code/local_train/constants.py

from __future__ import annotations

from pathlib import Path

from local_eval.constants import (
    DEFAULT_DATA_ROOT,
    DEFAULT_MODEL_DIR,
    DEFAULT_RUNS_DIR,
    REQUIRED_FILES,
    TOKENIZER_DIR,
)

GENESIS_DIR = DEFAULT_MODEL_DIR
DEFAULT_PACK_DIR = Path("/workspace/data/sft-packs")
DEFAULT_ADAPTER_DIR = Path("/workspace/models/albedo-challenger-sft-adapter")
DEFAULT_EXPORT_DIR = Path("/workspace/models/albedo-challenger-sft")
DEFAULT_RL_ADAPTER_DIR = Path("/workspace/models/albedo-challenger-rl-adapter")
DEFAULT_RL_EXPORT_DIR = Path("/workspace/models/albedo-challenger-rl")
DEFAULT_V2_ADAPTER_DIR = Path("/workspace/models/albedo-challenger-v2-adapter")
DEFAULT_V2_EXPORT_DIR = Path("/workspace/models/albedo-challenger-v2")
DEFAULT_V4_ADAPTER_DIR = Path("/workspace/models/albedo-challenger-v4-adapter")
DEFAULT_V4_EXPORT_DIR = Path("/workspace/models/albedo-challenger-v4")
DEFAULT_CHAIN_PACK = DEFAULT_PACK_DIR / "sft-chain-chain-pack.jsonl"
DEFAULT_DPO_PACK = DEFAULT_PACK_DIR / "dpo-live-fault-v16.jsonl"
DEFAULT_DPO_ADAPTER_DIR = Path("/workspace/models/albedo-challenger-v16-dpo-adapter")
DEFAULT_DPO_EXPORT_DIR = Path("/workspace/models/albedo-challenger-v16")

SOURCES = ("mini-coder", "mini-coder-rs", "open-swe-traces", "swe-hero")
MAX_PREFIX_CHARS = 54_000
KEEP_ORIGINAL_RATIO = 0.25
NON_PYTHON_FRACTION = 0.30

METADATA_FILES = REQUIRED_FILES
ALLOWED_EXPORT_EXTRAS = (
    "model.safetensors.index.json",
    ".gitattributes",
    "LICENSE",
    "README.md",
)

# HF / trainer leftovers that fail the live file manifest.
STRIP_EXPORT_NAMES = (
    "merges.txt",
    "vocab.json",
    "configuration.json",
    "special_tokens_map.json",
    "added_tokens.json",
    "chat_template.json",
    "training_args.bin",
    "trainer_state.json",
    "optimizer.pt",
    "scheduler.pt",
    "rng_state.pth",
    "adapter_config.json",
    "adapter_model.safetensors",
    "adapter_model.bin",
)

LORA_RANK = 64
LORA_ALPHA = 128
DEFAULT_MAX_SEQ_LEN = 16384
SMOKE_MAX_SEQ_LEN = 4096