blumia
85b206f1cd
Add an append-only streaming path that feeds each audio chunk as a new conversation turn, so nothing is ever recomputed: the KV grows by append, no prefix is re-decoded and no audio is re-encoded. Per-step cost becomes flat instead of growing with the segment (RTF 1.06 -> 0.26 on 62.6s of continuous speech), but accuracy requires ~8s chunks because the audio encoder needs its 8s attention window, so text only appears every 8s. The incremental mode stays the default. 新增追加式流式：每块音频作为一个新的对话轮追加，前缀不重算、旧音频不重 编码，单步成本不再随段落增长（连续 62.6s 语音 RTF 从 1.06 降到 0.26）。 但准确率要求 chunk≈8s，否则会被截断，所以出字粒度是 8s；增量式仍为默认。
91 lines
2.7 KiB
C++
91 lines
2.7 KiB
C++
#pragma once
|
|
|
|
#include <cctype>
|
|
#include <cstddef>
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
#include "iasr.h"
|
|
|
|
// 可脚本化的假推理后端：让流式策略与实时管线可以脱离真模型测试。
|
|
//
|
|
// dropLastTokens 用"空格分词"模拟词表：丢掉最后 N 个词，并保留原有格式
|
|
// （这点很重要 —— 续写开头那个空格是有意义的）。
|
|
class FakeAsr : public lc::IAsr {
|
|
public:
|
|
// 每次 streamAppend 依次返回这里的文本；用完后返回空串
|
|
std::vector<std::string> scripted;
|
|
|
|
// 记录（供断言用）
|
|
std::vector<std::string> prefixesSeen;
|
|
std::vector<std::string> turnsSeen;
|
|
std::vector<size_t> samplesSeen;
|
|
int resets = 0;
|
|
|
|
int sampleRate() const override { return 16000; }
|
|
|
|
lc::AsrResult transcribeWithPrefix(const float *, size_t,
|
|
const std::string &) override
|
|
{
|
|
lc::AsrResult r;
|
|
r.ok = true;
|
|
if (call_ < scripted.size()) {
|
|
r.raw = scripted[call_];
|
|
r.transcript = lc::parseAsrOutput(r.raw);
|
|
}
|
|
++call_;
|
|
return r;
|
|
}
|
|
|
|
lc::AsrStreamStep streamAppend(const float *, size_t nSamples,
|
|
const std::string & prefix, int) override
|
|
{
|
|
prefixesSeen.push_back(prefix);
|
|
samplesSeen.push_back(nSamples);
|
|
|
|
lc::AsrStreamStep s;
|
|
s.ok = true;
|
|
if (call_ < scripted.size()) s.newText = scripted[call_];
|
|
++call_;
|
|
s.newTokens = 1;
|
|
s.encodeSeconds = 0.001;
|
|
s.inferSeconds = 0.001;
|
|
return s;
|
|
}
|
|
|
|
lc::AsrStreamStep streamAppendTurn(const float *, size_t nSamples, int) override
|
|
{
|
|
samplesSeen.push_back(nSamples);
|
|
|
|
lc::AsrStreamStep s;
|
|
s.ok = true;
|
|
if (call_ < scripted.size()) s.newText = scripted[call_];
|
|
++call_;
|
|
turnsSeen.push_back(s.newText);
|
|
s.newTokens = 1;
|
|
s.inferSeconds = 0.001;
|
|
return s;
|
|
}
|
|
|
|
void streamReset() override { ++resets; }
|
|
|
|
std::string dropLastTokens(const std::string & text, int dropLast) const override
|
|
{
|
|
if (dropLast <= 0 || text.empty()) return text;
|
|
|
|
const auto isSpace = [](char c) { return std::isspace((unsigned char) c) != 0; };
|
|
|
|
size_t pos = text.size();
|
|
for (int i = 0; i < dropLast; ++i) {
|
|
while (pos > 0 && isSpace(text[pos - 1])) --pos; // 跳过尾部空白
|
|
while (pos > 0 && !isSpace(text[pos - 1])) --pos; // 跳过一个词
|
|
if (pos == 0) return {};
|
|
}
|
|
while (pos > 0 && isSpace(text[pos - 1])) --pos;
|
|
return text.substr(0, pos);
|
|
}
|
|
|
|
private:
|
|
size_t call_ = 0;
|
|
};
|