trade for 10k vol at .4525. So that savy hungry buy nearly 2mil worth at .45c was impressive. someone with real money is buying up the shares.
| Sum of Volume | Column Labels | |||
| Row Labels | 0.4475 | 0.45 | 0.4525 | Grand Total |
| 3:51:21 PM | 9159 | 1849355 | 1858514 | |
| 3:51:41 PM | 20000 | 20000 | ||
| Grand Total | 9159 | 1849355 | 20000 | 1878514 |
Falling Knife
Regular
Three trades covered the 2.9 million shares on offerHey Esq, where's the stats for today? Or are you busy tucking into that vanilla infused brew of yours?
How big was the wall at 45 when it fell, still around 2.5 million shares?..
COMMSEC COURSE OF SALE DL
Soaked up a few more before pulling back.
I'm no expert but it the Bid is sitting there at 45c and they are gone in these three trades it indicates buying up not selling down
Huge volume for the day.
CES next week. What will it bring or reveal.
Other companies might demo something new revealing who may have been in an DNA whilst in development.
ie Smart Airbags previously.
Boab
I wish I could paint like Vincent
buena suerte :-)
BOB Bank of Brainchip
You look beautiful… 
Looking good 
Germany
,
Germany
, 
DingoBorat
Slim
This is a recent clip, which is a good amalgamation of various interviews with Musk, from the past year..
Visuals are all, already in the public domain, but it's an enjoyable watch.
From 1min 15secs (under safety concerns) Musk describes the need for some kind of safety switch, that is completely independant of the internet/Cloud.
Obviously, straight away, I thought of AKIDA, as it would need (in my opinion) to be an "intelligent" adaptive switch, that was monitoring the robotic systems, for anomalies.
There would probably be many varied solutions though, but something that could learn and adapt, without Cloud updates?..
A bit like how Sonny, in iRobot was built with 2 independent operating systems (brains) so it was able to reason against the nefarious orders of V.I.K.I.(basically Terminator's Skynet, for those unfamiliar with iRobot).
There's a scene where Sonny is running and V.I.K.I. is trying to control "him" and he basically says "Yeah I hear you, but it just doesn't make sense to me".
This is Sonny, lying about "murdering" his Creator.
This is all extremely close, to no longer being science fiction..
Absolute Classic film by the way..
Visuals are all, already in the public domain, but it's an enjoyable watch.
From 1min 15secs (under safety concerns) Musk describes the need for some kind of safety switch, that is completely independant of the internet/Cloud.
Obviously, straight away, I thought of AKIDA, as it would need (in my opinion) to be an "intelligent" adaptive switch, that was monitoring the robotic systems, for anomalies.
There would probably be many varied solutions though, but something that could learn and adapt, without Cloud updates?..
A bit like how Sonny, in iRobot was built with 2 independent operating systems (brains) so it was able to reason against the nefarious orders of V.I.K.I.(basically Terminator's Skynet, for those unfamiliar with iRobot).
There's a scene where Sonny is running and V.I.K.I. is trying to control "him" and he basically says "Yeah I hear you, but it just doesn't make sense to me".
This is Sonny, lying about "murdering" his Creator.
This is all extremely close, to no longer being science fiction..
Absolute Classic film by the way..
Last edited:
Boab
I wish I could paint like Vincent
Das ist gut! C'est fantastique! Taught to me by Ian Dury and the block heads IYKYKLooking good
Germany
Smoothsailing
Regular
At the 4min mark he talks about local LLM running on a raspberry pie.
The move is still only a 'baby' at 6 days of age with a red day thrown in. Momentum is slowing but still bullish.
It appears to be driven by expectation of deals.
For those not aware how shorting works:
Shorter believes the stock is overbought/overvalued.
Shorter then borrows the stock at a cost to the shorter.
The shorter then sells the stock on the ASX - just like we buy and sell - no different.
The shorter hopes the share price will fall. If it falls the shorter buys the stock at a lower price, hands them back to the lender and takes a profit.
If the share price goes up the shorter buys back the stock at a lower price than he/she sold it for, hands them to the lender and takes a loss.
If enough shorters enter the market in a similar time frame it puts a fair bit of downward pressure on the share price and likely falls. That is what is meant by posters when they say shorters are manipulating the price.
Conversely, if the price goes against the shorters and they exit in a similar time frame the SP likely goes up. This is what some posters call a short squeeze.
So far shorters are being overwhelmed by the rush of buyers.
Shorter believes the stock is overbought/overvalued.
Shorter then borrows the stock at a cost to the shorter.
The shorter then sells the stock on the ASX - just like we buy and sell - no different.
The shorter hopes the share price will fall. If it falls the shorter buys the stock at a lower price, hands them back to the lender and takes a profit.
If the share price goes up the shorter buys back the stock at a lower price than he/she sold it for, hands them to the lender and takes a loss.
If enough shorters enter the market in a similar time frame it puts a fair bit of downward pressure on the share price and likely falls. That is what is meant by posters when they say shorters are manipulating the price.
Conversely, if the price goes against the shorters and they exit in a similar time frame the SP likely goes up. This is what some posters call a short squeeze.
So far shorters are being overwhelmed by the rush of buyers.
Pom down under
Top 20
Glad I’m thick skinned
Same reason why I don't post selfies myself.Glad I’m thick skinned
I am doing everyone here a favour
RobjHunt
Regular
In my opinion it has becom quite apparent that mobs out there are taking notice with what’s happened with our little nipper over the past few weeks. A couple of “contracts” so to speak. I wonder how long they were under an NDA?
How many more of the NDA mobs have tested enough to reveal themselves directly?
Pantene Peeps
How many more of the NDA mobs have tested enough to reveal themselves directly?
Pantene Peeps
Fullmoonfever
Top 20
Thoughts if you get the chance @Diogenese
TIA.
All sounds somewhat familiar from Carnegie.
Aswin Ak
-
December 29, 2024
Designing neuromorphic sensory processing units (NSPUs) based on Temporal Neural Networks (TNNs) is a highly challenging task due to the reliance on manual, labor-intensive hardware development processes. TNNs have been identified as highly promising for real-time edge AI applications, mainly because they are energy-efficient and bio-inspired. However, available methodologies lack automation and are not very accessible. Consequently, the design process becomes complex, time-consuming, and requires specialized knowledge. It is through overcoming these challenges that one can unlock the full potential of TNNs for efficient and scalable processing of sensory signals.
The current approaches to TNN development are fragmented workflows, as software simulations and hardware designs are handled separately. Advancements such as ASAP7 and TNN7 libraries made some aspects of hardware efficient but remain proprietary tools that require significant expertise. The fragmentation of the process restricts usability, prevents the easier exploration of design configurations with increased computational overhead, and can’t be used for more application-specific rapid prototyping or large-scale deployment purposes.
Researchers at Carnegie Mellon University introduce TNNGen, a unified and automated framework for designing TNN-based NSPUs. The innovation lies in the integration of software-based functional simulation with hardware generation in a single streamlined workflow. It combines a PyTorch-based simulator, modeling spike-timing dynamics and evaluating application-specific metrics, with a hardware generator that automates RTL generation and layout design using PyVerilog. Through the utilization of TNN7 custom macros and the integration of a variety of libraries, this framework realizes considerable enhancements in simulation velocity as well as physical design. Additionally, its predictive abilities facilitate precise forecasting of silicon metrics, thereby diminishing the dependency on computationally demanding EDA tools.
TNNGen is organized around two principal elements. The functional simulator, constructed using PyTorch, accommodates adaptable TNN configurations, allowing for swift examination of various model architectures. It has GPU acceleration and accurate spike-timing modeling, thus ensuring high simulation speed and accuracy. The hardware generator converts PyTorch models into optimized RTL and physical layouts. Using libraries such as TNN7 and customized TCL scripts, it automates synthesis and place-and-route processes while being compatible with multiple technology nodes like FreePDK45 and ASAP7.
TNNGen achieves excellent performance in both clustering accuracy and hardware efficiency. The TNN designs for time-series clustering tasks show competitive performance with the best deep-learning techniques while drastically reducing the utilization of computational resources. The approach brings major energy efficiency improvements, obtaining a reduction in die area and leakage power compared to conventional approaches. In addition, the runtime of the design is dramatically reduced, especially for larger designs, which benefit most from the optimized workflows. Moreover, the comprehensive forecasting instrument provides accurate estimations of hardware parameters, allowing researchers to evaluate design viability without the necessity of engaging in physical hardware procedures. Taken together, these findings position TNNGen as a viable approach for streamlining and expediting the creation of energy-efficient neuromorphic systems.
TNNGen is the next step in the fully automated development of TNN-based NSPUs by unifying simulation and hardware generation into an accessible, efficient framework. The approach addressed key challenges in the manual design process and made this tool much more scalable and usable for edge AI applications. Future work would involve extending its capabilities toward support for more complex TNN architectures and a much wider range of applications to become a critical enabler of sustainable neuromorphic computing.
Check out the Paper. All credit for this research goes to the researchers of this project
TIA.
All sounds somewhat familiar from Carnegie.
Redirect Notice
www.google.com.au
CMU Researchers Introduce TNNGen: An AI Framework that Automates Design of Temporal Neural Networks (TNNs) from PyTorch Software Models to Post-Layout Netlists
ByAswin Ak
-
December 29, 2024
Designing neuromorphic sensory processing units (NSPUs) based on Temporal Neural Networks (TNNs) is a highly challenging task due to the reliance on manual, labor-intensive hardware development processes. TNNs have been identified as highly promising for real-time edge AI applications, mainly because they are energy-efficient and bio-inspired. However, available methodologies lack automation and are not very accessible. Consequently, the design process becomes complex, time-consuming, and requires specialized knowledge. It is through overcoming these challenges that one can unlock the full potential of TNNs for efficient and scalable processing of sensory signals.
The current approaches to TNN development are fragmented workflows, as software simulations and hardware designs are handled separately. Advancements such as ASAP7 and TNN7 libraries made some aspects of hardware efficient but remain proprietary tools that require significant expertise. The fragmentation of the process restricts usability, prevents the easier exploration of design configurations with increased computational overhead, and can’t be used for more application-specific rapid prototyping or large-scale deployment purposes.
Researchers at Carnegie Mellon University introduce TNNGen, a unified and automated framework for designing TNN-based NSPUs. The innovation lies in the integration of software-based functional simulation with hardware generation in a single streamlined workflow. It combines a PyTorch-based simulator, modeling spike-timing dynamics and evaluating application-specific metrics, with a hardware generator that automates RTL generation and layout design using PyVerilog. Through the utilization of TNN7 custom macros and the integration of a variety of libraries, this framework realizes considerable enhancements in simulation velocity as well as physical design. Additionally, its predictive abilities facilitate precise forecasting of silicon metrics, thereby diminishing the dependency on computationally demanding EDA tools.
TNNGen is organized around two principal elements. The functional simulator, constructed using PyTorch, accommodates adaptable TNN configurations, allowing for swift examination of various model architectures. It has GPU acceleration and accurate spike-timing modeling, thus ensuring high simulation speed and accuracy. The hardware generator converts PyTorch models into optimized RTL and physical layouts. Using libraries such as TNN7 and customized TCL scripts, it automates synthesis and place-and-route processes while being compatible with multiple technology nodes like FreePDK45 and ASAP7.
TNNGen achieves excellent performance in both clustering accuracy and hardware efficiency. The TNN designs for time-series clustering tasks show competitive performance with the best deep-learning techniques while drastically reducing the utilization of computational resources. The approach brings major energy efficiency improvements, obtaining a reduction in die area and leakage power compared to conventional approaches. In addition, the runtime of the design is dramatically reduced, especially for larger designs, which benefit most from the optimized workflows. Moreover, the comprehensive forecasting instrument provides accurate estimations of hardware parameters, allowing researchers to evaluate design viability without the necessity of engaging in physical hardware procedures. Taken together, these findings position TNNGen as a viable approach for streamlining and expediting the creation of energy-efficient neuromorphic systems.
TNNGen is the next step in the fully automated development of TNN-based NSPUs by unifying simulation and hardware generation into an accessible, efficient framework. The approach addressed key challenges in the manual design process and made this tool much more scalable and usable for edge AI applications. Future work would involve extending its capabilities toward support for more complex TNN architectures and a much wider range of applications to become a critical enabler of sustainable neuromorphic computing.
Check out the Paper. All credit for this research goes to the researchers of this project
DingoBorat
Slim
Supply and Demand.
Supply and Demand.
Supply and Demand.
Demand is currently very, very strong and going by today's volume, which was the most in 10 months and the highest in this run, it doesn't look like slowing on a Friday, which is "traditionally" a strong day.
Again
.. Friday's action is going to be interesting..
From the strength of buying, the Big Buyer/s? Is/are, quite happy taking large chunks, up to at least 45 cents.
If the volume at these levels, is no longer offered, the price could go higher again, as there is an apparent urgency, to build a substantial position.
There's been around 10% of the number of shares on issue traded now, in the high 30's to 40's range, that's a good soak.
This will soon become a bit of a new trading range, if Dave is again unable to knock this stiff rise on the head, with a cold spoon.
And he really has no choice but to try, so additional borrowed shares, will have again been taken out today.
Personally, even without further contracts or licences, I think we "should" have been trading in the 50 to 60 cent price range, just on everything, the Company has put in place, over the last couple of years.
Normally, technology Companies trade at very large valuations, even pre-profit, due to their growth potential (if they have the goods, which BrainChip has proven, but this is not "usually" appreciated on a resource focused exchange, like the ASX. BrainChip is changing that however).
This is very different, from the Mercedes run, as we are in a far stronger position, both Technology offering wise and Ecosystem wise.
I don't think we can get to those heights "now" (which would equate to around $1.50 after dilution since) but we can build a solid base for launch, on additional IP licence wins.
Or this could all just be taking the BRN thoroughbred horse, out for a good run, burning some poor Daves along the way and making a stack of cash.
It doesn't look like that, seems too strong, but there are obviously Entities and Individuals, with the financial backing, to play that game, if they wish.
Disregarding that possibility..
Supply and Demand, is the only thing that matters right now and what we are seeing, is just a taste, of what's to come.
(with additional solid progress).
Good Fortune to all Holders.
Supply and Demand.
Supply and Demand.
Demand is currently very, very strong and going by today's volume, which was the most in 10 months and the highest in this run, it doesn't look like slowing on a Friday, which is "traditionally" a strong day.
Again
.. Friday's action is going to be interesting..From the strength of buying, the Big Buyer/s? Is/are, quite happy taking large chunks, up to at least 45 cents.
If the volume at these levels, is no longer offered, the price could go higher again, as there is an apparent urgency, to build a substantial position.
There's been around 10% of the number of shares on issue traded now, in the high 30's to 40's range, that's a good soak.
This will soon become a bit of a new trading range, if Dave is again unable to knock this stiff rise on the head, with a cold spoon.
And he really has no choice but to try, so additional borrowed shares, will have again been taken out today.
Personally, even without further contracts or licences, I think we "should" have been trading in the 50 to 60 cent price range, just on everything, the Company has put in place, over the last couple of years.
Normally, technology Companies trade at very large valuations, even pre-profit, due to their growth potential (if they have the goods, which BrainChip has proven, but this is not "usually" appreciated on a resource focused exchange, like the ASX. BrainChip is changing that however).
This is very different, from the Mercedes run, as we are in a far stronger position, both Technology offering wise and Ecosystem wise.
I don't think we can get to those heights "now" (which would equate to around $1.50 after dilution since) but we can build a solid base for launch, on additional IP licence wins.
Or this could all just be taking the BRN thoroughbred horse, out for a good run, burning some poor Daves along the way and making a stack of cash.
It doesn't look like that, seems too strong, but there are obviously Entities and Individuals, with the financial backing, to play that game, if they wish.
Disregarding that possibility..
Supply and Demand, is the only thing that matters right now and what we are seeing, is just a taste, of what's to come.
(with additional solid progress).
Good Fortune to all Holders.
Last edited:
Similar threads
