found python version of mnist code

[Abhinav-Kumar012]

looks like this the code : https://github.com/pytorch/examples/tree/main/mnist

Rust version : https://burn.dev/books/burn/basic-workflow/index.html

check @valmikGit @Transyltooniaa

[Transyltooniaa]

lgtm! will discuss this tomorrow

[Transyltooniaa]

@Abhinav-Kumar012 the architecture is different

Aspect	PyTorch	Burn
Conv channels	1 → 32 → 64	1 → 8 → 16
Activation	After each conv	After conv2 only
Pooling	MaxPool2d (2×2)	AdaptiveAvgPool2d (8×8)
Flatten size	9216	1024
FC hidden size	128 (fixed)	Configurable
Dropout	Two, different rates	One, reused
Output	log_softmax	logits

[Abhinav-Kumar012]

Rust Result:

======================== Learner Summary ========================
Model:
Model {
  conv1: Conv2d {ch_in: 1, ch_out: 8, stride: [1, 1], kernel_size: [3, 3], dilation: [1, 1], groups: 1, padding: Valid, params: 80}
  conv2: Conv2d {ch_in: 8, ch_out: 16, stride: [1, 1], kernel_size: [3, 3], dilation: [1, 1], groups: 1, padding: Valid, params: 1168}
  pool: AdaptiveAvgPool2d {output_size: [8, 8]}
  dropout: Dropout {prob: 0.5}
  linear1: Linear {d_input: 1024, d_output: 512, bias: true, params: 524800}
  linear2: Linear {d_input: 512, d_output: 10, bias: true, params: 5130}
  activation: Relu
  params: 531178
}
Total Epochs: 10

Split	Metric	Min.	Epoch	Max.	Epoch
Train	Accuracy	83.232	1	97.603	10
Train	CPU Memory	7.333	1	7.526	6
Train	CPU Temperature	82.667	10	85.402	3
Train	CPU Usage	7.361	8	7.853	1
Train	FBetaScore (1) @ TopK(1) [Macro]	81.526	1	97.368	10
Train	Loss	0.075	10	0.585	1
Train	Precision @ TopK(1) [Macro]	83.807	1	97.600	10
Train	Recall @ TopK(1) [Macro]	82.980	1	97.626	10
Train	Top-K Accuracy @ TopK(5)	97.913	1	99.965	10
Valid	Accuracy	93.720	1	98.490	10
Valid	CPU Memory	7.335	1	7.530	6
Valid	CPU Temperature	84.917	9	87.783	3
Valid	CPU Usage	10.312	4	11.961	6
Valid	FBetaScore (1) @ TopK(1) [Macro]	93.047	1	98.372	10
Valid	Loss	0.046	10	0.209	1
Valid	Precision @ TopK(1) [Macro]	93.840	1	98.535	10
Valid	Recall @ TopK(1) [Macro]	93.525	1	98.525	10
Valid	Top-K Accuracy @ TopK(5)	99.810	1	100.000	6

[valmikGit]

For running the codes.

[valmikGit]

This is for running the PyTorch and Tensorflow implementations for the MNIST dataset.

[valmikGit]

This is for the PyTorch implementation of the text classification task.

[valmikGit]

We just have to check whether the entire dataset is being used for training and testing. Currently, I think, the PyTorch implementation is using 2 sample datapoints for training.

[valmikGit]

This is for the Tensorflow implementation of the text classification task.

[valmikGit]

We just have to check whether the entire dataset is being used for training and testing. Currently, I think, the Tensorflow implementation is using 2 sample datapoints for training.

[Abhinav-Kumar012]

@Transyltooniaa @valmikGit please check code before commiting and pushing
you can watch the rust code like nodemon by using

bacon run-release -- mnist_ml

this command needs to be run at rust_ml directory
replace mnist_ml when running other packages

[Abhinav-Kumar012]

you can install bacon with

cargo install --locked bacon

website : https://dystroy.org/bacon/

[Abhinav-Kumar012]

@valmikGit
this is the command for running mnist

cargo run -p mnist_ml --release

[Abhinav-Kumar012]

rust benchmark code written in python_ml/benchmark folder
install locust : pip install locust
docker build : docker build -t mnist -f ./Dockerfile.mnist_inf .
docker container up : docker run -p 9050:9050 --name rust_mnist_infer mnist:latest
locust run : locust -f ./rust_benchmark.py

please check @valmikGit @Transyltooniaa

[Abhinav-Kumar012]

Please create a folder called test_data in python_ml/benchmark folder and put some images preferably more that 20