rswaminathan38/llmbench-teacher-8b-gsm8k-ce-20260410

libraryname: transformers pipelinetag: text-generation base_model: meta-llama/Meta-Llama-3.1-8B datasets:

• —gsm8k tags:
• —gsm8k
• —transformers
• —vllm
• —text-generation
• —teacher-model ---

Teacher 8B CE

This repo contains the cross-entropy fine-tune export for the teacher model from the GSM8K workflow in this project.

Quick Use

Transformers:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "rswaminathan38/llmbench-teacher-8b-gsm8k-ce-20260410"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto")

vLLM:

vllm serve rswaminathan38/llmbench-teacher-8b-gsm8k-ce-20260410 --dtype auto

Metrics

• —Test relaxed exact-match accuracy: 0.1554
• —Correct / examples: 205 / 1319
• —Avg generated tokens: 238.6634
• —Prompt style used during evaluation: cot_step_by_step

Training Details

• —Base model: meta-llama/Meta-Llama-3.1-8B
• —Variant: cross-entropy fine-tune
• —Output source: /storage/ice1/3/3/rswaminathan38/LLM_Bench/LLMOptimization/Model_Optimizations/outputs/hf_teacher_8b_gsm8k_2026-04-10
• —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

Notes

• —Teacher model fine-tuned on GSM8K with the repository's chain-of-thought prompt format.
• —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.
• —run_config.json and eval/test_summary.json are uploaded alongside the weights when available.