code/Abforge_Training/scripts/cast_sft_bf16.py

#!/usr/bin/env python
"""Cast an fp32 HF checkpoint to bf16, one safetensors shard at a time (low memory).

Reads SRC (read-only), writes a bf16 copy to DST. Copies config (torch_dtype patched
to bfloat16), tokenizer, generation_config, and the weight index (total_size updated).
Originals are never modified.
"""
import json, os, shutil, sys
import torch
from safetensors.torch import load_file, save_file

SRC, DST = sys.argv[1], sys.argv[2]
os.makedirs(DST, exist_ok=True)

idx_path = os.path.join(SRC, "model.safetensors.index.json")
with open(idx_path) as f:
    index = json.load(f)
shards = sorted(set(index["weight_map"].values()))

total = 0
for shard in shards:
    sd = load_file(os.path.join(SRC, shard))
    out = {}
    for k, v in sd.items():
        out[k] = v.to(torch.bfloat16) if v.is_floating_point() else v
        total += out[k].numel() * out[k].element_size()
    save_file(out, os.path.join(DST, shard), metadata={"format": "pt"})
    print(f"  cast {shard}: {len(out)} tensors")
    del sd, out

# index with updated total_size
index["metadata"] = index.get("metadata", {})
index["metadata"]["total_size"] = total
with open(os.path.join(DST, "model.safetensors.index.json"), "w") as f:
    json.dump(index, f, indent=2)

# config.json with torch_dtype patched
with open(os.path.join(SRC, "config.json")) as f:
    cfg = json.load(f)
cfg["torch_dtype"] = "bfloat16"
with open(os.path.join(DST, "config.json"), "w") as f:
    json.dump(cfg, f, indent=2)

# copy aux files verbatim
for fn in ("generation_config.json", "tokenizer.json", "tokenizer_config.json",
           "vocab.json", "merges.txt", "special_tokens_map.json",
           "added_tokens.json", "chat_template.jinja", "tokenizer.model"):
    s = os.path.join(SRC, fn)
    if os.path.isfile(s):
        shutil.copy2(s, os.path.join(DST, fn))

print(f"done: {DST}  (~{total/1e9:.1f} GB bf16)")