You can deploy this model and start earning money today!

rswaminathan38/
llmbench-student-3b-gsm8k-ce-20260414-checkpoint-2000

Name: llmbench-student-3b-gsm8k-ce-20260414-checkpoint-2000
Availability: InStock

safetensors

llama

Views

No views yet

Student 3B CE Checkpoint 2000

This repo contains the training checkpoint at step 2000 export for the student model from the GSM8K workflow in this project.

Quick Use

Transformers:

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "rswaminathan38/llmbench-student-3b-gsm8k-ce-20260414-checkpoint-2000"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto")

vLLM:

vllm serve rswaminathan38/llmbench-student-3b-gsm8k-ce-20260414-checkpoint-2000 --dtype auto

Metrics

Best validation relaxed accuracy: 0.3583
Best checkpoint: checkpoint-2000
Full held-out test summary is not available because the source training job ended before the planned post-train evaluation step.

Training Details

Base model: meta-llama/Llama-3.2-3B
Variant: training checkpoint at step 2000
Output source: /storage/ice1/3/3/rswaminathan38/LLM_Bench/LLMOptimization/Model_Optimizations/outputs/hf_student_ce_3b_2026-04-14
num_train_epochs: 20
per_device_train_batch_size: 2
gradient_accumulation_steps: 8
learning_rate: 1e-05
warmup_ratio: 0.05
max_seq_length: 1024
alpha: 1.0
temperature: 2.0
kd_method: full
top_k: 64
hidden_match: False
hidden_match_weight: 0.05

Notes

Preserves the exact checkpoint directory from the interrupted 2026-04-14 CE run.
Prompt format in this repo is question + "\n\nLet's think step by step.\n".
Original Meta Llama license and access requirements still apply to downstream use.
Training metadata and evaluation summaries are uploaded alongside the weights when available.