Colab#4

Draft

ironmanizawesome wants to merge 15 commits into

CandleLabAI:mainfrom

ironmanizawesome:colab

Commits on May 5, 2026

fix(models): unpack get_encoder tuple in PCBClassNet build

ironmanizawesome
and
claude
committed
docs(colab): add notebook + readme for Colab training

ironmanizawesome
and
claude
committed
docs(colab): use condacolab to pin Python 3.10 + TF 2.10 stack

ironmanizawesome
and
claude
committed

Commits on May 10, 2026

docs(colab): preprocess in Colab + cap full training at 40 epochs

ironmanizawesome
and
claude
committed
fix(colab): pin condacolab to Python 3.10 explicitly

ironmanizawesome
and
claude
committed
docs(colab): switch to TF 2.15 + clean notebook duplicates

ironmanizawesome
and
claude
committed
fix(colab): use tensorflow[and-cuda] extra so TF 2.15 sees the GPU

ironmanizawesome
and
claude
committed
fix(colab): install nvidia CUDA wheels directly (and-cuda extra is broken)

ironmanizawesome
and
claude
committed
fix(colab): pin install + run to system python3.11 binary

ironmanizawesome
and
claude
committed
fix(colab): set TF_GPU_ALLOCATOR=cuda_malloc_async on every training cell

ironmanizawesome
and
claude
committed
fix(colab): install python3.11 if missing + pin CUDA wheel versions

ironmanizawesome
and
claude
committed
config(colab): bump full training to 80 epochs each (was 40)

ironmanizawesome
and
claude
committed

Commits on May 11, 2026