[EPD] Allow KV consumers to omit MM embeddings (#52697)
Signed-off-by: zhenwei-intel <zhenwei.liu@intel.com>
This commit is contained in:
@@ -183,6 +183,7 @@ run_epd_1e_1pd() {
|
||||
--port "$PREFILL_DECODE_PORT" \
|
||||
--max-model-len "$MAX_MODEL_LEN" \
|
||||
--enforce-eager \
|
||||
--enable-mm-embeds \
|
||||
--gpu-memory-utilization "$GPU_MEMORY_UTILIZATION" \
|
||||
--enable-request-id-headers \
|
||||
--max-num-seqs "$MAX_NUM_SEQS" \
|
||||
@@ -389,6 +390,7 @@ run_epd_1e_1p_1d() {
|
||||
--port "$PREFILL_PORT" \
|
||||
--max-model-len "$MAX_MODEL_LEN" \
|
||||
--enforce-eager \
|
||||
--enable-mm-embeds \
|
||||
--gpu-memory-utilization "$GPU_MEMORY_UTILIZATION" \
|
||||
--enable-request-id-headers \
|
||||
--max-num-seqs "$MAX_NUM_SEQS" \
|
||||
@@ -415,6 +417,7 @@ run_epd_1e_1p_1d() {
|
||||
--port "$DECODE_PORT" \
|
||||
--max-model-len "$MAX_MODEL_LEN" \
|
||||
--enforce-eager \
|
||||
--enable-mm-embeds \
|
||||
--gpu-memory-utilization "$GPU_MEMORY_UTILIZATION" \
|
||||
--enable-request-id-headers \
|
||||
--max-num-seqs "$MAX_NUM_SEQS" \
|
||||
|
||||
Reference in New Issue
Block a user