nazdef/20260713_resume-gpt2medium-step22000-d1800-step23100

GPT2Medium EN/IT NanoChat — 22k decay-only checkpoint step_23100

This is an ordinary, non-official checkpoint release from a decay-only continual-pretraining branch started from step_22000. It is a candidate for comparison in the future 1gpu-llm-medium-v2 selection; it is not the definitive v2 release.

Released checkpoint

• checkpoint: step_23100.pt
• branch: 20260713_resume-gpt2medium-gpt2preln-k20-wsddecayonly-cpt14700-step22000-lr5e5-final1e5-webwiki-d1800
• parent checkpoint: step_22000.pt
• decay schedule: 1800 steps, final target step_23800
• languages: English + Italian
• context window: 2500 tokens
• architecture: GPT-2-style decoder with pre-layernorm blocks
• architecture identifiers: architecture: gpt2, block_type: gpt2_prelayernorm
• parameter count: approximately 337.7M native training parameters; approximately 337.6M in the Transformers export
• hardware: single RTX 4060 Ti 16GB

Selection and position

step_23100 is the scalar winner of the branch, reached after 1100 of the 1800 decay steps:

• parent step_22000: val_loss_mixed = 4.5058
• winner step_23100: val_loss_mixed = 4.4675
• improvement: -0.0383
• endpoint step_23800: val_loss_mixed = 4.4918

The endpoint must therefore not be promoted automatically. The current global medium scalar champion remains the no-decay-branch CPT checkpoint step_34000 at 4.4401. This checkpoint is a branch winner and comparison candidate, not an official family release.

Training data

The model was trained on the bilingual EN/IT web + wiki corpus:

• English FineWeb-HQ (epfml/FineWeb-HQ)
• Italian FineWeb2-HQ (epfml/FineWeb2-HQ)
• English and Italian Wiki40B (google/wiki40b)
• local dataset: 202605141153_fineweb50_wiki50_50en_50it_score100_2500context_5Btokens_tok_20260515_en50it50_webwiki_stratified_500M

Quick start

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo_id = "nazdef/20260713_resume-gpt2medium-step22000-d1800-step23100"
tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(repo_id)

prompt = "La capitale d'Italia è"
prompt_ids = tokenizer(prompt, return_tensors="pt", add_special_tokens=False)
bos = torch.tensor([[tokenizer.bos_token_id]], dtype=prompt_ids["input_ids"].dtype)
input_ids = torch.cat([bos, prompt_ids["input_ids"]], dim=1)
attention_mask = torch.ones_like(input_ids)
outputs = model.generate(
    input_ids=input_ids,
    attention_mask=attention_mask,
    do_sample=True,
    max_new_tokens=64,
    temperature=0.8,
    top_k=50,
    top_p=0.95,
    repetition_penalty=1.1,
    eos_token_id=tokenizer.eos_token_id,
    pad_token_id=tokenizer.pad_token_id,
)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

This is a base pretraining checkpoint, not an instruction-tuned chat model.

License

This release uses CC-BY-SA-4.0 as the practical downstream posture for the mixed training corpus. The corpus combines FineWeb-HQ/FineWeb2-HQ web data and Wiki40B slices, whose upstream terms and attribution/share-alike obligations may apply to downstream use and redistribution. Users are responsible for checking that their intended use and derivative packaging comply with the upstream dataset terms.