Skip to content

Instantly share code, notes, and snippets.

View Tagussan's full-sized avatar
🥝
Need kiwi

Tagussan

🥝
Need kiwi
32 followers · 28 following
  • The University of Tokyo
  • Tokyo
View GitHub Profile
@Tagussan
Tagussan / pdfcropIt.bat
Created January 5, 2023 04:50
pdfcrop %~nx1
@Tagussan
Tagussan / keymap_googleIME.txt
Created July 18, 2022 01:07
status key command
Composition Backspace Backspace
Composition Ctrl a MoveCursorToBeginning
Composition Ctrl Backspace Backspace
Composition Ctrl d MoveCursorRight
Composition Ctrl Down MoveCursorToEnd
Composition Ctrl e MoveCursorToBeginning
Composition Ctrl Enter Commit
Composition Ctrl f MoveCursorToEnd
Composition Ctrl g Delete
@Tagussan
Tagussan / debug.hpp
Last active April 2, 2022 11:10
C++ debug header. Use -DDBUG_ option when compiling. Many thanks for https://www.creativ.xyz/dump-cpp-652/
#include <iostream>
#include <map>
#include <set>
#include <string>
#include <vector>
using namespace std;
#define repi(itr, ds) for (auto itr = ds.begin(); itr != ds.end(); itr++)
@Tagussan
Tagussan / compile_fbnet.py
Created December 3, 2021 09:23
Code to reproduce error on https://github.com/aws/aws-neuron-sdk/issues/369
import torch
import timm
import torch.neuron
# original model
model = timm.create_model("fbnetc_100", pretrained=True)
model.eval()
# number of parameters
@Tagussan
Tagussan / log
Created November 2, 2021 01:56
INFO:Neuron:There are 1 ops of 1 different types in the TorchScript that are not compiled by neuron-cc: aten::adaptive_max_pool2d, (For more information see https://github.com/aws/aws-neuron-sdk/blob/master/release-notes/neuron-cc-ops/neuron-cc-ops-pytorch.md)
INFO:Neuron:Number of arithmetic operators (pre-compilation) before = 306, fused = 295, percent fused = 96.41%
INFO:Neuron:Compiling function _NeuronGraph$868 with neuron-cc
INFO:Neuron:Compiling with command line: '/home/ubuntu/anaconda3/envs/aws_neuron_pytorch_p36/bin/neuron-cc compile /tmp/tmp1qen5wtt/graph_def.pb --framework TENSORFLOW --pipeline compile SaveTemps --output /tmp/tmp1qen5wtt/graph_def.neff --io-config {"inputs": {"0:0": [[1, 5, 3, 384, 384], "float32"]}, "outputs": ["Reshape_2:0"]} --verbose 35'
.[01:48:52] /opt/brazil-pkg-cache/packages/DmlcTvm/DmlcTvm-1.7.2.0/AL2_x86_64/generic-flavor/src/src/relay/pass/do_replication.cc:66: Replication is not applied to group convolution. Kernel and data input channels do not match[3, 3, 4, 128] [
@Tagussan
Tagussan / pdf_remove_margin.sh
Created September 26, 2021 06:55
#!/bin/bash
tmpdir=$(mktemp -d)
gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -o $tmpdir/tmp.pdf $1
pdfcrop $tmpdir/tmp.pdf $1
@Tagussan
Tagussan / 9-43.v
Created June 22, 2014 14:17
Ltac split_all := try _split_all
with _split_all :=
match goal with
| |- (_ /\ _) => try (repeat split) || split_all
end.
(* 以下は動作確認用 *)
Lemma bar :
forall P Q R S : Prop,
@Tagussan
Tagussan / 9-42.v
Created June 15, 2014 11:03
Require Import Arith.
Require Import Omega.
Require Import Recdef.
Function log(n:nat) {wf lt n} :=
if le_lt_dec n 1 then
0
else
S (log (Div2.div2 n)).
Proof.
@Tagussan
Tagussan / 9-41.v
Created June 15, 2014 11:02
Ltac split_all := try (repeat split)||split_all.
(* 以下は動作確認用 *)
Lemma bar :
forall P Q R S : Prop,
R -> Q -> P -> S -> (P /\ R) /\ (S /\ Q).
Proof.
intros P Q R S H0 H1 H2 H3.
split_all.
@Tagussan
Tagussan / 8-39.v
Created June 8, 2014 12:03
(* bindの記法を予約 *)
Reserved Notation "x >>= y" (at level 110, left associativity).
(* モナド *)
Class Monad(M:Type -> Type) := {
bind {A B} : M A -> (A -> M B) -> M B
where "x >>= f" := (bind x f);
ret {A} : A -> M A;
left_id : forall {A:Type} {B:Type} (x:A) (f:A -> M B), ((ret x) >>= f) = (f x);
right_id : forall {A:Type} (x:M A), (x >>= ret) = x;