1pp-1.7b-asst-base

One Persona Pretraining (1PP) experiment model: 1.66B parameters, pretraining condition rewritten conversations, assistant-turn loss.

Part of a 3 × 3 study: three sizes (0.5B, 1B, 1.7B) × three pretraining conditions on the same 47.8M source documents in the same order (original documents; rewritten conversations with loss on assistant turns; rewritten conversations with loss on user and assistant turns). Every run saw the identical batch sequence, so the conditions differ only in the document text and the loss mask. Models are grouped in the 1pp collection.

Architecture

Llama-style decoder, 24 layers, hidden 2,048, FFN 8,192 (SwiGLU), attention heads / KV heads 16 / 4 (head dim 128), RMSNorm, RoPE base 10,000, untied embeddings, no biases, no QK-norm, sequence length 4,096. Tokenizer: SmolLM2 vocabulary (49,152) plus <|pad|>; <|endoftext|> is the end-of-document token.

Pretraining

Data: the 1PP conversations rewritten from those documents; loss only on assistant turns (no loss on user turns or <|endoftext|>). One pass over 47.8M documents (66.2B tokens of original documents; 63.0B tokens as conversations), 31,777 steps at global batch 512 × 4,096 tokens, cross-document attention masking, best-fit packing with step-aligned document assignment. Optimizer: Muon (shape scaling, matrix LR 0.005) with Adam for embeddings and norms, warmup 2,000 steps, constant, linear decay over the last 10% to 1/100, weight decay 0.1, bf16.

Validation loss (per token, 2,433 held-out documents, final checkpoint):

assistant text	user text	document text
1.451	6.730	3.178

Chat format

ChatML without a system turn (the models never saw one):

<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n{reply}<|im_end|>\n

The bundled chat_template renders exactly this. Generation stops at <|im_end|> (id 2) or <|endoftext|> (id 0); both are listed in eos_token_id. This is a base model; the conversation conditions produce chat-formatted text, the raw baseline plain text.

Verification

The HF weights were checked against the Megatron checkpoint by recomputing validation losses with this model:

set	HF loss	Megatron reference	abs. diff
val50m segments [3]	1.4495	1.4509	0.0014
raw_val50m segments [8]	3.1794	3.1779	0.0015

Links

• Training logs: wandb projects 1pp-training and 1pp-sft
• Research artifact from the 1PP project (EPFL DLAB); not a general-purpose assistant.