Skip to content
Other

ResNet-50 ImageNet training trajectory

103 checkpoints trained from scratch

2 min readupdated

A collection of 103 sequential saves (checkpoints) of the classic ResNet-50 image recognition model spanning its entire training run on the ImageNet-1k dataset. Rather than just giving you the final weights, this repository from LSL-datasets lets you look inside the training process itself—from random weight initialization at step zero all the way to epoch 90. Each file stores not only the PyTorch weights, but also the current step, epoch, learning rate, and validation accuracy. The release is aimed primarily at researchers, students, and ML engineers studying neural network convergence dynamics.

Key features

  • Full training trajectory: 103 weight files (~103 MB each), covering step zero before training begins, frequent early snapshots (steps 16–1000, epochs 1.5–4.5), and the end of every epoch from 1 through 90.
  • Ready-to-use metrics log: includes a trajectory.csv table tracking val_top1 and val_top5 scores for every saved step.
  • Reproducible recipe: trained using torchvision.models.resnet50 with the SGD optimizer, a batch size of 1024, and cosine learning rate decay.
  • Solid final accuracy: the final checkpoint (step 112,590) hits 76.7% top-1 and 93.3% top-5 accuracy across 50,000 ImageNet validation images.

What you can do

  • Explore loss landscapes and weight dynamics: analyze how internal layer representations evolve during early versus late stages of training.
  • Benchmark pruning and quantization methods: test which training epoch responds best to model compression or fine-tuning on custom classes.
  • Use in coursework and labs: give students a hands-on look at warmup effects and cosine learning rate schedules without burning GPU hours training from scratch.

Example

import torch, torchvision
from huggingface_hub import hf_hub_download

path = hf_hub_download("LSL-datasets/resnet50-imagenet-trajectory", "checkpoints/step0112590.pt")
ckpt = torch.load(path, map_location="cpu", weights_only=True)
model = torchvision.models.resnet50()
model.load_state_dict(ckpt["model"])
print(ckpt["epoch"], ckpt["val_top1"])  # 90.0 0.76658
Read moreDetails
Final metrics: 76.7% top-1 and 93.3% top-5 (224×224 center crop).
Hardware and setup: 2 GPUs with DistributedDataParallel, bfloat16 precision, batch size of 1024 images.
Limitations: ImageNet license — strictly for non-commercial research and educational use.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.6 GB
  • 4-bitweights 0.0 GB0.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # LSL-datasets/resnet50-imagenet-trajectory
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download LSL-datasets/resnet50-imagenet-trajectory --local-dir ~/models/resnet50-imagenet-trajectory
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of resnet50-imagenet-trajectory (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

LSL-datasets/resnet50-imagenet-trajectory
Author
LSL-datasets
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments