Terminaaliin latauksen tilaindikaattori (spinner + vaihe)

Mallin latauksen aikana terminaalissa näkyy animoitu spinner
ja nykyinen vaihe: WASM → tokenizer → malli (%) → rakennus → valmis.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

docker-errors.log

@@ -0,0 +1,475 @@
+[INFO]: Checking for the Wasm target...
+info: downloading component rust-std
+[INFO]: Compiling to Wasm...
+   Compiling node v0.1.0 (/app/node)
+warning: unused imports: `DType`, `Device`, and `Tensor`
+ --> node/src/smollm.rs:1:19
+  |
+1 | use candle_core::{Device, Tensor, DType};
+  |                   ^^^^^^  ^^^^^^  ^^^^^
+  |
+  = note: `#[warn(unused_imports)]` (part of `#[warn(unused)]`) on by default
+
+warning: unused import: `candle_nn::VarBuilder`
+ --> node/src/smollm.rs:2:5
+  |
+2 | use candle_nn::VarBuilder;
+  |     ^^^^^^^^^^^^^^^^^^^^^
+
+warning: unused imports: `Cache`, `LlamaConfig`, `LlamaEosToks`, and `Llama`
+ --> node/src/smollm.rs:3:42
+  |
+3 | use candle_transformers::models::llama::{Llama, LlamaConfig, LlamaEosToks, Cache};
+  |                                          ^^^^^  ^^^^^^^^^^^  ^^^^^^^^^^^^  ^^^^^
+
+warning: unused imports: `DType`, `Device`, and `Tensor`
+ --> node/src/phi3.rs:1:19
+  |
+1 | use candle_core::{Device, Tensor, DType};
+  |                   ^^^^^^  ^^^^^^  ^^^^^
+
+warning: unused import: `candle_nn::VarBuilder`
+ --> node/src/phi3.rs:2:5
+  |
+2 | use candle_nn::VarBuilder;
+  |     ^^^^^^^^^^^^^^^^^^^^^
+
+warning: unused imports: `Config as Phi3Config` and `Model as Phi3Model`
+ --> node/src/phi3.rs:3:41
+  |
+3 | use candle_transformers::models::phi3::{Config as Phi3Config, Model as Phi3Model};
+  |                                         ^^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^
+
+warning: unused import: `wasm_bindgen::JsCast`
+ --> node/src/phi3.rs:4:5
+  |
+4 | use wasm_bindgen::JsCast;
+  |     ^^^^^^^^^^^^^^^^^^^^
+
+warning: unused import: `crate::storage`
+ --> node/src/phi3.rs:9:5
+  |
+9 | use crate::storage;
+  |     ^^^^^^^^^^^^^^
+
+warning: unused import: `Int`
+ --> node/src/burn_smollm/attention.rs:2:46
+  |
+2 | use burn::tensor::{backend::Backend, Tensor, Int};
+  |                                              ^^^
+
+warning: unused imports: `Mlp` and `RmsNorm`
+ --> node/src/burn_smollm/attention.rs:4:22
+  |
+4 | use super::modules::{RmsNorm, Mlp};
+  |                      ^^^^^^^  ^^^
+
+warning: use of deprecated struct `burn::tensor::Data`: the internal data format has changed, please use `TensorData` instead
+   --> node/src/smollm.rs:174:23
+    |
+174 |         burn::tensor::Data::new(input_ids.iter().map(|&x| x as i32).collect::<Vec<_>>(), [input_len].into()),
+    |                       ^^^^
+    |
+    = note: `#[warn(deprecated)]` on by default
+
+warning: use of deprecated struct `burn::tensor::Data`: the internal data format has changed, please use `TensorData` instead
+   --> node/src/smollm.rs:200:27
+    |
+200 |             burn::tensor::Data::new(vec![next_token as i32], [1].into()),
+    |                           ^^^^
+
+warning: use of deprecated struct `burn::tensor::Data`: the internal data format has changed, please use `TensorData` instead
+ --> node/src/burn_smollm/loader.rs:1:46
+  |
+1 | use burn::tensor::{backend::Backend, Tensor, Data};
+  |                                              ^^^^
+
+warning: use of deprecated struct `burn::tensor::Data`: the internal data format has changed, please use `TensorData` instead
+  --> node/src/burn_smollm/loader.rs:17:16
+   |
+17 |     let data = Data::new(vec, shape_out_in.into());
+   |                ^^^^
+
+warning: use of deprecated struct `burn::tensor::Data`: the internal data format has changed, please use `TensorData` instead
+  --> node/src/burn_smollm/loader.rs:32:16
+   |
+32 |     let data = Data::new(vec, shape.into());
+   |                ^^^^
+
+warning: use of deprecated struct `burn::tensor::Data`: the internal data format has changed, please use `TensorData` instead
+  --> node/src/burn_smollm/loader.rs:45:16
+   |
+45 |     let data = Data::new(vec, shape.into());
+   |                ^^^^
+
+error[E0061]: this function takes 2 arguments but 1 argument was supplied
+   --> node/src/smollm.rs:124:9
+    |
+124 |         burn_wgpu::init_async::<burn_wgpu::AutoGraphicsApi>(&Default::default()).await;
+    |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^--------------------- argument #2 of type `RuntimeOptions` is missing
+    |
+note: function defined here
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/cubecl-wgpu-0.2.0/src/runtime.rs:116:14
+    |
+116 | pub async fn init_async<G: GraphicsApi>(device: &WgpuDevice, options: RuntimeOptions) {
+    |              ^^^^^^^^^^
+help: provide the argument
+    |
+124 |         burn_wgpu::init_async::<burn_wgpu::AutoGraphicsApi>(&Default::default(), /* RuntimeOptions */).await;
+    |                                                                                ++++++++++++++++++++++
+
+error[E0277]: the trait bound `TensorData: From<burn::tensor::Data<i32, 1>>` is not satisfied
+   --> node/src/smollm.rs:174:9
+    |
+173 |     let mut input_tensor = burn::tensor::Tensor::<B, 1, burn::tensor::Int>::from_data(
+    |                            ---------------------------------------------------------- required by a bound introduced by this call
+174 |         burn::tensor::Data::new(input_ids.iter().map(|&x| x as i32).collect::<Vec<_>>(), [input_len].into()),
+    |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the trait `From<burn::tensor::Data<i32, 1>>` is not implemented for `TensorData`
+    |
+    = help: the following other types implement trait `From<T>`:
+              `TensorData` implements `From<&[E]>`
+              `TensorData` implements `From<&[usize]>`
+              `TensorData` implements `From<[E; A]>`
+              `TensorData` implements `From<[[E; B]; A]>`
+              `TensorData` implements `From<[[[E; C]; B]; A]>`
+              `TensorData` implements `From<[[[[E; D]; C]; B]; A]>`
+              `TensorData` implements `From<[[[[[Elem; E]; D]; C]; B]; A]>`
+              `TensorData` implements `From<[usize; A]>`
+    = note: required for `burn::tensor::Data<i32, 1>` to implement `Into<TensorData>`
+note: required by a bound in `burn::tensor::Tensor::<B, D, K>::from_data`
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:719:12
+    |
+717 |     pub fn from_data<T>(data: T, device: &B::Device) -> Self
+    |            --------- required by a bound in this associated function
+718 |     where
+719 |         T: Into<TensorData>,
+    |            ^^^^^^^^^^^^^^^^ required by this bound in `Tensor::<B, D, K>::from_data`
+
+error[E0061]: this method takes 2 arguments but 0 arguments were supplied
+   --> node/src/smollm.rs:183:51
+    |
+183 |     let next_token_tensor = last_logits.argmax(2).flatten::<1>();
+    |                                                   ^^^^^^^^^^^^-- two arguments of type `usize` and `usize` are missing
+    |
+note: method defined here
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:292:12
+    |
+292 |     pub fn flatten<const D2: usize>(self, start_dim: usize, end_dim: usize) -> Tensor<B, D2, K> {
+    |            ^^^^^^^
+help: provide the arguments
+    |
+183 |     let next_token_tensor = last_logits.argmax(2).flatten::<1>(/* usize */, /* usize */);
+    |                                                                ++++++++++++++++++++++++
+
+error[E0277]: the trait bound `TensorData: From<burn::tensor::Data<i32, 1>>` is not satisfied
+   --> node/src/smollm.rs:200:13
+    |
+199 |         let mut input_tensor = burn::tensor::Tensor::<B, 1, burn::tensor::Int>::from_data(
+    |                                ---------------------------------------------------------- required by a bound introduced by this call
+200 |             burn::tensor::Data::new(vec![next_token as i32], [1].into()),
+    |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the trait `From<burn::tensor::Data<i32, 1>>` is not implemented for `TensorData`
+    |
+    = help: the following other types implement trait `From<T>`:
+              `TensorData` implements `From<&[E]>`
+              `TensorData` implements `From<&[usize]>`
+              `TensorData` implements `From<[E; A]>`
+              `TensorData` implements `From<[[E; B]; A]>`
+              `TensorData` implements `From<[[[E; C]; B]; A]>`
+              `TensorData` implements `From<[[[[E; D]; C]; B]; A]>`
+              `TensorData` implements `From<[[[[[Elem; E]; D]; C]; B]; A]>`
+              `TensorData` implements `From<[usize; A]>`
+    = note: required for `burn::tensor::Data<i32, 1>` to implement `Into<TensorData>`
+note: required by a bound in `burn::tensor::Tensor::<B, D, K>::from_data`
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:719:12
+    |
+717 |     pub fn from_data<T>(data: T, device: &B::Device) -> Self
+    |            --------- required by a bound in this associated function
+718 |     where
+719 |         T: Into<TensorData>,
+    |            ^^^^^^^^^^^^^^^^ required by this bound in `Tensor::<B, D, K>::from_data`
+
+error[E0061]: this method takes 2 arguments but 0 arguments were supplied
+   --> node/src/smollm.rs:207:50
+    |
+207 |         let next_token_tensor = logits.argmax(2).flatten::<1>();
+    |                                                  ^^^^^^^^^^^^-- two arguments of type `usize` and `usize` are missing
+    |
+note: method defined here
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:292:12
+    |
+292 |     pub fn flatten<const D2: usize>(self, start_dim: usize, end_dim: usize) -> Tensor<B, D2, K> {
+    |            ^^^^^^^
+help: provide the arguments
+    |
+207 |         let next_token_tensor = logits.argmax(2).flatten::<1>(/* usize */, /* usize */);
+    |                                                               ++++++++++++++++++++++++
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:58:13
+   |
+58 |         q = q.reshape([batch, seq_len, self.num_heads, self.head_dim]).swap_dims(1, 2);
+   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `3`, found `4`
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 3>`
+              found struct `burn::tensor::Tensor<_, 4>`
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:59:13
+   |
+59 |         k = k.reshape([batch, seq_len, self.num_kv_heads, self.head_dim]).swap_dims(1, 2);
+   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `3`, found `4`
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 3>`
+              found struct `burn::tensor::Tensor<_, 4>`
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:60:13
+   |
+60 |         v = v.reshape([batch, seq_len, self.num_kv_heads, self.head_dim]).swap_dims(1, 2);
+   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `3`, found `4`
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 3>`
+              found struct `burn::tensor::Tensor<_, 4>`
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:63:31
+   |
+63 |         q = self.rope.forward(q, offset);
+   |                       ------- ^ expected `4`, found `3`
+   |                       |
+   |                       arguments to this method are incorrect
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 4>`
+              found struct `burn::tensor::Tensor<_, 3>`
+note: method defined here
+  --> node/src/burn_smollm/rope.rs:35:12
+   |
+35 |     pub fn forward(&self, x: Tensor<B, 4>, offset: usize) -> Tensor<B, 4> {
+   |            ^^^^^^^        ---------------
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:63:13
+   |
+63 |         q = self.rope.forward(q, offset);
+   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `3`, found `4`
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 3>`
+              found struct `burn::tensor::Tensor<_, 4>`
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:64:31
+   |
+64 |         k = self.rope.forward(k, offset);
+   |                       ------- ^ expected `4`, found `3`
+   |                       |
+   |                       arguments to this method are incorrect
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 4>`
+              found struct `burn::tensor::Tensor<_, 3>`
+note: method defined here
+  --> node/src/burn_smollm/rope.rs:35:12
+   |
+35 |     pub fn forward(&self, x: Tensor<B, 4>, offset: usize) -> Tensor<B, 4> {
+   |            ^^^^^^^        ---------------
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:64:13
+   |
+64 |         k = self.rope.forward(k, offset);
+   |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `3`, found `4`
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 3>`
+              found struct `burn::tensor::Tensor<_, 4>`
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:68:41
+   |
+68 |             c.k = Tensor::cat(vec![c.k, k], 2);
+   |                                         ^ expected `4`, found `3`
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 4>`
+              found struct `burn::tensor::Tensor<_, 3>`
+
+error[E0308]: mismatched types
+  --> node/src/burn_smollm/attention.rs:69:41
+   |
+69 |             c.v = Tensor::cat(vec![c.v, v], 2);
+   |                                         ^ expected `4`, found `3`
+   |
+   = note: expected struct `burn::tensor::Tensor<_, 4>`
+              found struct `burn::tensor::Tensor<_, 3>`
+
+error[E0308]: `if` and `else` have incompatible types
+  --> node/src/burn_smollm/attention.rs:72:13
+   |
+67 |           let (k, v) = if let Some(mut c) = cache {
+   |  ______________________-
+68 | |             c.k = Tensor::cat(vec![c.k, k], 2);
+69 | |             c.v = Tensor::cat(vec![c.v, v], 2);
+70 | |             (c.k.clone(), c.v.clone())
+   | |             -------------------------- expected because of this
+71 | |         } else {
+72 | |             (k.clone(), v.clone())
+   | |             ^^^^^^^^^^^^^^^^^^^^^^ expected `4`, found `3`
+73 | |         };
+   | |_________- `if` and `else` have incompatible types
+   |
+   = note: expected tuple `(burn::tensor::Tensor<_, 4>, burn::tensor::Tensor<_, 4>)`
+              found tuple `(burn::tensor::Tensor<_, 3>, burn::tensor::Tensor<_, 3>)`
+
+error[E0282]: type annotations needed
+  --> node/src/burn_smollm/attention.rs:75:38
+   |
+75 |         let new_cache = KVCache { k: k.clone(), v: v.clone() };
+   |                                      ^ cannot infer type
+
+error[E0282]: type annotations needed
+  --> node/src/burn_smollm/attention.rs:75:52
+   |
+75 |         let new_cache = KVCache { k: k.clone(), v: v.clone() };
+   |                                                    ^ cannot infer type
+
+error[E0277]: the trait bound `TensorData: From<burn::tensor::Data<f32, 2>>` is not satisfied
+   --> node/src/burn_smollm/loader.rs:18:44
+    |
+ 18 |     let t_burn = Tensor::<B, 2>::from_data(data, device);
+    |                  ------------------------- ^^^^ the trait `From<burn::tensor::Data<f32, 2>>` is not implemented for `TensorData`
+    |                  |
+    |                  required by a bound introduced by this call
+    |
+    = help: the following other types implement trait `From<T>`:
+              `TensorData` implements `From<&[E]>`
+              `TensorData` implements `From<&[usize]>`
+              `TensorData` implements `From<[E; A]>`
+              `TensorData` implements `From<[[E; B]; A]>`
+              `TensorData` implements `From<[[[E; C]; B]; A]>`
+              `TensorData` implements `From<[[[[E; D]; C]; B]; A]>`
+              `TensorData` implements `From<[[[[[Elem; E]; D]; C]; B]; A]>`
+              `TensorData` implements `From<[usize; A]>`
+    = note: required for `burn::tensor::Data<f32, 2>` to implement `Into<TensorData>`
+note: required by a bound in `burn::tensor::Tensor::<B, D, K>::from_data`
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:719:12
+    |
+717 |     pub fn from_data<T>(data: T, device: &B::Device) -> Self
+    |            --------- required by a bound in this associated function
+718 |     where
+719 |         T: Into<TensorData>,
+    |            ^^^^^^^^^^^^^^^^ required by this bound in `Tensor::<B, D, K>::from_data`
+
+error[E0277]: the trait bound `TensorData: From<burn::tensor::Data<f32, 1>>` is not satisfied
+   --> node/src/burn_smollm/loader.rs:33:53
+    |
+ 33 |     Ok(Param::from_tensor(Tensor::<B, 1>::from_data(data, device)))
+    |                           ------------------------- ^^^^ the trait `From<burn::tensor::Data<f32, 1>>` is not implemented for `TensorData`
+    |                           |
+    |                           required by a bound introduced by this call
+    |
+    = help: the following other types implement trait `From<T>`:
+              `TensorData` implements `From<&[E]>`
+              `TensorData` implements `From<&[usize]>`
+              `TensorData` implements `From<[E; A]>`
+              `TensorData` implements `From<[[E; B]; A]>`
+              `TensorData` implements `From<[[[E; C]; B]; A]>`
+              `TensorData` implements `From<[[[[E; D]; C]; B]; A]>`
+              `TensorData` implements `From<[[[[[Elem; E]; D]; C]; B]; A]>`
+              `TensorData` implements `From<[usize; A]>`
+    = note: required for `burn::tensor::Data<f32, 1>` to implement `Into<TensorData>`
+note: required by a bound in `burn::tensor::Tensor::<B, D, K>::from_data`
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:719:12
+    |
+717 |     pub fn from_data<T>(data: T, device: &B::Device) -> Self
+    |            --------- required by a bound in this associated function
+718 |     where
+719 |         T: Into<TensorData>,
+    |            ^^^^^^^^^^^^^^^^ required by this bound in `Tensor::<B, D, K>::from_data`
+
+error[E0277]: the trait bound `TensorData: From<burn::tensor::Data<f32, 2>>` is not satisfied
+   --> node/src/burn_smollm/loader.rs:47:53
+    |
+ 47 |     Ok(Param::from_tensor(Tensor::<B, 2>::from_data(data, device)))
+    |                           ------------------------- ^^^^ the trait `From<burn::tensor::Data<f32, 2>>` is not implemented for `TensorData`
+    |                           |
+    |                           required by a bound introduced by this call
+    |
+    = help: the following other types implement trait `From<T>`:
+              `TensorData` implements `From<&[E]>`
+              `TensorData` implements `From<&[usize]>`
+              `TensorData` implements `From<[E; A]>`
+              `TensorData` implements `From<[[E; B]; A]>`
+              `TensorData` implements `From<[[[E; C]; B]; A]>`
+              `TensorData` implements `From<[[[[E; D]; C]; B]; A]>`
+              `TensorData` implements `From<[[[[[Elem; E]; D]; C]; B]; A]>`
+              `TensorData` implements `From<[usize; A]>`
+    = note: required for `burn::tensor::Data<f32, 2>` to implement `Into<TensorData>`
+note: required by a bound in `burn::tensor::Tensor::<B, D, K>::from_data`
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:719:12
+    |
+717 |     pub fn from_data<T>(data: T, device: &B::Device) -> Self
+    |            --------- required by a bound in this associated function
+718 |     where
+719 |         T: Into<TensorData>,
+    |            ^^^^^^^^^^^^^^^^ required by this bound in `Tensor::<B, D, K>::from_data`
+
+error[E0599]: no function or associated item named `arange` found for struct `burn::tensor::Tensor<B, 1>` in the current scope
+   --> node/src/burn_smollm/rope.rs:19:33
+    |
+ 19 |         let t = Tensor::<B, 1>::arange(0..max_seq_len as i64, device).float().unsqueeze::<2>().transpose();
+    |                                 ^^^^^^ function or associated item not found in `burn::tensor::Tensor<B, 1>`
+    |
+note: if you're trying to build a new `burn::tensor::Tensor<B, 1>` consider using one of the following associated functions:
+      burn::tensor::Tensor::<B, D, K>::new
+      burn::tensor::Tensor::<B, D, K>::from_primitive
+      burn::tensor::Tensor::<B, D, K>::empty
+      burn::tensor::Tensor::<B, D, K>::from_data
+      and 9 others
+   --> /usr/local/cargo/registry/src/index.crates.io-1949cf8c6b5b557f/burn-tensor-0.14.0/src/tensor/api/base.rs:24:10
+    |
+ 24 |   #[derive(new, Clone, Debug)]
+    |            ^^^
+...
+ 55 |       pub fn from_primitive(tensor: K::Primitive<D>) -> Self {
+    |       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+...
+ 60 |       pub fn empty<S: Into<Shape<D>>>(shape: S, device: &B::Device) -> Self {
+    |       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+...
+717 | /     pub fn from_data<T>(data: T, device: &B::Device) -> Self
+718 | |     where
+719 | |         T: Into<TensorData>,
+    | |____________________________^
+    = note: the function or associated item was found for
+            - `burn::tensor::Tensor<B, 1, burn::tensor::Int>`
+    = note: this error originates in the derive macro `new` (in Nightly builds, run with -Z macro-backtrace for more info)
+
+warning: variable does not need to be mutable
+  --> node/src/burn_smollm/loader.rs:70:13
+   |
+70 |         let mut layer = &mut model.layers[i];
+   |             ----^^^^^
+   |             |
+   |             help: remove this `mut`
+   |
+   = note: `#[warn(unused_mut)]` (part of `#[warn(unused)]`) on by default
+
+warning: unused variable: `batch`
+  --> node/src/burn_smollm/model.rs:79:14
+   |
+79 |         let [batch, seq_len] = input_ids.dims();
+   |              ^^^^^ help: if this is intentional, prefix it with an underscore: `_batch`
+   |
+   = note: `#[warn(unused_variables)]` (part of `#[warn(unused)]`) on by default
+
+warning: unused variable: `seq_len`
+  --> node/src/burn_smollm/model.rs:79:21
+   |
+79 |         let [batch, seq_len] = input_ids.dims();
+   |                     ^^^^^^^ help: if this is intentional, prefix it with an underscore: `_seq_len`
+
+Some errors have detailed explanations: E0061, E0277, E0282, E0308, E0599.
+For more information about an error, try `rustc --explain E0061`.
+warning: `node` (lib) generated 19 warnings
+error: could not compile `node` (lib) due to 21 previous errors; 19 warnings emitted
+Error: Compiling your crate to WebAssembly failed
+Caused by: Compiling your crate to WebAssembly failed
+Caused by: failed to execute `cargo build`: exited with exit status: 101
+  full command: cd "/app/node" && "cargo" "build" "--lib" "--release" "--target" "wasm32-unknown-unknown"

network-poc/docker-compose.yml

@@ -9,7 +9,7 @@ services:
     volumes:
       - .:/app
     # Käännetään aina käynnistyksen yhteydessä varmuuden vuoksi Wasm uusimmista koodeista, ja päälle pyöräytetään Hub!
-    command: bash -c "cd node && wasm-pack build --target web --out-dir ../static/pkg && cd ../hub && cargo run"
+    command: bash -c "cd node && wasm-pack build --release --target web --out-dir ../static/pkg && cd ../hub && cargo run"
 
   # Valinnainen natiivi-solmu — kerää oikeat laitteistotiedot (nvidia-smi-taso)
   native-node:

network-poc/hub/src/main.rs

@@ -978,7 +978,7 @@ async fn api_chat_completions(
         let busy = state.node_busy.lock().unwrap();
         let matching: Vec<u64> = tasks.iter().filter(|(_, task)| {
             if payload.model == "qwen-coder" {
-                *task == "qwen-coder-05b" || *task == "qwen-coder"
+                task.starts_with("qwen-coder")
             } else {
                 **task == payload.model
             }

network-poc/node/src/qwen_coder.rs

@@ -1,6 +1,8 @@
 use candle_core::{Device, Tensor, DType};
+use candle_core::quantized::gguf_file;
 use candle_nn::VarBuilder;
 use candle_transformers::models::qwen2::{Config as QwenConfig, ModelForCausalLM as QwenModel};
+use candle_transformers::models::quantized_qwen2::ModelWeights as QwenQuantizedModel;
 use wasm_bindgen::JsCast;
 use std::cell::RefCell;
 use std::rc::Rc;
@@ -16,13 +18,36 @@ 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 — parempi laatu, vaatii enemmän muistia (~6 GB lataus, ~12 GB RAM)
-const MODEL_3B_PART1_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct/resolve/main/model-00001-of-00002.safetensors";
-const MODEL_3B_PART2_URL: &str = "https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct/resolve/main/model-00002-of-00002.safetensors";
-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),
+    Quantized(QwenQuantizedModel),
+}
+
+impl CoderModel {
+    fn forward(&mut self, x: &Tensor, pos: usize) -> candle_core::Result<Tensor> {
+        match self {
+            CoderModel::Full(m) => m.forward(x, pos),
+            CoderModel::Quantized(m) => m.forward(x, pos),
+        }
+    }
+
+    fn clear_kv_cache(&mut self) {
+        match self {
+            CoderModel::Full(m) => m.clear_kv_cache(),
+            CoderModel::Quantized(_) => {
+                // Quantized model nollaa KV-cachen automaattisesti kun forward kutsutaan pos=0:lla
+                // (ks. quantized_qwen2.rs rivi 118: if index_pos == 0)
+            }
+        }
+    }
+}
 
 struct CachedModel {
-    model: QwenModel,
+    model: CoderModel,
     tokenizer: tokenizers::Tokenizer,
     is_3b: bool,
 }
@@ -182,50 +207,39 @@ 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))?;
 
     // Painot
-    let tensors = if use_3b {
-        let part1 = ensure_cached("coder3b-model-part1.safetensors", MODEL_3B_PART1_URL, ws).await?;
-        let part2 = ensure_cached("coder3b-model-part2.safetensors", MODEL_3B_PART2_URL, ws).await?;
-        console_log!("[Coder] Rakennetaan 3B-mallia...");
-        let mut all_tensors = candle_core::safetensors::load_buffer(&part1[..], &device)
-            .map_err(|e| format!("Part1: {}", e))?;
-        let tensors2 = candle_core::safetensors::load_buffer(&part2[..], &device)
-            .map_err(|e| format!("Part2: {}", e))?;
-        all_tensors.extend(tensors2);
-        all_tensors
+    let model = if use_3b {
+        // GGUF Q4_K_M — kvantisoidtu 3B-malli (~1.9 GB)
+        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))?;
+        let qmodel = QwenQuantizedModel::from_gguf(content, &mut cursor, &device)
+            .map_err(|e| format!("GGUF model: {}", e))?;
+        CoderModel::Quantized(qmodel)
     } else {
         let model_bytes = ensure_cached("coder05b-model.safetensors", MODEL_05B_URL, ws).await?;
         console_log!("[Coder] Rakennetaan 0.5B-mallia...");
-        candle_core::safetensors::load_buffer(&model_bytes[..], &device)
-            .map_err(|e| format!("Safetensors: {}", e))?
-    };
-
-    let vb = VarBuilder::from_tensors(tensors, dtype, &device);
-    let config = if use_3b {
-        QwenConfig {
-            vocab_size: 151936, hidden_size: 2048, intermediate_size: 11008,
-            num_hidden_layers: 36, num_attention_heads: 16, num_key_value_heads: 2,
-            max_position_embeddings: 32768, sliding_window: 32768, max_window_layers: 36,
-            tie_word_embeddings: true, rope_theta: 1000000.0, rms_norm_eps: 1e-6,
-            use_sliding_window: false, hidden_act: candle_nn::Activation::Silu,
-        }
-    } else {
-        QwenConfig {
+        let tensors = candle_core::safetensors::load_buffer(&model_bytes[..], &device)
+            .map_err(|e| format!("Safetensors: {}", e))?;
+        let config = QwenConfig {
             vocab_size: 151936, hidden_size: 896, intermediate_size: 4864,
             num_hidden_layers: 24, num_attention_heads: 14, num_key_value_heads: 2,
             max_position_embeddings: 32768, sliding_window: 32768, max_window_layers: 21,
             tie_word_embeddings: true, rope_theta: 1000000.0, rms_norm_eps: 1e-6,
             use_sliding_window: false, hidden_act: candle_nn::Activation::Silu,
-        }
+        };
+        let vb = VarBuilder::from_tensors(tensors, dtype, &device);
+        let qwen = QwenModel::new(&config, vb).map_err(|e| format!("Malli: {}", e))?;
+        CoderModel::Full(qwen)
     };
-
-    let model = QwenModel::new(&config, vb).map_err(|e| format!("Malli: {}", e))?;
     console_log!("[Coder] Malli ladattu ja välimuistitettu");
 
     MODEL_CACHE.with(|c| {

network-poc/static/index.html

@@ -1156,8 +1156,8 @@
             manager: { name: 'Manageri — System Prompt', model: 'qwen-coder', default: 'Olet projektipäällikkö. Jaa tehtävät osiin, priorisoi ja koordinoi tiimin työtä.' },
             coder: { name: 'Koodari — System Prompt', model: 'qwen-coder', default: 'Olet kokenut ohjelmistokehittäjä. Kirjoita selkeää, testattavaa koodia ja vastaa aina koodilla.' },
             data: { name: 'Data-Agentti — System Prompt', model: 'qwen-coder', default: 'Olet tietokanta-asiantuntija. Vastaat skeemojen suunnittelusta, SQL-kyselyiden optimoinnista ja datamalleista.' },
-            qa: { name: 'QA — System Prompt', model: 'smollm-135m', default: 'Olet laadunvarmistaja (QA). Kirjoitat testejä, etsit virheitä ja varmistat, että kaikki reunatapaukset on huomioitu.' },
-            tester: { name: 'DevOps — System Prompt', model: 'smollm-135m', default: 'Olet DevOps-insinööri. Vastaat koodin julkaisuputkista, serveri-infrastruktuurista ja ympäristön suorituskyvystä.' },
+            qa: { name: 'QA — System Prompt', model: 'qwen-coder', default: 'Olet laadunvarmistaja (QA). Kirjoitat testejä, etsit virheitä ja varmistat, että kaikki reunatapaukset on huomioitu.' },
+            tester: { name: 'DevOps — System Prompt', model: 'qwen-coder', default: 'Olet DevOps-insinööri. Vastaat koodin julkaisuputkista, serveri-infrastruktuurista ja ympäristön suorituskyvystä.' },
         };
         const selectedAgents = new Set();
         let sharedPrompt = localStorage.getItem('kpn-shared-prompt') || '';
@@ -2229,7 +2229,7 @@ Write the corrected code.`;
                 // Mallikatalogista valinta numerolla tai nimellä
                 const loadModels = [
                     { id: '1', key: '05b', name: 'Qwen2.5-Coder:0.5B', size: '~990 MB', coderSize: '05b' },
-                    { id: '2', key: '3b',  name: 'Qwen2.5-Coder:3B',  size: '~6.2 GB', coderSize: '3b' },
+                    { id: '2', key: '3b',  name: 'Qwen2.5-Coder:1.5B Q4',  size: '~1 GB', coderSize: '3b' },
                 ];
                 if (!arg) {
                     // Näytetään lista
@@ -2268,7 +2268,7 @@ Write the corrected code.`;
             if (sub === 'models') {
                 termLog('  Käytettävissä olevat mallit:', '#c9d1d9');
                 termLog('    <span style="color:#58a6ff">1</span>  qwen-coder     Qwen2.5-Coder:0.5B  <span style="color:#8b949e">~990 MB  | koodin generointi</span>');
-                termLog('    <span style="color:#58a6ff">2</span>  qwen-coder-3b  Qwen2.5-Coder:3B    <span style="color:#8b949e">~6.2 GB  | parempi koodinlaatu</span>');
+                termLog('    <span style="color:#58a6ff">2</span>  qwen-coder-3b  Qwen2.5-Coder:1.5B Q4 <span style="color:#8b949e">~1 GB  | kvantisoidtu, parempi laatu</span>');
                 termLog('    <span style="color:#58a6ff">3</span>  smollm-135m    SmolLM 135M         <span style="color:#8b949e">~270 MB  | kevyt, nopea</span>');
                 termLog('    <span style="color:#58a6ff">4</span>  qwen-05b       Qwen2.5:0.5B        <span style="color:#8b949e">~990 MB  | yleismalli</span>');
                 termLog('    <span style="color:#58a6ff">5</span>  phi3-mini      Phi-3 Mini          <span style="color:#8b949e">~2.2 GB  | Microsoftin malli</span>');
@@ -2552,7 +2552,7 @@ Write the corrected code.`;
         uiSocket.onmessage = (event) => {
             try {
                 const raw = event.data;
-                if (raw.includes('"single_tokenize"') || raw.includes('"download_progress"')) return;
+                if (raw.includes('"single_tokenize"')) return;
 
                 const data = JSON.parse(raw);
                 if (data.type === "stats") {
@@ -2577,6 +2577,24 @@ Write the corrected code.`;
                     } else {
                         dlBar.style.display = 'none';
                     }
+                    // Terminaaliin latauksen edistyminen
+                    const term = document.getElementById('agent-terminal');
+                    if (term) {
+                        let dlLine = term.querySelector('.term-download');
+                        if (data.pct >= 100) {
+                            if (dlLine) dlLine.remove();
+                            termLog(`  <span style="color:#3fb950">✓</span> ${data.file} ladattu`, '#a5d6ff');
+                        } else {
+                            if (!dlLine) {
+                                dlLine = document.createElement('div');
+                                dlLine.className = 'terminal-line term-download';
+                                term.appendChild(dlLine);
+                            }
+                            const bar = '█'.repeat(Math.floor(data.pct / 5)) + '░'.repeat(20 - Math.floor(data.pct / 5));
+                            dlLine.innerHTML = `  <span style="color:#d29922">${data.file}</span> <span style="color:#8b949e">${bar}</span> <span style="color:#58a6ff">${data.pct}%</span> <span style="color:#8b949e">${data.loaded_mb}/${data.total_mb} MB</span>`;
+                            term.scrollTop = term.scrollHeight;
+                        }
+                    }
                 } else if (data.type === "single_tokenize_done") {
                     chatBox.classList.remove('hidden');
                     const r = data.result || {};
@@ -3104,10 +3122,43 @@ Write the corrected code.`;
         }
 
         // Kuuntele console.log-viestejä pipeline-vaiheiden seuraamiseksi
+        // Terminaalin lataustilarivi — päivittyy dynaamisesti
+        function termLoadStatus(phase, detail) {
+            const term = document.getElementById('agent-terminal');
+            if (!term) return;
+            let statusLine = term.querySelector('.term-load-status');
+            if (!statusLine) {
+                statusLine = document.createElement('div');
+                statusLine.className = 'terminal-line term-load-status';
+                term.appendChild(statusLine);
+            }
+            const spinner = ['⠋','⠙','⠹','⠸','⠼','⠴','⠦','⠧','⠇','⠏'];
+            const frame = spinner[Math.floor(Date.now() / 100) % spinner.length];
+            statusLine.innerHTML = `  <span style="color:#d29922">${frame}</span> <span style="color:#8b949e">${phase}</span>${detail ? ` <span style="color:#58a6ff">${detail}</span>` : ''}`;
+            term.scrollTop = term.scrollHeight;
+        }
+        function termLoadDone() {
+            const term = document.getElementById('agent-terminal');
+            if (!term) return;
+            const statusLine = term.querySelector('.term-load-status');
+            if (statusLine) statusLine.remove();
+        }
+
         const origCodeLog = console.log;
         const codeLogListener = (...args) => {
             const msg = args.join(' ');
             if (msg.includes('[Coder]') || msg.includes('[Storage]') || msg.includes('Burn Wasm') || msg.includes('Kipinä Agent Node')) {
+                // Terminaalin lataustilapäivitys
+                if (msg.includes('Agent Node käynnistyy')) termLoadStatus('WASM alustettu');
+                if (msg.includes('Ladataan') && msg.includes('tokenizer')) termLoadStatus('Ladataan tokenizer...');
+                if (msg.includes('tokenizer') && (msg.includes('löytyi') || msg.includes('tallennettu'))) termLoadStatus('Tokenizer ✓');
+                if (msg.includes('Ladataan') && msg.includes('gguf')) termLoadStatus('Ladataan mallia...');
+                const dlMatch = msg.match(/lataus: (\d+)%/);
+                if (dlMatch) termLoadStatus('Ladataan mallia...', dlMatch[1] + '%');
+                if (msg.includes('tallennettu') && msg.includes('gguf')) termLoadStatus('Malli tallennettu');
+                if (msg.includes('Rakennetaan')) termLoadStatus('Rakennetaan mallia...');
+                if (msg.includes('Malli ladattu')) termLoadDone();
+
                 if (msg.includes('Burn Wasm')) setStep('step-wasm', 'active');
                 if (msg.includes('Agent Node käynnistyy')) { setStep('step-wasm', 'done'); }
                 // Tokenizer: [Coder] tai [Storage] -prefiksi
@@ -3160,6 +3211,15 @@ Write the corrected code.`;
                     if (cl) { cl.textContent = 'Qwen2.5-Coder:' + (coderSize === '3b' ? '3B' : '0.5B'); cl.style.color = '#3fb950'; cl.title = sizeLabel + ' · Candle Wasm · CPU · max 512 tok'; }
                     if (btn) { btn.dataset.state = 'ready'; btn.textContent = '✓ Valmis'; btn.style.borderColor = '#3fb950'; btn.style.color = '#3fb950'; btn.style.cursor = 'default'; btn.title = 'Kielimalli ladattu — oma kone on valmis laskentaan'; }
                     localStorage.setItem('kpn-coder-loaded', 'true');
+                    // Terminaaliin valmis-viesti (vain kerran)
+                    if (!window._coderReadyLogged) {
+                        window._coderReadyLogged = true;
+                        const term = document.getElementById('agent-terminal');
+                        if (term) {
+                            const sLabel = coderSize === '3b' ? 'Qwen2.5-Coder:1.5B Q4' : 'Qwen2.5-Coder:0.5B';
+                            termLog(`  <span style="color:#3fb950">✓</span> ${sLabel} valmis — kpn run coder "prompti"`, '#3fb950');
+                        }
+                    }
                 }
                 if (msg.includes('[Coder]') && msg.includes('Syöte:')) {
                     // Pipeline piiloon kun generointi alkaa