github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-06-21 21:19:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove this as unnecessary.

Browse Source
This commit is contained in:
Adam Treat 2023-06-19 18:23:05 -04:00 committed by AT
parent 7c28e79644
commit 0cfe225506
3 changed files with 11 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
gpt4all-chat/chatlistmodel.cpp
View File

@ -6,7 +6,7 @@
#include <QDataStream> #include <QDataStream>
#define CHAT_FORMAT_MAGIC 0xF5D553CC #define CHAT_FORMAT_MAGIC 0xF5D553CC
#define CHAT_FORMAT_VERSION 3 #define CHAT_FORMAT_VERSION 4
ChatListModel::ChatListModel(QObject *parent) ChatListModel::ChatListModel(QObject *parent)
: QAbstractListModel(parent) : QAbstractListModel(parent)

21
gpt4all-chat/chatllm.cpp
View File

@ -91,7 +91,6 @@ ChatLLM::ChatLLM(Chat *parent, bool isServer)
: QObject{nullptr} : QObject{nullptr}
, m_promptResponseTokens(0) , m_promptResponseTokens(0)
, m_promptTokens(0) , m_promptTokens(0)
, m_responseLogits(0)
, m_isRecalc(false) , m_isRecalc(false)
, m_chat(parent) , m_chat(parent)
, m_timer(nullptr) , m_timer(nullptr)
@ -300,12 +299,9 @@ void ChatLLM::regenerateResponse()
else else
m_ctx.n_past -= m_promptResponseTokens; m_ctx.n_past -= m_promptResponseTokens;
m_ctx.n_past = std::max(0, m_ctx.n_past); m_ctx.n_past = std::max(0, m_ctx.n_past);
// FIXME: This does not seem to be needed in my testing and llama models don't to it. Remove?
m_ctx.logits.erase(m_ctx.logits.end() -= m_responseLogits, m_ctx.logits.end());
m_ctx.tokens.erase(m_ctx.tokens.end() -= m_promptResponseTokens, m_ctx.tokens.end()); m_ctx.tokens.erase(m_ctx.tokens.end() -= m_promptResponseTokens, m_ctx.tokens.end());
m_promptResponseTokens = 0; m_promptResponseTokens = 0;
m_promptTokens = 0; m_promptTokens = 0;
m_responseLogits = 0;
m_response = std::string(); m_response = std::string();
emit responseChanged(); emit responseChanged();
} }
@ -314,7 +310,6 @@ void ChatLLM::resetResponse()
{ {
m_promptTokens = 0; m_promptTokens = 0;
m_promptResponseTokens = 0; m_promptResponseTokens = 0;
m_responseLogits = 0;
m_response = std::string(); m_response = std::string();
emit responseChanged(); emit responseChanged();
} }
@ -368,7 +363,7 @@ void ChatLLM::modelNameChangeRequested(const QString &modelName)
bool ChatLLM::handlePrompt(int32_t token) bool ChatLLM::handlePrompt(int32_t token)
{ {
// m_promptResponseTokens and m_responseLogits are related to last prompt/response not // m_promptResponseTokens is related to last prompt/response not
// the entire context window which we can reset on regenerate prompt // the entire context window which we can reset on regenerate prompt
#if defined(DEBUG) #if defined(DEBUG)
qDebug() << "prompt process" << m_chat->id() << token; qDebug() << "prompt process" << m_chat->id() << token;
@ -393,7 +388,7 @@ bool ChatLLM::handleResponse(int32_t token, const std::string &response)
return false; return false;
} }
// m_promptResponseTokens and m_responseLogits are related to last prompt/response not // m_promptResponseTokens is related to last prompt/response not
// the entire context window which we can reset on regenerate prompt // the entire context window which we can reset on regenerate prompt
++m_promptResponseTokens; ++m_promptResponseTokens;
m_timer->inc(); m_timer->inc();
@ -458,7 +453,6 @@ bool ChatLLM::prompt(const QString &prompt, const QString &prompt_template, int3
fflush(stdout); fflush(stdout);
#endif #endif
m_timer->stop(); m_timer->stop();
m_responseLogits += m_ctx.logits.size() - logitsBefore;
std::string trimmed = trim_whitespace(m_response); std::string trimmed = trim_whitespace(m_response);
if (trimmed != m_response) { if (trimmed != m_response) {
m_response = trimmed; m_response = trimmed;
@ -595,7 +589,10 @@ bool ChatLLM::serialize(QDataStream &stream, int version)
stream << response(); stream << response();
stream << generatedName(); stream << generatedName();
stream << m_promptResponseTokens; stream << m_promptResponseTokens;
stream << m_responseLogits; if (version <= 3) {
int responseLogits;
stream << responseLogits;
}
stream << m_ctx.n_past; stream << m_ctx.n_past;
stream << quint64(m_ctx.logits.size()); stream << quint64(m_ctx.logits.size());
stream.writeRawData(reinterpret_cast<const char*>(m_ctx.logits.data()), m_ctx.logits.size() * sizeof(float)); stream.writeRawData(reinterpret_cast<const char*>(m_ctx.logits.data()), m_ctx.logits.size() * sizeof(float));
@ -624,7 +621,10 @@ bool ChatLLM::deserialize(QDataStream &stream, int version)
stream >> nameResponse; stream >> nameResponse;
m_nameResponse = nameResponse.toStdString(); m_nameResponse = nameResponse.toStdString();
stream >> m_promptResponseTokens; stream >> m_promptResponseTokens;
stream >> m_responseLogits; if (version <= 3) {
int responseLogits;
stream >> responseLogits;
}
stream >> m_ctx.n_past; stream >> m_ctx.n_past;
quint64 logitsSize; quint64 logitsSize;
stream >> logitsSize; stream >> logitsSize;
@ -639,7 +639,6 @@ bool ChatLLM::deserialize(QDataStream &stream, int version)
stream >> compressed; stream >> compressed;
m_state = qUncompress(compressed); m_state = qUncompress(compressed);
} else { } else {
stream >> m_state; stream >> m_state;
} }
#if defined(DEBUG) #if defined(DEBUG)

1
gpt4all-chat/chatllm.h
View File

@ -150,7 +150,6 @@ protected:
LLModelType m_modelType; LLModelType m_modelType;
std::string m_response; std::string m_response;
std::string m_nameResponse; std::string m_nameResponse;
quint32 m_responseLogits;
QString m_modelName; QString m_modelName;
Chat *m_chat; Chat *m_chat;
TokenTimer *m_timer; TokenTimer *m_timer;