You can deploy this model and start earning money today!

Raghav-Singhal/
tulu3sft-normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data

Name: tulu3sft-normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data
Availability: InStock

safetensors

llama

Views

No views yet

tulu3sft-normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data

This repository contains the final Hugging Face SFT model produced from the corrected normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data pretraining checkpoint.

Training

Base model: normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data
Posttraining recipe: full-model SFT on allenai/tulu-3-sft-mixture
Sequence length: 2048
Epochs: 1.0
Learning rate: 1e-5
Global batch size: 128

Notes

Export format: standard Hugging Face LlamaForCausalLM
Precision: bfloat16
Tokenizer assets are bundled
Default chat system prompt: You are a helpful AI assistant.

Raghav-Singhal/tulu3sft-normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data

tulu3sft-normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data

Training

Notes

Raghav-Singhal/
tulu3sft-normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data