OMNIML-4668 — hidden_state_dump_support: step2_hidden.yaml for Qwen3-8B

Agent-authored via pensieve-intern's hidden_state_dump_support
stage on Epic OMNIML-4666. Faithful extraction of task_1
(dump_offline_data.sh — hidden-state capture from target model)
from hf_offline_eagle3.yaml's monolithic pipeline, renamed task_0
for the standalone step convention.

Signed-off-by: Chenhan D. Yu <chenhany@nvidia.com>

[ci-retrigger] previous run had a cancelled matrix variant (runner pre-emption, not a real test failure)

Author: ChenhanYu

tools/launcher/examples/Qwen/Qwen3-8B/step2_hidden.yaml

@@ -0,0 +1,40 @@
+# EAGLE3 offline speculative decoding pipeline for Qwen3-8B.
+#
+# 4-step pipeline:
+#   task_0: Data synthesis — query TRT-LLM server to generate prompt samples
+#   task_1: Dump hidden states — run target model to capture hidden states
+#   task_2: Offline training — train the EAGLE3 draft head
+#   task_3: Benchmark — evaluate speculative decoding speedup via VLLM
+#
+# All tasks share /scratchspace to pass artifacts between steps.
+#
+# Usage:
+#   uv run launch.py --yaml examples/Qwen/Qwen3-8B/hf_offline_eagle3.yaml --yes
+#   uv run slurm.py --yaml modules/Model-Optimizer/tools/launcher/examples/Qwen/Qwen3-8B/hf_offline_eagle3.yaml --yes
+
+job_name: Qwen3-8B_EAGLE3_hidden_dump
+pipeline:
+  allow_to_fail: false
+  skip: false
+  note:
+
+  global_vars:
+    hf_model: /hf-local/Qwen/Qwen3-8B
+
+  # Step 2: Dump hidden states from target model
+  task_0:
+    script: common/eagle3/dump_offline_data.sh
+    args:
+      - --input-data /scratchspace/data
+      - --output-dir /scratchspace/offline_hidden_states
+      - --max-seq-len 8192
+      - --tp 8
+      - --moe-ep 8
+    environment:
+      - HF_MODEL_CKPT: <<global_vars.hf_model>>
+    slurm_config:
+      _factory_: "slurm_factory"
+      nodes: 1
+      ntasks_per_node: 8
+      gpus_per_node: 8
+      container: nvcr.io/nvidia/tensorrt-llm/release:1.2.0