The Autocomplete on VsCode seems not work in the right line

[RedIce007]

Before submitting your bug report

• I believe this is a bug. I'll try to join the Continue Discord for questions
• I'm not able to find an open issue that reports the same bug
• I've seen the troubleshooting guide on the Continue Docs

Relevant environment info

- OS:win10
- Continue:v0.8.27
- IDE:VSCode 1.89.0

Description

I add new line at first bold place
but the completion(latatic) seems based on second bold place which is completely irrelevant with my input

here is the code

int hcos_get_usage(const pid_t pid, __uint64_t seconds, HCOS_USAGE *usage)
{
*****
*****
usage->status = fstat.processStatus;
usage->cpuUsage = (sstat.processTotoalTime - fstat.processTotoalTime) /
(sstat.cpuTotalTime - fstat.cpuTotalTime);
usage->memUsage = fstat.rss;
usage->fdUsage = fd_count;
usage free(fd_list_iter->fdnext);
return 0;
}

int hcos_check_fd(const pid_t pid, char *file_path, __uint64_t *max_counts)
{
******
total 85 line
******
for (struct fdinfo *fd_list_iter = fd_list_head; fd_list_iter != NULL; fd_list_iter = fd_list_head)
{
fd_list_head = fd_list_head->fdnext;
free(fd_list_iter);
}
******
return 0;
}

To reproduce

1.setup local ollama
2.config continue

{
"title": "ollama",
"model": "llama3",
"completionOptions": {},
"apiBase": "http://172.18.16.72:11434",
"provider": "ollama"
}
],
"customCommands": [
{
"name": "test",
"prompt": "{{{ input }}}\n\nWrite a comprehensive set of unit tests for the selected code. It should setup, run tests that check for correctness including important edge cases, and teardown. Ensure that the tests are complete and sophisticated. Give the tests just as chat output, don't edit any file.",
"description": "Write unit tests for highlighted code"
}
],
"tabAutocompleteModel": {
"title": "Llama3",
"provider": "ollama",
"model": "llama3",
"apiBase": "http://172.18.16.72:11434"
},
"tabAutocompleteOptions": {
"maxPromptTokens": 4000
},
"allowAnonymousTelemetry": true,
"contextProviders": [
{
"name": "code",
"params": {}
},
{
"name": "docs",
"params": {}
}

3.enable auto completion
4.write the code and wait completion

Log output

No recent errors in log console.

[sestinj]

@RedIce007 there is a chance that this happens because the model being used is Llama 3, which was not trained for tab-autocomplete. I would recommend trying starcoder2:3b, which is trained on a "Fill in the middle" format. See here for a related explanation of why this is important