Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Rebase to current master + some minor fixes #4

Open
wants to merge 291 commits into
base: pynq
Choose a base branch
from
Open

Rebase to current master + some minor fixes #4

wants to merge 291 commits into from

Conversation

jmduarte
Copy link

@jmduarte jmduarte commented Feb 6, 2021

  • Rebase to current master + some minor fixes

vloncar and others added 30 commits September 28, 2020 16:17
@vloncar
Add global pooling layers to config util
8ab6f0a
@vloncar
Fuse steps in Conv2D to reduce latency
eb59f82
@vloncar
Support for strategy and ReuseFactor in Conv2D
c5c6e9d
@thesps
Add support for trace of ApplyAlpha layer.
eed3e2f
@vloncar
Merge pull request fastmachinelearning#222 from thesps/config_fix
254ee89 
Fix config generation for activations again...
@vloncar
Pipeline one level up only if we don't use packing
0ac3795
@vloncar
Pipeline on pack level only if using latency strategy
934bddc
@vloncar
Merge pull request fastmachinelearning#215 from Duchstf/garnet_fix
63b12c1 
Move pointer to include garnet test, update example fetching scheme
@vloncar
Restore submodule
46f0bff
@vloncar
New streaming implementation of the Dense layer
7d348c9
@vloncar
ReuseFactor in streaming BN
358c904
@vloncar
Support for binary/ternary optimizations
18fc2ac
@vloncar
Make dense_latency behave the same ith streams
b3e00b8
@vloncar
Add numeric_limits so compile() function works
3abb8e4
@vloncar
Some pooling fixes
cf14689
@vloncar
Global pooling cleanup
8df9ad4
@vloncar
Add loop names to clear things up during synthesis
b28925d
@vloncar
Merge branch 'master' into cnn_stream
491b236
@vloncar
Stringify precision before plotting
fee4323
@vloncar
Proper imports in conv1d header
28d42ad
@vloncar
Handle no-bias fusing of Dense and BN
46edf5a
@vloncar @thesps
Handle no-bias fusing of Dense and BN
8619037
@vloncar
Stable softmax implementation for io_stream
4ce99b4
@vloncar
"Legacy" softmax implementation
9ca877f
@thesps
Merge pull request fastmachinelearning#229 from thesps/use_bias_fix
06d569f 
Handle no-bias fusing of Dense and BN
@vloncar @thesps
SkipOptimizers config parameter
b5c12ba
@vloncar @thesps
Print cosim latency report
f785847
@vloncar @thesps
Fix test bench event counter
8c42af0
@vloncar
Consistent file naming for conv2d implementations
0430067
@vloncar
Rename Conv1D headers
9d73d17
@jmduarte
Copy link
Author

Are you sure this is safe?

Actually, I was wondering what the logic is for the original mapping. In the Tiny ResNet model, I see identical results with the mapping I updated above.

But maybe there are other/corner cases I'm not considering?

@thesps
Copy link
Owner

thesps commented Apr 27, 2021

I just did a bit more testing with isolated quantized_bits(W,I). It does indeed seem to be safe 👍 it could be that what originally seemed like the need for that extra bit was actually just 'alpha'...

thesps and others added 12 commits May 11, 2021 10:19
thesps pushed a commit that referenced this pull request Jul 22, 2021
@Keb-L
Merge pull request #4 from vloncar/line_buffer_conv
ae50bfe 
Fix Conv1D encoded impl after function renaming
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@thesps thesps thesps left review comments

@GiuseppeDiGuglielmo GiuseppeDiGuglielmo GiuseppeDiGuglielmo left review comments

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.