Add screenshots.

README.md

@@ -11,6 +11,9 @@ This is an inference engine for the [language model of RWKV](https://github.com/
 - Int8 quantization.
 - Very fast.
 
+![chat](screenshots/chat.gif)
+![batch](screenshots/batch.gif)
+
 ## Compile and Run
 1. [Install Rust](https://rustup.rs/).
 2. Run `cargo run --release --example gen` to generate 100 tokens and measure the time cost.
@@ -49,5 +52,16 @@ You may download the official RWKV World series models from [HuggingFace](https:
 
 An already-converted 0.4B model can be found under [`assets/models`](assets/models/RWKV-4-World-0.4B-v1-20230529-ctx4096.st).
 
+## Troubleshoot
+- "thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: HeaderTooLarge'"
+
+  Your model is broken, mainly because you cloned the repo but did not set up git-lfs.Please download the model manually and overwrite that one in `assets/models`.
+
+- "thread 'main' panicked at 'wgpu error: Validation Error"
+
+  It's most likely that you are using the D3D backend.
+  Please use Vulkan backend instead.
+
+
 ## Credits
 - Tokenizer is implemented by [@koute](https://github.com/koute/rwkv_tokenizer).

examples/batch.rs

@@ -1,12 +1,14 @@
 use anyhow::Result;
 use clap::Parser;
+#[cfg(not(debug_assertions))]
 use crossterm::terminal::{
     disable_raw_mode, enable_raw_mode, EnterAlternateScreen, LeaveAlternateScreen,
 };
 #[cfg(not(debug_assertions))]
 use dialoguer::{theme::ColorfulTheme, Select};
 use itertools::Itertools;
 use memmap2::Mmap;
+#[cfg(not(debug_assertions))]
 use ratatui::{
     prelude::{Constraint, CrosstermBackend, Direction, Layout},
     style::{Color, Modifier, Style, Stylize},
@@ -16,7 +18,7 @@ use ratatui::{
 };
 use std::{
     fs::File,
-    io::{BufReader, Read, Stdout},
+    io::{BufReader, Read},
     path::PathBuf,
     str::FromStr,
 };
@@ -96,21 +98,24 @@ fn load_model(context: &Context, model: PathBuf, quant: Option<u64>) -> Result<M
     Ok(model)
 }
 
-fn setup_terminal() -> Result<Terminal<CrosstermBackend<Stdout>>> {
+#[cfg(not(debug_assertions))]
+fn setup_terminal() -> Result<Terminal<CrosstermBackend<std::io::Stdout>>> {
     let mut stdout = std::io::stdout();
     enable_raw_mode()?;
     crossterm::execute!(stdout, EnterAlternateScreen)?;
     Ok(Terminal::new(CrosstermBackend::new(stdout))?)
 }
 
-fn restore_terminal(terminal: &mut Terminal<CrosstermBackend<Stdout>>) -> Result<()> {
+#[cfg(not(debug_assertions))]
+fn restore_terminal(terminal: &mut Terminal<CrosstermBackend<std::io::Stdout>>) -> Result<()> {
     disable_raw_mode()?;
     crossterm::execute!(terminal.backend_mut(), LeaveAlternateScreen,)?;
     Ok(terminal.show_cursor()?)
 }
 
 async fn run(cli: Cli) -> Result<()> {
     let context = create_context().await?;
+    #[cfg(not(debug_assertions))]
     let mut terminal = setup_terminal()?;
 
     let tokenizer = load_tokenizer()?;
@@ -146,6 +151,7 @@ async fn run(cli: Cli) -> Result<()> {
         .repeat((cli.batch + prompts.len() - 1) / prompts.len())[..cli.batch]
         .to_vec();
     loop {
+        #[cfg(not(debug_assertions))]
         terminal.draw(|frame| {
             let size = frame.size();
 
@@ -193,6 +199,11 @@ async fn run(cli: Cli) -> Result<()> {
             }
         })?;
 
+        #[cfg(debug_assertions)]
+        for (index, prompt) in prompts.iter().enumerate() {
+            println!("{index}: {prompt}");
+        }
+
         let logits = model.run(&mut tokens, &state)?;
         let probs = model.softmax(logits)?;
         for (index, probs) in probs.into_iter().enumerate().filter(|(_, v)| !v.is_empty()) {
@@ -212,6 +223,7 @@ async fn run(cli: Cli) -> Result<()> {
         }
     }
 
+    #[cfg(not(debug_assertions))]
     restore_terminal(&mut terminal)?;
     Ok(())
 }