Claude
fe4d911a00
Replaces fixed tier table with real hardware probing and dynamic model selection.
gpu_detect.py — complete rewrite:
- Reads torch.cuda.get_device_properties + mem_get_info for actual free VRAM
- Detects: fp16 (CC≥6.0), bf16 (CC≥8.0), fp8 (CC≥8.9 Ada/Hopper),
int8 (CC≥7.0), tensor_cores (CC≥7.0), xformers presence
- Pre-Pascal (CC<6.0): effective_vram halved (fp32 weights are 2× larger)
- Subtracts 400MB driver overhead from free VRAM before model selection
- _select_txt2img / _select_inpaint / _select_img2img / _select_upscale:
eff≥20GB → FLUX.1-schnell (no offload)
eff≥10GB → FLUX.1-schnell (model_cpu_offload)
eff≥7.5GB → SDXL
eff≥5.5GB → SDXL + attention_slicing
eff≥3.5GB → SD 2.1
eff≥2.5GB → SD 2.1-base + attention_slicing
eff≥1.7GB → SD 1.5
else → SD 1.5 + sequential_cpu_offload
- ModelSpec carries: model_id, family, memory_opt, native_res, vram_fp16_gb
- Warnings: old CC, pre-Pascal fp32, fp8 upgrade hint, xformers install tip
- Compatibility shim get_model_ids() retained for existing callers
- infer_spec_from_model_id() auto-detects family from HF_MODEL_* overrides
local_diffusion.py — refactored to use ModelSpec:
- Reads spec from GpuCapabilities.recommended[op] instead of tier table
- FLUX.1-schnell: FluxPipeline / FluxImg2ImgPipeline, 4 steps, guidance=0.0
- SD families: family-aware pipeline class selection (sd15/sd2x/sdxl)
- Memory opts applied per ModelSpec.memory_opt field
- xformers attention enabled automatically when xformers detected
gpu_status.py — richer response:
- Exposes all feature flags (fp16/bf16/fp8/int8/tensor_cores/xformers)
- Returns full ModelSpec per operation (model_id, family, memory_opt, native_res)
ai_tools.py — /api/config exposes:
- gpu_vram_total, gpu_vram_free, gpu_cc, gpu_fp16, gpu_bf16, gpu_fp8,
gpu_tensor_cores, gpu_eff_vram, local_gpu_warnings
requirements.gpu.txt:
- diffusers bumped to >=0.29.0 (FLUX pipeline added in 0.29)
- transformers bumped to >=4.40.0
- sentencepiece added (FLUX T5 tokenizer)
scripts/gpu_setup.py:
- Prints full model table at startup (op → model_id, family, memory_opt, res)
- Shows all feature flags in one line
https://claude.ai/code/session_01WVDg7amsy1TTtxvpku7bcM
Classic Eyes Scripts
This directory contains scripts for generating and importing classic eye images into the AI Photo Edit patch library.
Overview
The scripts generate a variety of classic eye styles that can be used as reusable patches for photo editing:
- Realistic eyes - Detailed eyes with gradient irises and realistic highlights
- Anime eyes - Large, expressive eyes in anime style with prominent highlights
- Cartoon eyes - Simple, bold cartoon-style eyes
Each style is available in multiple iris colors: blue, green, brown, hazel, grey, and amber.
Scripts
1. generate_classic_eyes_ppm.py
Generates classic eye images using only the Python standard library (no dependencies required).
python scripts/generate_classic_eyes_ppm.py
This creates PPM format images in data/classic_eyes_ppm/. PPM is a simple image format that can be converted to PNG later.
2. download_classic_eyes.py
Full-featured script that generates eyes and imports them directly into the patch library. Requires PIL/Pillow.
# Run inside the backend container
docker exec -it ai-photo-edit-backend python /app/../scripts/download_classic_eyes.py
# Or with the API running
python scripts/download_classic_eyes.py --api --base-url http://localhost:8101
# Or save to a directory for later import
python scripts/download_classic_eyes.py --output-dir ./my_eyes
3. startup_import_eyes.py
Converts PPM files to PNG and imports them into the patch library. Run this inside the backend container.
docker exec -it ai-photo-edit-backend python /app/../scripts/startup_import_eyes.py
4. import_saved_eyes.py
Import previously saved eye images from a directory.
python scripts/import_saved_eyes.py /path/to/saved/eyes
Quick Start
-
Generate the eye images (no dependencies needed):
python scripts/generate_classic_eyes_ppm.py -
Start the application:
docker compose up -d -
Import the eyes:
docker exec -it ai-photo-edit-backend python /scripts/startup_import_eyes.py -
Verify in the app: Open the patch library in the UI to see the imported classic eyes.
Generated Eyes
| Style | Colors Available | Size |
|---|---|---|
| Realistic | Blue, Green, Brown, Hazel, Grey, Amber | 200x200 |
| Anime | Blue, Green, Brown, Hazel, Grey, Amber | 200x200 |
| Cartoon | Blue, Green, Brown, Hazel, Grey, Amber | 200x200 |
Total: 18 unique eye variants
Extending
To add more eye styles or colors, edit the generate_classic_eyes_ppm.py or download_classic_eyes.py scripts:
# Add new color
colors["purple"] = (128, 0, 128)
# Add new style in create_classic_eye() function
elif style == "fantasy":
# Your custom eye drawing code
pass
File Formats
- PPM: Portable Pixmap format - simple, universal, generated without dependencies
- PNG: Preferred format for the patch library - converted from PPM using PIL
License
The generated eye images are created programmatically and are free to use without restrictions.