#6607 [Engine] Inkling-2-Mini setup stops on a PC with two NVIDIA cards: AttributeError 'int' object has no attribute 'get' in install_inkling2

open · @stuchapin909 · 3 commentaires · Sur GitHub

needs-format

Description

### Area

[Engine]

### Desktop app, Terminal or source?

Terminal (the `strata` command)

### What happened

On a PC with two NVIDIA cards, `strata launch` for Inkling-2-Mini stops in setup before it writes a config. The engine download and its test start pass; setup then raises in `install_inkling2`:

```
File "...\components\control\0.1.65\setup.py", line 11376, in main
    return install_inkling2(a, family, gpu, hip, amd, cuda_tk, multi, port, ram, adopted)
File "...\components\control\0.1.65\setup.py", line 10197, in install_inkling2
    dual_gpus = [g for g in (multi or []) if g.get("vendor", "nvidia") == "nvidia"] if (not hip and not amd and not unified) else []
AttributeError: 'int' object has no attribute 'get'
```

`strata launch` reports it as: `The model could not be installed. Setup stopped with an error (AttributeError: 'int' object has no attribute 'get')`.

Step 1 of the setup log has already chosen both cards ("Main: GPU 2 RTX 3090. Helper: GPU 1 RTX 3090", "GPU 0 + GPU 1 together (the model's layers are split across them)"), so `multi` reaches `install_inkling2` as a list of GPU indices (ints), and line 10197 expects dicts with a `vendor` key.

Adding `--gpu 0` avoids it: setup then finishes, verifies the 20.0 GB of files and serves SQ-20 on one card.

Expected: setup either builds the two-card config the 0.2.19 notes describe (`--gpus 2`) or falls back to one card, instead of stopping.

### Smallest steps or command that reproduce it

```
strata launch --model sq-20 --family inkling-2-mini --context 128k --access local --port 8080 --save-preset ink-sq20 --no-tui --yes
```

on a PC with two NVIDIA GPUs (the SQ-20 files were already in `Documents\Strata\models\inkling-2-mini`, SHA-256 matching `SHA256SUMS.txt`).

Workaround: the same command with `--gpu 0`.

### Reproduction status

Reproduced locally

### Strata version

0.2.19 (Control 0.1.65, strata-inkling 0.1.65 windows-x86_64-cuda-cuda13)

### GPU, CPU, RAM, OS

2x RTX 3090 24 GB (sm_86, driver 617.42), i9-9920X, 128 GB RAM, Windows 11 (10.0.26300)

### Model, size and settings

Inkling-2-Mini SQ-20 (`strata-engine/Inkling-2-Mini-Strata`), context 128k, defaults otherwise

### Logs

`%LOCALAPPDATA%\Strata\launch\ink-sq20.setup.log`, from Step 1 to the traceback:

```
=== Step 1: checking your PC ===
  Your NVIDIA GPUs:
    GPU 0: NVIDIA GeForce RTX 3090, 24 GB VRAM - can be used
    GPU 1: NVIDIA GeForce RTX 3090, 24 GB VRAM - can be used
    GPU 1  RTX 3090                 24 GB   Helper
    GPU 2  RTX 3090                 24 GB   Main
  [ok] Main: GPU 2 RTX 3090. Helper: GPU 1 RTX 3090.
  [ok] GPUs: GPU 0 (NVIDIA GeForce RTX 3090, 24 GB) + GPU 1 (NVIDIA GeForce RTX 3090, 24 GB) together (the model's layers are split across them)
  [ok] GPU: NVIDIA GeForce RTX 3090, 24.0 GB VRAM, compute capability 8.6, driver 617.42
  [ok] RAM: 128 GB
=== Step 2: your choices ===
  [ok] model: Inkling-2-Mini
Traceback (most recent call last):
  File "...\setup.py", line 11913, in <module>
  File "...\setup.py", line 11376, in main
  File "...\setup.py", line 10197, in install_inkling2
AttributeError: 'int' object has no attribute 'get'
```

Related: #6592 (the `--gpus 2` engine crash after a config is written by hand) and its note that setup writes a one-GPU config on two cards.

Sur le site

Liens install, modèles, releases.