huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 10:38:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add a verbose-prompt mode, similar to llama.cpp. (#489)

Browse Source
This commit is contained in:
Laurent Mazare
2023-08-17 15:26:44 +01:00
committed by GitHub
parent 5d99026fd2
commit ad7c53953b
1 changed files with 13 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
candle-examples/examples/quantized/main.rs
View File

@ -303,6 +303,10 @@ struct Args {
/// Enable tracing (generates a trace-timestamp.json file).
#[arg(long)]
tracing: bool,
/// Display the token for the specified prompt.
#[arg(long)]
verbose_prompt: bool,
}
impl Args {
@ -375,11 +379,15 @@ fn main() -> anyhow::Result<()> {
let tokenizer = args.tokenizer()?;
let prompt = args.prompt.as_ref().map_or(DEFAULT_PROMPT, |p| p.as_str());
let mut tokens = tokenizer
.encode(prompt, true)
.map_err(anyhow::Error::msg)?
.get_ids()
.to_vec();
let tokens = tokenizer.encode(prompt, true).map_err(anyhow::Error::msg)?;
if args.verbose_prompt {
for (token, id) in tokens.get_tokens().iter().zip(tokens.get_ids().iter()) {
let token = token.replace('▁', " ").replace("<0x0A>", "\n");
println!("{id:7} -> '{token}'");
}
}
let mut tokens = tokens.get_ids().to_vec();
let mut index_pos = 0;
let mut logits_processor = LogitsProcessor::new(args.seed, args.temperature);
let start_gen = std::time::Instant::now();