Skip to content

crane-serve: Preserve the anyhow context chain in TTS error responses #116

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lucasjinreal merged 2 commits into from
Aug 31, 2026
+24 −15
Merged

crane-serve: Preserve the anyhow context chain in TTS error responses #116

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f451df1
example: Format align_qwen35_vl.rs
cryptomilk Aug 31, 2026
0e14937
crane-serve: Preserve the anyhow context chain in TTS error responses
cryptomilk Aug 31, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions crane-serve/src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -253,14 +253,14 @@ fn generate_audio(
let started = std::time::Instant::now();
let result = tts
.generate_voice_clone(&req.input, &req.language, ref_audio_path, ref_text, &opts)
.map_err(|e| e.to_string());
.map_err(|e| format!("{e:#}"));
log_generate_result(&result, started.elapsed());
result
} else {
let started = std::time::Instant::now();
let result = tts
.generate_speech(&req.input, &req.language, req.voice.as_deref(), &opts)
.map_err(|e| e.to_string());
.map_err(|e| format!("{e:#}"));
log_generate_result(&result, started.elapsed());
result
}
Expand Down Expand Up @@ -351,8 +351,8 @@ fn stream_tts(
match tts.generate_speech_stream(&req.input, &req.language, req.voice.as_deref(), &opts) {
Ok(s) => s,
Err(e) => {
tracing::error!("TTS stream setup failed: {e}");
let _ = meta.send(Err(e.to_string()));
tracing::error!("TTS stream setup failed: {e:#}");
let _ = meta.send(Err(format!("{e:#}")));
return;
},
};
Expand Down Expand Up @@ -381,8 +381,8 @@ fn stream_tts(
},
Ok(None) => break,
Err(e) => {
tracing::error!("{model_name} TTS stream failed after {n_chunks} chunks: {e}");
let _ = chunks.send(Err(e.to_string()));
tracing::error!("{model_name} TTS stream failed after {n_chunks} chunks: {e:#}");
let _ = chunks.send(Err(format!("{e:#}")));
break;
},
}
Expand Down
27 changes: 18 additions & 9 deletions example/src/align_qwen35_vl.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,7 +140,10 @@ fn main() -> Result<()> {

// ===== 1. vision embedding alignment (identical pixel inputs) =====
let emb_crane = model.encode_images(&pv_t, &grid_t)?;
let emb_crane_f = emb_crane.to_dtype(DType::F32)?.flatten_all()?.to_vec1::<f32>()?;
let emb_crane_f = emb_crane
.to_dtype(DType::F32)?
.flatten_all()?
.to_vec1::<f32>()?;
let rows = emb_crane_f.len() / 2048;
let mut row_cos = Vec::with_capacity(rows);
for r in 0..rows {
Expand Down Expand Up @@ -241,7 +244,11 @@ fn main() -> Result<()> {
matches += 1;
}
}
let acc = if n > 0 { matches as f64 / n as f64 } else { 0.0 };
let acc = if n > 0 {
matches as f64 / n as f64
} else {
0.0
};

// also produce decoded text for eyeballing
let text = model
Expand All @@ -264,8 +271,10 @@ fn main() -> Result<()> {

// ===== print report =====
println!("\n========== ALIGNMENT: Crane vs PyTorch (Qwen3.5-2B-VL) ==========");
println!("device={device:?} dtype={dtype:?} prompt_len={s} gen_len(crane)={}",
generated.len());
println!(
"device={device:?} dtype={dtype:?} prompt_len={s} gen_len(crane)={}",
generated.len()
);
println!();
println!("--- vision tower (identical pixel_values) ---");
println!(" vision emb mean row cosine : {mean_row_cos:.6}");
Expand All @@ -275,8 +284,10 @@ fn main() -> Result<()> {
println!();
println!("--- prefill logits (last position) ---");
println!(" logits cosine : {logits_cos:.6}");
println!(" top-1 argmax match : {} (crane={crane_argmax} ref={ref_argmax})",
crane_argmax == ref_argmax);
println!(
" top-1 argmax match : {} (crane={crane_argmax} ref={ref_argmax})",
crane_argmax == ref_argmax
);
println!(" top-5 overlap : {top5_overlap}/5");
println!();
println!("--- greedy decode (same prompt + pixels) ---");
Expand All @@ -285,9 +296,7 @@ fn main() -> Result<()> {
println!(" crane decoded text : {text}");
println!();
println!("--- speed (same machine) ---");
println!(
" load model : {load_ms} ms (crane)"
);
println!(" load model : {load_ms} ms (crane)");
println!(
" prefill (multimodal) : {:.1} ms -> {:.1} t/s (crane) | pytorch {:.1} t/s",
prefill_ms,
Expand Down
Loading