syadegari

root@computeinstance-e00ccn3b72xn1s8y6y:/workspace# mkdir -p /workspace/syad && cd /workspace/syad
git clone --depth=1 --branch main --single-branch https://github.com/HazyResearch/ThunderKittens ThunderKittens 2>/dev/null || true

# --- versions ---
. /etc/os-release 2>/dev/null; UBU="${NAME:-Linux} ${VERSION_ID:-$(uname -r)}"
CUDA=$(nvcc --version 2>/dev/null | sed -n 's/.*release \([0-9.]\+\).*/\1/p' | head -1 || echo "")
TORCH=$(python -c "import torch; print(torch.__version__)" 2>/dev/null || echo "not found")
TRT=$(python -c "import tensorrt as trt; print(trt.__version__)" 2>/dev/null || echo "not found")
GPU=$(nvidia-smi --query-gpu=name --format=csv,noheader 2>/dev/null | paste -sd ' / ' - || echo "not found")
HASH=$(git -C ThunderKittens rev-parse --short=12 HEAD 2>/dev/null || echo "unknown")

syadegari

syadegari

syadegari

syadegari

syadegari

syadegari

syadegari

import torch
import torch.nn as nn
import torch.nn.functional as F

class Encoder(nn.Module):
    def __init__(self, nonlin_fn=F.relu, dropout=0.2):
        super(Encoder, self).__init__()
        self.conv1 = nn.Conv2d(1, 32, 2)
        self.conv2 = nn.Conv2d(32, 16, 2)
        self.conv3 = nn.Conv2d(16, 4, 2)

syadegari

syadegari

class model(nn.Module):
    '''
    Convolutional part
    conv -> relu -> pool -> dropout
    FC part
    fc -> relu'''
    def __init__(self, inp, out, conv_layers, fc_layers,
                 kernel_size=3, pool_size=2, p=0.2):
        super(model, self).__init__()