mirror of
https://github.com/NVIDIA/Model-Optimizer.git
synced 2026-11-08 09:03:20 +08:00
skierat
c4129b6e03
### What does this PR do? Type of change: new example Adds an end-to-end Cosmos3 Nano DFlash training recipe for multimodal speculative decoding. - Adds a notebook that prepares data, launches synthetic generation in Slurm, trains a DFlash draft model, exports it, and provides a vLLM smoke-test command. - Adds PAI-Understanding, VQA v2, and multilingual prompt sharding and distributed-generation helpers. - Adds an atomic, multimodal-safe merge and conservative deduplication flow. - Extends the VLM data collator to handle structured image/video messages, configurable visual bounds, and fixed DFlash sequence lengths. - Hardens generation launch scripts and preserves truncated generated responses. ### Usage ```bash cd examples/speculative_decoding/recipes export MODEL_PATH=/path/to/cosmos3-nano export PLAIN_TEXT_INPUT=/path/to/nemotron-chat-or-approved-user-data.jsonl jupyter lab train_dflash_cosmos3_nano.ipynb Run the notebook in order: 1. Configure paths. 2. Prepare prompts on a CPU-only node and generate target completions in a Slurm GPU allocation. 3. Merge the four required sources and submit training. 4. Export a saved checkpoint and run the vLLM deployment smoke test. ### Testing - jq empty examples/speculative_decoding/recipes/train_dflash_cosmos3_nano.ipynb - bash -n on the modified launch, worker, and recipe shell scripts. - Ran a two-step Cosmos3 Nano DFlash Slurm smoke job; it completed and wrote modelopt_state.pth. - Not run: pytest tests/unit/torch/speculative/plugins/test_hf_speculative_offline.py (pytest is unavailable in the current environment). ### Before your PR is "Ready for review" - Is this change backward compatible?: ✅ - If you copied code from any other sources or added a new PIP dependency, did you follow guidance in CONTRIBUTING.md?: N/A - Did you write any new necessary tests?: ✅ - Did you update CHANGELOG.rst?: N/A - Did you get Claude approval on this PR?: N/A ### Additional Information Security follow-up required before marking ready: the notebook hardcodes model.trust_remote_code=true and --trust_remote_code. Either parameterize this with a default of false, or obtain and document a security exception. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **New Features** * Added end-to-end multimodal workflows for dataset preparation, distributed generation, result merging, training, export, and deployment testing. * Added support for image and video inputs, multiple dataset formats, resumable JSONL generation, configurable serving, and parallel processing. * Added configurable prompt, media, token, sequence, temperature, and tensor-parallel settings. * **Bug Fixes** * Improved truncated-response handling, assistant-label processing, validation, health checks, cleanup, deduplication, media resolution, atomic outputs, and failure reporting. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Signed-off-by: Slawomir Kierat <skierat@nvidia.com>
81 lines
2.8 KiB
Bash
81 lines
2.8 KiB
Bash
#!/bin/bash
|
|
# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
|
|
# SPDX-License-Identifier: Apache-2.0
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
set -euo pipefail
|
|
|
|
if [ $# -lt 9 ]; then
|
|
echo "Usage: $0 <job_id> <backend> <model_path> <data_path> <output_path> <scripts_path> <start_shard> <jobs_per_node> <comma_separated_node_names> <system_prompt>"
|
|
echo "Example: $0 245387 vllm /model/ /input_data/ /output_data/ /scripts/ 0 20 cluster-01,cluster-02 "\"You are a helpful assistant.\"""
|
|
exit 1
|
|
fi
|
|
|
|
JOB_ID=$1
|
|
BACKEND=$2
|
|
MODEL_PATH=$3
|
|
DATA_PATH=$4
|
|
OUTPUT_PATH=$5
|
|
SCRIPTS_PATH=$6
|
|
START_SHARD=$7
|
|
JOBS_PER_NODE=$8
|
|
NODE_NAME=$9
|
|
SYSTEM_PROMPT="${10:-}"
|
|
IFS=',' read -r -a NODE_LIST <<< "$NODE_NAME"
|
|
|
|
# Pyxis requires bind-mount sources to exist before srun creates the container.
|
|
mkdir -p "$OUTPUT_PATH"
|
|
|
|
# backend needs to be either vllm or sglang
|
|
if [ "$BACKEND" != "vllm" ] && [ "$BACKEND" != "sglang" ]; then
|
|
echo "Invalid backend: $BACKEND"
|
|
exit 1
|
|
fi
|
|
|
|
if [ "$BACKEND" == "vllm" ]; then
|
|
DEFAULT_CONTAINER_IMAGE="vllm/vllm-openai:v0.24.0"
|
|
else
|
|
DEFAULT_CONTAINER_IMAGE="lmsysorg/sglang:v0.5.3-cu129"
|
|
fi
|
|
CONTAINER_IMAGE=${CONTAINER_IMAGE:-$DEFAULT_CONTAINER_IMAGE}
|
|
counter=$START_SHARD
|
|
worker_pids=()
|
|
for node in "${NODE_LIST[@]}"; do
|
|
echo "Processing node: $node"
|
|
srun --output="srun_worker_${node}.log" --jobid="$JOB_ID" -N 1 --ntasks=1 --ntasks-per-node=1 -w "$node" \
|
|
--mpi pmix --overlap --container-image="$CONTAINER_IMAGE" \
|
|
--container-mounts="$MODEL_PATH":/model/,"$DATA_PATH":/input_data/,"$OUTPUT_PATH":/output_data/,"$SCRIPTS_PATH":/scripts/ \
|
|
bash /scripts/distributed_generate/worker.sh "$counter" "$BACKEND" "$JOBS_PER_NODE" "$SYSTEM_PROMPT" &
|
|
|
|
echo "srun command for node $node started with PID $!" >> srun_launch.log
|
|
worker_pids+=("$!")
|
|
|
|
# increment counter by JOBS_PER_NODE
|
|
counter=$((counter + JOBS_PER_NODE))
|
|
done
|
|
|
|
echo "Started workers, each processing $JOBS_PER_NODE shards of data. Will process shards $START_SHARD through $((counter - 1))."
|
|
|
|
worker_status=0
|
|
for worker_pid in "${worker_pids[@]}"; do
|
|
if ! wait "$worker_pid"; then
|
|
worker_status=1
|
|
fi
|
|
done
|
|
|
|
if [ "$worker_status" -ne 0 ]; then
|
|
echo "ERROR: one or more workers failed." >&2
|
|
fi
|
|
exit "$worker_status"
|