ChatON+MetaHpp: Had forgotten to conv reverse-prompt

Also has dump was using get_value calls with fallback to default,
so it wasnt identifying the missed field.

Have fixed both of those. Also reconverted meta json file.

Misc: interesting avesham and aattam
This commit is contained in:
HanishKVC 2024-05-12 15:59:37 +05:30
parent 4232ec1fb9
commit f94fed92d3
3 changed files with 29 additions and 5 deletions

View file

@ -781,6 +781,11 @@ inline bool _chaton_meta_dump(std::string &tmpl) {
std::string assistantPrefix = gCT.get_value<std::string>(tmpl, { K_ASSISTANT, K_PREFIX }); std::string assistantPrefix = gCT.get_value<std::string>(tmpl, { K_ASSISTANT, K_PREFIX });
std::string assistantSuffix = gCT.get_value<std::string>(tmpl, { K_ASSISTANT, K_SUFFIX }); std::string assistantSuffix = gCT.get_value<std::string>(tmpl, { K_ASSISTANT, K_SUFFIX });
std::string assistantEnd = gCT.get_value<std::string>(tmpl, { K_ASSISTANT, K_END }); std::string assistantEnd = gCT.get_value<std::string>(tmpl, { K_ASSISTANT, K_END });
std::string reversePrompt = gCT.get_value<std::string>(tmpl, { K_REVERSE_PROMPT });
bool systemHasSuffix = gCT.get_value<bool>(tmpl, { K_SYSTEMUSER_SYSTEM_HAS_SUFFIX });
bool systemHasEnd = gCT.get_value<bool>(tmpl, { K_SYSTEMUSER_SYSTEM_HAS_END });
bool userHasBegin = gCT.get_value<bool>(tmpl, { K_SYSTEMUSER_1ST_USER_HAS_BEGIN });
bool userHasPrefix = gCT.get_value<bool>(tmpl, { K_SYSTEMUSER_1ST_USER_HAS_PREFIX });
LOGXLN("INFO:%s:%s:%s", __func__, "global->begin", globalBegin.c_str()); LOGXLN("INFO:%s:%s:%s", __func__, "global->begin", globalBegin.c_str());
LOGXLN("INFO:%s:%s:%s", __func__, "global->end", globalEnd.c_str()); LOGXLN("INFO:%s:%s:%s", __func__, "global->end", globalEnd.c_str());
@ -796,11 +801,11 @@ inline bool _chaton_meta_dump(std::string &tmpl) {
LOGXLN("INFO:%s:%s:%s", __func__, "assistant->prefix", assistantPrefix.c_str()); LOGXLN("INFO:%s:%s:%s", __func__, "assistant->prefix", assistantPrefix.c_str());
LOGXLN("INFO:%s:%s:%s", __func__, "assistant->suffix", assistantSuffix.c_str()); LOGXLN("INFO:%s:%s:%s", __func__, "assistant->suffix", assistantSuffix.c_str());
LOGXLN("INFO:%s:%s:%s", __func__, "assistant->end", assistantEnd.c_str()); LOGXLN("INFO:%s:%s:%s", __func__, "assistant->end", assistantEnd.c_str());
LOGXLN("INFO:%s:%s:%s", __func__, K_REVERSE_PROMPT, chaton_tmpl_kv(tmpl, K_REVERSE_PROMPT).c_str()); LOGXLN("INFO:%s:%s:%s", __func__, K_REVERSE_PROMPT, reversePrompt.c_str());
LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_SYSTEM_HAS_SUFFIX, chaton_tmpl_kv_bool(tmpl, K_SYSTEMUSER_SYSTEM_HAS_SUFFIX)); LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_SYSTEM_HAS_SUFFIX, systemHasSuffix);
LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_SYSTEM_HAS_END, chaton_tmpl_kv_bool(tmpl, K_SYSTEMUSER_SYSTEM_HAS_END)); LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_SYSTEM_HAS_END, systemHasEnd);
LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_1ST_USER_HAS_BEGIN, chaton_tmpl_kv_bool(tmpl, K_SYSTEMUSER_1ST_USER_HAS_BEGIN)); LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_1ST_USER_HAS_BEGIN, userHasBegin);
LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_1ST_USER_HAS_PREFIX, chaton_tmpl_kv_bool(tmpl, K_SYSTEMUSER_1ST_USER_HAS_PREFIX)); LOGXLN("INFO:%s:%s:%d", __func__, K_SYSTEMUSER_1ST_USER_HAS_PREFIX, userHasPrefix);
if (!userEnd.empty()) { if (!userEnd.empty()) {
LOG_TEELN("WARN:%s:User->End seems to be set to [%s], do cross check if this is proper and needed", __func__, userEnd.c_str()); LOG_TEELN("WARN:%s:User->End seems to be set to [%s], do cross check if this is proper and needed", __func__, userEnd.c_str());

View file

@ -19,6 +19,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "" }, { "assistant-prefix", "" },
{ "assistant-suffix", "" }, { "assistant-suffix", "" },
{ "assistant-end", "</s>" }, { "assistant-end", "</s>" },
{ "reverse-prompt", "</s>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", false }, { "systemuser-system-has-end", false },
{ "systemuser-1st-user-has-begin", false }, { "systemuser-1st-user-has-begin", false },
@ -39,6 +40,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "<|start_header_id|>assistant<|end_header_id|>\n" }, { "assistant-prefix", "<|start_header_id|>assistant<|end_header_id|>\n" },
{ "assistant-suffix", "<|eot_id|>\n\n" }, { "assistant-suffix", "<|eot_id|>\n\n" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<|eot_id|>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -59,6 +61,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "<|im_start|>assistant\n" }, { "assistant-prefix", "<|im_start|>assistant\n" },
{ "assistant-suffix", "<|im_end|>\n" }, { "assistant-suffix", "<|im_end|>\n" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<|im_start|>user\n" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -79,6 +82,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "<|assistant|>\n" }, { "assistant-prefix", "<|assistant|>\n" },
{ "assistant-suffix", "" }, { "assistant-suffix", "" },
{ "assistant-end", "</s>\n" }, { "assistant-end", "</s>\n" },
{ "reverse-prompt", "</s>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -99,6 +103,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "<start_of_turn>model\n" }, { "assistant-prefix", "<start_of_turn>model\n" },
{ "assistant-suffix", "<end_of_turn>\n" }, { "assistant-suffix", "<end_of_turn>\n" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<eos>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", false }, { "systemuser-system-has-end", false },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -119,6 +124,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "### Response:\n" }, { "assistant-prefix", "### Response:\n" },
{ "assistant-suffix", "\n<|EOT|>\n" }, { "assistant-suffix", "\n<|EOT|>\n" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<|EOT|>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", false }, { "systemuser-system-has-end", false },
{ "systemuser-1st-user-has-begin", false }, { "systemuser-1st-user-has-begin", false },
@ -139,6 +145,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "Assistant: " }, { "assistant-prefix", "Assistant: " },
{ "assistant-suffix", " <end▁of▁sentence>\n" }, { "assistant-suffix", " <end▁of▁sentence>\n" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<end▁of▁sentence>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", false }, { "systemuser-system-has-end", false },
{ "systemuser-1st-user-has-begin", false }, { "systemuser-1st-user-has-begin", false },
@ -159,6 +166,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "assistant\n" }, { "assistant-prefix", "assistant\n" },
{ "assistant-suffix", "" }, { "assistant-suffix", "" },
{ "assistant-end", "</s>\n" }, { "assistant-end", "</s>\n" },
{ "reverse-prompt", "</s>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -179,6 +187,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "" }, { "assistant-prefix", "" },
{ "assistant-suffix", "" }, { "assistant-suffix", "" },
{ "assistant-end", "</s> \n" }, { "assistant-end", "</s> \n" },
{ "reverse-prompt", "</s>" },
{ "systemuser-system-has-suffix", false }, { "systemuser-system-has-suffix", false },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", false }, { "systemuser-1st-user-has-begin", false },
@ -199,6 +208,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "<|assistant|>\n" }, { "assistant-prefix", "<|assistant|>\n" },
{ "assistant-suffix", "<|end|>\n" }, { "assistant-suffix", "<|end|>\n" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<|end|>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -219,6 +229,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>" }, { "assistant-prefix", "<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>" },
{ "assistant-suffix", "<|END_OF_TURN_TOKEN|>" }, { "assistant-suffix", "<|END_OF_TURN_TOKEN|>" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<|END_OF_TURN_TOKEN|>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -239,6 +250,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "" }, { "assistant-prefix", "" },
{ "assistant-suffix", "</s>" }, { "assistant-suffix", "</s>" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "</s>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", false }, { "systemuser-1st-user-has-begin", false },
@ -259,6 +271,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "GPT4 Correct Assistant: " }, { "assistant-prefix", "GPT4 Correct Assistant: " },
{ "assistant-suffix", "<|end_of_turn|>" }, { "assistant-suffix", "<|end_of_turn|>" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "<|end_of_turn|>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },
@ -279,6 +292,7 @@ ChatTemplates gCT = {{
{ "assistant-prefix", "ASSISTANT: " }, { "assistant-prefix", "ASSISTANT: " },
{ "assistant-suffix", "</s>\n" }, { "assistant-suffix", "</s>\n" },
{ "assistant-end", "" }, { "assistant-end", "" },
{ "reverse-prompt", "</s>" },
{ "systemuser-system-has-suffix", true }, { "systemuser-system-has-suffix", true },
{ "systemuser-system-has-end", true }, { "systemuser-system-has-end", true },
{ "systemuser-1st-user-has-begin", true }, { "systemuser-1st-user-has-begin", true },

View file

@ -9,6 +9,9 @@ import json
def kkv_str(j, tmpl, k1, k2, comma): def kkv_str(j, tmpl, k1, k2, comma):
print("\t\t{{ \"{}\", \"{}\" }}{}".format("{}-{}".format(k1,k2), repr(j[tmpl][k1][k2])[1:-1], comma)) print("\t\t{{ \"{}\", \"{}\" }}{}".format("{}-{}".format(k1,k2), repr(j[tmpl][k1][k2])[1:-1], comma))
def kv_str(j, tmpl, k1, comma):
print("\t\t{{ \"{}\", \"{}\" }}{}".format(k1, repr(j[tmpl][k1])[1:-1], comma))
def kv_bool(j, tmpl, k1, comma): def kv_bool(j, tmpl, k1, comma):
print("\t\t{{ \"{}\", {} }}{}".format(k1, repr(j[tmpl][k1]).lower(), comma)) print("\t\t{{ \"{}\", {} }}{}".format(k1, repr(j[tmpl][k1]).lower(), comma))
@ -39,6 +42,8 @@ for tmpl in j:
kkv_str(j, tmpl, "assistant", "suffix", ",") kkv_str(j, tmpl, "assistant", "suffix", ",")
kkv_str(j, tmpl, "assistant", "end", ",") kkv_str(j, tmpl, "assistant", "end", ",")
kv_str(j, tmpl, "reverse-prompt", ",")
kv_bool(j, tmpl, "systemuser-system-has-suffix", ",") kv_bool(j, tmpl, "systemuser-system-has-suffix", ",")
kv_bool(j, tmpl, "systemuser-system-has-end", ",") kv_bool(j, tmpl, "systemuser-system-has-end", ",")
kv_bool(j, tmpl, "systemuser-1st-user-has-begin", ",") kv_bool(j, tmpl, "systemuser-1st-user-has-begin", ",")