"--model " + mailQuery.model.filesystemPath,
"--threads " + configuration.getThreadCount(),
"--threads-batch " + configuration.getBatchThreadCount(),
- "--mirostat 2",
- "--flash-attn",
+ "--top-k 40",
+ "--top-p 0.95",
+ "--min-p 0.1",
+ "--repeat-penalty 1.1",
+ "--dry-multiplier 0.5",
+ "--presence-penalty 0.1",
+ "--mirostat 0",
+ "--samplers 'top_k;top_p;min_p;temperature;dry;typ_p;xtc'",
"--cache-type-k q8_0",
"--cache-type-v q8_0",
"--no-display-prompt",
"--no-warmup",
- "--temp " + temperature,
+ "--flash-attn",
+ "--temp 0.6",
"--ctx-size " + mailQuery.model.contextSizeTokens,
"--batch-size 8",
"--no-conversation",
"-n -1",
- "--repeat_penalty 1.1",
"--file " + inputFile
);
+
+// Old solution that sometimes went into an infinite loop, for backup purposes:
+//
+// return join(" ",
+// "nice", "-n", Integer.toString(niceValue),
+// executablePath,
+// "--model " + mailQuery.model.filesystemPath,
+// "--threads " + configuration.getThreadCount(),
+// "--threads-batch " + configuration.getBatchThreadCount(),
+// "--mirostat 2",
+// "--flash-attn",
+// "--cache-type-k q8_0",
+// "--cache-type-v q8_0",
+// "--no-display-prompt",
+// "--no-warmup",
+// "--temp " + temperature,
+// "--ctx-size " + mailQuery.model.contextSizeTokens,
+// "--batch-size 8",
+// "--no-conversation",
+// "-n -1",
+// "--repeat_penalty 1.1",
+// "--file " + inputFile
+// );
+
}
/**