jaakko/agentic-studio
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

1.5B Q4_K_M: vaihdettu 3B→1.5B koska 3B ei mahdu WASM:iin (~1 GB vs ~2 GB)

Browse Source 
3B GGUF vaati ~5 GB muistia parsinnassa → SIGILL WASM:n 4 GB rajalla.
1.5B Q4_K_M on ~1 GB ja mahtuu turvallisesti selaimeen.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
jaakko
2026-04-06 16:14:41 +03:00
parent eb69893124
commit e652bf7ab6
2 changed files with 10 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
network-poc/node/src/qwen_coder.rs
View File

@@ -18,9 +18,9 @@ macro_rules! console_log {
const MODEL_05B_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-0.5B-Instruct/resolve/main/model.safetensors";
const TOKENIZER_05B_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-0.5B-Instruct/resolve/main/tokenizer.json";
// 3B GGUF Q4_K_M — kvantisoidtu, mahtuu selaimeen (~1.9 GB)
const MODEL_3B_GGUF_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct-GGUF/resolve/main/qwen2.5-coder-3b-instruct-q4_k_m.gguf";
const TOKENIZER_3B_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct/resolve/main/tokenizer.json";
// 1.5B GGUF Q4_K_M — kvantisoidtu, mahtuu selaimeen (~1 GB)
const MODEL_GGUF_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct-GGUF/resolve/main/qwen2.5-coder-1.5b-instruct-q4_k_m.gguf";
const TOKENIZER_GGUF_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct/resolve/main/tokenizer.json";
enum CoderModel {
Full(QwenModel),
@@ -207,8 +207,8 @@ async fn get_or_build_model(use_3b: bool, ws: &Rc<RefCell<WebSocket>>) -> Result
let dtype = DType::F32;
// Tokenizer
let tok_url = if use_3b { TOKENIZER_3B_URL } else { TOKENIZER_05B_URL };
let tok_key = if use_3b { "coder3b-tokenizer.json" } else { "coder05b-tokenizer.json" };
let tok_url = if use_3b { TOKENIZER_GGUF_URL } else { TOKENIZER_05B_URL };
let tok_key = if use_3b { "coder15b-tokenizer.json" } else { "coder05b-tokenizer.json" };
let tok_bytes = ensure_cached(tok_key, tok_url, ws).await?;
let tokenizer = tokenizers::Tokenizer::from_bytes(&tok_bytes[..])
.map_err(|e| format!("Tokenizer: {}", e))?;
@@ -216,8 +216,8 @@ async fn get_or_build_model(use_3b: bool, ws: &Rc<RefCell<WebSocket>>) -> Result
// Painot
let model = if use_3b {
// GGUF Q4_K_M — kvantisoidtu 3B-malli (~1.9 GB)
let gguf_bytes = ensure_cached("coder3b-q4km.gguf", MODEL_3B_GGUF_URL, ws).await?;
console_log!("[Coder] Rakennetaan kvantisoidun 3B-mallia (Q4_K_M)...");
let gguf_bytes = ensure_cached("coder15b-q4km.gguf", MODEL_GGUF_URL, ws).await?;
console_log!("[Coder] Rakennetaan kvantisoidun 1.5B-mallia (Q4_K_M)...");
let mut cursor = std::io::Cursor::new(&gguf_bytes[..]);
let content = gguf_file::Content::read(&mut cursor)
.map_err(|e| format!("GGUF parse: {}", e))?;