train_grpo: prebuilt flash-attn wheel + verbose training rollouts

- Install flash-attn 2.7.4.post1 via prebuilt wheel matched to
torch 2.5 / py3.11 / cu12 instead of building from source. The HF
Job container (pytorch/pytorch:*-runtime) has no nvcc, so the
source build always fails with CUDA_HOME unset. Triple fallback:
prebuilt -> source build -> skip and use sdpa.
- Flip verbose=False to verbose=True in the training-loop rollouts so
per-day generation progress is visible while a round runs (~minutes
per round at 6 eps x 15 days x 2 phases).

Made-with: Cursor

training/train_grpo.ipynb

@@ -33,7 +33,10 @@
         "!pip install -q matplotlib pandas\n",
         "!pip install -q \"typing_extensions>=4.13.0\" pydantic httpx\n",
         "!pip install -q \"openenv-core[core]>=0.2.2\"\n",
-        "!pip install -q flash-attn --no-build-isolation || echo \"flash-attn install skipped; will use sdpa\""
+        "# flash-attn: install prebuilt wheel matched to torch 2.5 + py3.11 + cu12 (HF Job container).\n",
+        "# This avoids the from-source build that fails when the container has no nvcc / CUDA_HOME.\n",
+        "# Falls back to sdpa if the wheel install fails (e.g. on a different env).\n",
+        "!pip install -q \"https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.5cxx11abiFALSE-cp311-cp311-linux_x86_64.whl\" || pip install -q flash-attn --no-build-isolation || echo \"flash-attn install skipped; will use sdpa\""
       ],
       "execution_count": null,
       "outputs": []
@@ -881,7 +884,7 @@
         "        tasks_seeds = [(TASKS[ep % len(TASKS)], 42 + ep + round_idx * 10) for ep in range(EPISODES_PER_ROUND)]\n",
         "        t_roll = time.time()\n",
         "        results = run_llm_episodes_batched(\n",
-        "            peft_model, tokenizer, tasks_seeds, verbose=False, eval=False,\n",
+        "            peft_model, tokenizer, tasks_seeds, verbose=True, eval=False,\n",
         "            system=sys_prompt, hint_peak_hours=use_hint, reward_mode=reward_mode,\n",
         "            log_tag=f\"{phase_name}_r{round_idx}\",\n",
         "        )\n",