您需要编写自己的Spliterator，如下所示：

public final class ChunkingInputStreamSpliterator implements Spliterator<byte[]> {

    private final InputStream is;
    private final int bufferSize;

    public ChunkingInputStreamSpliterator(InputStream is,int bufferSize) {
        this.is = is;
        this.bufferSize = bufferSize;
    }

    @Override
    public boolean tryAdvance(Consumer<? super byte[]> action) {
        byte[] bytes;
        try {
            bytes = this.is.readNBytes(this.bufferSize);
        } catch (IOException e) {
            throw new UncheckedIOException(e);
        }
        if (bytes.length == 0)
            return false;
        action.accept(bytes);
        return true;
    }

    @Override
    public Spliterator<byte[]> trySplit() {
        return null; // cannot split an InputStream
    }

    @Override
    public long estimateSize() {
        return Long.MAX_VALUE; // unknown
    }

    @Override
    public int characteristics() {
        return Spliterator.ORDERED | Spliterator.NONNULL;
    }

}

然后像这样实现您的方法：

public static Stream<byte[]> toStream(InputStream is,int bufferSize) {
    return StreamSupport.stream(new ChunkingInputStreamSpliterator(is,bufferSize),false);
}

如果您没有Java 11，那么就没有非常方便的readNBytes方法，那么您就可以自己这样做：

public boolean tryAdvance(Consumer<? super byte[]> action) {
    byte[] bytes = new byte[this.bufferSize];
    int len = 0;
    try {
        for (int read; len < bytes.length; len += read)
            if ((read = this.is.read(bytes,len,bytes.length - len)) <= 0)
                break;
    } catch (IOException e) {
        throw new UncheckedIOException(e);
    }
    if (len == 0)
        return false;
    if (len < bytes.length)
        bytes = Arrays.copyOfRange(bytes,len);
    action.accept(bytes);
    return true;
}