CoolFace
Modelpublic

the-cramer-project/cpt-adapters-t1

sourceHugging Faceotherupdated 4mo agoView on Hugging Face
0likes
Model Card

CPT adapters (run_id 1)

This repo bundles 27 LoRA adapters produced by the CPT pipeline. Each adapter lives in its own subfolder; load it by passing subfolder= to PEFT.

Layout

<base_model>/<variant>/top<K>/
    adapter_config.json
    adapter_model.safetensors
    metadata.json     # per-adapter hyperparameters + losses
    tokenizer.json    # (optional, may be omitted to save space)

Loading an adapter

python
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_id   = "meta-llama/Llama-3.1-8B"     # see metadata.json for the right base
subfolder = "Llama-3.1-8B/FT-KY/top1"

base  = AutoModelForCausalLM.from_pretrained(base_id, torch_dtype="bfloat16", device_map="auto")
model = PeftModel.from_pretrained(base, "the-cramer-project/cpt-adapters-t1", subfolder=subfolder)
tok   = AutoTokenizer.from_pretrained(base_id)

Adapter index

SubfolderBase modelLanguageLoRA rLRFinal eval loss (nats)
Llama-3.1-8B/FT-KY/top1meta-llama/Llama-3.1-8BKyrgyz2565e-051.0290
Llama-3.1-8B/FT-KY/top2meta-llama/Llama-3.1-8BKyrgyz1280.00011.0118
Llama-3.1-8B/FT-KY/top3meta-llama/Llama-3.1-8BKyrgyz1285e-051.0090
Llama-3.1-8B/FT-KZ/top1meta-llama/Llama-3.1-8BKazakh2565e-051.0093
Llama-3.1-8B/FT-KZ/top2meta-llama/Llama-3.1-8BKazakh1280.00010.9951
Llama-3.1-8B/FT-KZ/top3meta-llama/Llama-3.1-8BKazakh1285e-050.9937
Llama-3.1-8B/FT-PL/top1meta-llama/Llama-3.1-8BPolish645e-051.6365
Llama-3.1-8B/FT-PL/top2meta-llama/Llama-3.1-8BPolish325e-051.6234
Llama-3.1-8B/FT-PL/top3meta-llama/Llama-3.1-8BPolish320.00011.6322
Qwen3-8B-Base/FT-KY/top1Qwen/Qwen3-8B-BaseKyrgyz2560.00011.0866
Qwen3-8B-Base/FT-KY/top2Qwen/Qwen3-8B-BaseKyrgyz1280.00021.5180
Qwen3-8B-Base/FT-KY/top3Qwen/Qwen3-8B-BaseKyrgyz2565e-051.0851
Qwen3-8B-Base/FT-KZ/top1Qwen/Qwen3-8B-BaseKazakh2560.00011.0292
Qwen3-8B-Base/FT-KZ/top2Qwen/Qwen3-8B-BaseKazakh2565e-051.0318
Qwen3-8B-Base/FT-KZ/top3Qwen/Qwen3-8B-BaseKazakh1280.00021.0245
Qwen3-8B-Base/FT-PL/top1Qwen/Qwen3-8B-BasePolish2565e-051.7490
Qwen3-8B-Base/FT-PL/top2Qwen/Qwen3-8B-BasePolish1280.00011.7275
Qwen3-8B-Base/FT-PL/top3Qwen/Qwen3-8B-BasePolish640.00021.7133
gemma-2-9b/FT-KY/top1google/gemma-2-9bKyrgyz1285e-051.2077
gemma-2-9b/FT-KY/top2google/gemma-2-9bKyrgyz640.00011.1936
gemma-2-9b/FT-KY/top3google/gemma-2-9bKyrgyz645e-051.1952
gemma-2-9b/FT-KZ/top1google/gemma-2-9bKazakh1285e-051.3012
gemma-2-9b/FT-KZ/top2google/gemma-2-9bKazakh640.00011.2847
gemma-2-9b/FT-KZ/top3google/gemma-2-9bKazakh645e-051.2867
gemma-2-9b/FT-PL/top1google/gemma-2-9bPolish325e-051.9314
gemma-2-9b/FT-PL/top2google/gemma-2-9bPolish320.00011.9516
gemma-2-9b/FT-PL/top3google/gemma-2-9bPolish645e-051.9781

For machine-readable details see `manifest.json`.