Skip to content

GitLab

Unverified Commit 7e55a734 authored by Indrajit Bhosale's avatar Indrajit Bhosale Committed by GitHub
Browse files

chore: Remove llava-hf recipies (#6954)

parent dc14ce05
Hide whitespace changes
Inline Side-by-side
Showing with 4 additions and 70 deletions
examples/backends/trtllm/launch/disagg_multimodal.sh
View file @ 7e55a734
...@@ -4,14 +4,13 @@ ...@@ -4,14 +4,13 @@
# Environment variables with defaults # Environment variables with defaults
export DYNAMO_HOME=${DYNAMO_HOME:-"/workspace"} export DYNAMO_HOME=${DYNAMO_HOME:-"/workspace"}
export MODEL_PATH=${MODEL_PATH:-"llava-hf/llava-v1.6-mistral-7b-hf"} export MODEL_PATH=${MODEL_PATH:-"Qwen/Qwen3-VL-2B-Instruct"}
export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"llava-v1.6-mistral-7b-hf"} export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"Qwen3-VL-2B-Instruct"}
export PREFILL_ENGINE_ARGS=${PREFILL_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/llava-v1.6-mistral-7b-hf/prefill.yaml"} export PREFILL_ENGINE_ARGS=${PREFILL_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen3-vl-2b-instruct/prefill.yaml"}
export DECODE_ENGINE_ARGS=${DECODE_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/llava-v1.6-mistral-7b-hf/decode.yaml"} export DECODE_ENGINE_ARGS=${DECODE_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen3-vl-2b-instruct/decode.yaml"}
export PREFILL_CUDA_VISIBLE_DEVICES=${PREFILL_CUDA_VISIBLE_DEVICES:-"0"} export PREFILL_CUDA_VISIBLE_DEVICES=${PREFILL_CUDA_VISIBLE_DEVICES:-"0"}
export DECODE_CUDA_VISIBLE_DEVICES=${DECODE_CUDA_VISIBLE_DEVICES:-"1"} export DECODE_CUDA_VISIBLE_DEVICES=${DECODE_CUDA_VISIBLE_DEVICES:-"1"}
export MODALITY=${MODALITY:-"multimodal"} export MODALITY=${MODALITY:-"multimodal"}
export CUSTOM_TEMPLATE=${CUSTOM_TEMPLATE:-"$DYNAMO_HOME/examples/backends/trtllm/templates/llava_multimodal.jinja"}
# Setup cleanup trap # Setup cleanup trap
cleanup() { cleanup() {
...@@ -34,7 +33,6 @@ CUDA_VISIBLE_DEVICES=$PREFILL_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \ ...@@ -34,7 +33,6 @@ CUDA_VISIBLE_DEVICES=$PREFILL_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \
--served-model-name "$SERVED_MODEL_NAME" \ --served-model-name "$SERVED_MODEL_NAME" \
--extra-engine-args "$PREFILL_ENGINE_ARGS" \ --extra-engine-args "$PREFILL_ENGINE_ARGS" \
--modality "$MODALITY" \ --modality "$MODALITY" \
--custom-jinja-template "$CUSTOM_TEMPLATE" \
--disaggregation-mode prefill & --disaggregation-mode prefill &
PREFILL_PID=$! PREFILL_PID=$!
...@@ -44,5 +42,4 @@ CUDA_VISIBLE_DEVICES=$DECODE_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \ ...@@ -44,5 +42,4 @@ CUDA_VISIBLE_DEVICES=$DECODE_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \
--served-model-name "$SERVED_MODEL_NAME" \ --served-model-name "$SERVED_MODEL_NAME" \
--extra-engine-args "$DECODE_ENGINE_ARGS" \ --extra-engine-args "$DECODE_ENGINE_ARGS" \
--modality "$MODALITY" \ --modality "$MODALITY" \
--custom-jinja-template "$CUSTOM_TEMPLATE" \
--disaggregation-mode decode --disaggregation-mode decode
examples/backends/trtllm/launch/epd_multimodal_image.sh deleted 100755 → 0
View file @ dc14ce05
#!/bin/bash
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
# Environment variables with defaults
export DYNAMO_HOME=${DYNAMO_HOME:-"/workspace"}
export MODEL_PATH=${MODEL_PATH:-"llava-hf/llava-v1.6-mistral-7b-hf"}
export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"llava-v1.6-mistral-7b-hf"}
export PREFILL_ENGINE_ARGS=${PREFILL_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/llava-v1.6-mistral-7b-hf/prefill.yaml"}
export DECODE_ENGINE_ARGS=${DECODE_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/llava-v1.6-mistral-7b-hf/decode.yaml"}
export ENCODE_ENGINE_ARGS=${ENCODE_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/llava-v1.6-mistral-7b-hf/encode.yaml"}
export PREFILL_CUDA_VISIBLE_DEVICES=${PREFILL_CUDA_VISIBLE_DEVICES:-"0"}
export DECODE_CUDA_VISIBLE_DEVICES=${DECODE_CUDA_VISIBLE_DEVICES:-"1"}
export ENCODE_CUDA_VISIBLE_DEVICES=${ENCODE_CUDA_VISIBLE_DEVICES:-"2"}
export ENCODE_ENDPOINT=${ENCODE_ENDPOINT:-"dyn://dynamo.tensorrt_llm_encode.generate"}
export MODALITY=${MODALITY:-"multimodal"}
export CUSTOM_TEMPLATE=${CUSTOM_TEMPLATE:-"$DYNAMO_HOME/examples/backends/trtllm/templates/llava_multimodal.jinja"}
# Setup cleanup trap
cleanup() {
echo "Cleaning up background processes..."
kill $DYNAMO_PID $PREFILL_PID $DECODE_PID $ENCODE_PID 2>/dev/null || true
wait $DYNAMO_PID $PREFILL_PID $DECODE_PID $ENCODE_PID 2>/dev/null || true
echo "Cleanup complete."
}
trap cleanup EXIT INT TERM
# run frontend
python3 -m dynamo.frontend --http-port 8000 &
DYNAMO_PID=$!
# run encode worker
CUDA_VISIBLE_DEVICES=$ENCODE_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \
--model-path "$MODEL_PATH" \
--served-model-name "$SERVED_MODEL_NAME" \
--extra-engine-args "$ENCODE_ENGINE_ARGS" \
--modality "$MODALITY" \
--disaggregation-mode encode &
ENCODE_PID=$!
# run prefill worker
CUDA_VISIBLE_DEVICES=$PREFILL_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \
--model-path "$MODEL_PATH" \
--served-model-name "$SERVED_MODEL_NAME" \
--extra-engine-args "$PREFILL_ENGINE_ARGS" \
--modality "$MODALITY" \
--disaggregation-mode prefill \
--encode-endpoint "$ENCODE_ENDPOINT" \
--custom-jinja-template "$CUSTOM_TEMPLATE" &
PREFILL_PID=$!
# run decode worker
CUDA_VISIBLE_DEVICES=$DECODE_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \
--model-path "$MODEL_PATH" \
--served-model-name "$SERVED_MODEL_NAME" \
--extra-engine-args "$DECODE_ENGINE_ARGS" \
--modality "$MODALITY" \
--disaggregation-mode decode \
--custom-jinja-template "$CUSTOM_TEMPLATE" &
DECODE_PID=$!
wait $DYNAMO_PID
\ No newline at end of file
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment