feat: training pipeline for all 10 ML ensemble models

- Add standalone training binaries: TGGN, KAN, xLSTM, Diffusion (DBN data)
- Update Dockerfile.training: 6 → 16 binaries (all 10 models + hyperopt + baseline)
- Expand train.sh: 4 → 10 models, fix registry URL and GPU pool nodeSelector
- Add GPU overlay manifests for trading-service and ml-training-service
- Create training data PVC and upload pod manifests
- Expand web-gateway model validation: 4 → 10 types (training + tune routes)
- Extend dashboard: 10 model cards grouped by category (RL/Temporal/Graph/Generative)
- Add training image build job to Gitea CI workflow
- Update GPU taint controller to exclude inference pool from tainting
- Fix job-template nodeSelector: gpu → gpu-training

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

infra/scripts/train.sh

@@ -11,7 +11,7 @@ set -euo pipefail
 # Presets:
 #   quick-test    - fast sanity check (max-steps=500, requires --model)
 #   single-model  - full training run for one model (requires --model)
-#   full-ensemble - trains all 4 models (dqn ppo tft mamba2) in parallel
+#   full-ensemble - trains all 10 models sequentially (dqn ppo tft mamba2 tggn tlob liquid kan xlstm diffusion)
 #   hyperopt      - hyperparameter optimisation (requires --model, uses --trials)
 # ---------------------------------------------------------------------------
 
@@ -22,19 +22,25 @@ MAX_STEPS=2000
 TIMEOUT=3600
 DATA_DIR="/data/futures-baseline"
 NAMESPACE="foxhunt"
-IMAGE="rg.nl-ams.scw.cloud/foxhunt/training:latest"
+IMAGE="rg.fr-par.scw.cloud/foxhunt/training:latest"
 MODEL=""
 PRESET=""
 TIMESTAMP="$(date +%s)"
 
-ALL_MODELS=(dqn ppo tft mamba2)
+ALL_MODELS=(dqn ppo tft mamba2 tggn tlob liquid kan xlstm diffusion)
 
 # --- Model-to-binary mapping ----------------------------------------------
 declare -A MODEL_BINARY=(
-  [dqn]=train_dqn
+  [dqn]=train_dqn_es_fut
   [ppo]=train_ppo_parquet
-  [tft]=hyperopt_tft_demo
-  [mamba2]=hyperopt_mamba2_demo
+  [tft]=train_tft_dbn
+  [mamba2]=train_mamba2_dbn
+  [tggn]=train_tggn_dbn
+  [tlob]=train_tlob
+  [liquid]=train_liquid_dbn
+  [kan]=train_kan_dbn
+  [xlstm]=train_xlstm_dbn
+  [diffusion]=train_diffusion_dbn
 )
 
 # --- Helpers --------------------------------------------------------------
@@ -45,11 +51,11 @@ Usage: ./infra/scripts/train.sh <preset> [OPTIONS]
 Presets:
   quick-test      Fast sanity check (requires --model, max-steps=500)
   single-model    Full training run (requires --model)
-  full-ensemble   Trains dqn, ppo, tft, mamba2 in parallel Jobs
+  full-ensemble   Trains all 10 models sequentially (dqn ppo tft mamba2 tggn tlob liquid kan xlstm diffusion)
   hyperopt        Hyperparameter search (requires --model, uses --trials)
 
 Options:
-  --model MODEL     Model name: dqn | ppo | tft | mamba2
+  --model MODEL     Model name: dqn | ppo | tft | mamba2 | tggn | tlob | liquid | kan | xlstm | diffusion
   --symbol SYMBOL   Trading symbol          (default: ES.FUT)
   --trials N        Hyperopt trial count    (default: 20)
   --max-steps N     Max steps per epoch     (default: 2000)
@@ -152,7 +158,7 @@ spec:
     spec:
       restartPolicy: Never
       nodeSelector:
-        k8s.scaleway.com/pool-name: gpu
+        k8s.scaleway.com/pool-name: gpu-training
       tolerations:
         - key: nvidia.com/gpu
           operator: Exists
@@ -218,6 +224,8 @@ case "$PRESET" in
   full-ensemble)
     for m in "${ALL_MODELS[@]}"; do
       submit_job "$m"
+      echo "  Waiting for ${m} to complete before next model..."
+      kubectl -n "${NAMESPACE}" wait --for=condition=complete "job/train-${m}-${TIMESTAMP}" --timeout="${TIMEOUT}s" 2>/dev/null || true
     done
     ;;
 esac