Skip to content

GitLab

Unverified Commit 45e38d3d authored by Michael Gathara's avatar Michael Gathara Committed by GitHub
Browse files

feat: HF_ENDPOINT addition (#2637)

parent 889d6529
Hide whitespace changes
Inline Side-by-side
Showing with 8 additions and 8 deletions
deploy/cloud/operator/internal/dynamo/backend_trtllm.go
View file @ 45e38d3d
...@@ -188,7 +188,7 @@ func getGPUsPerNode(resources *common.Resources) int32 { ...@@ -188,7 +188,7 @@ func getGPUsPerNode(resources *common.Resources) int32 {
// getCommonTRTLLMEnvVars returns a map of common environment variables for TRTLLM deployments // getCommonTRTLLMEnvVars returns a map of common environment variables for TRTLLM deployments
func getCommonTRTLLMEnvVars() map[string]bool { func getCommonTRTLLMEnvVars() map[string]bool {
return map[string]bool{ return map[string]bool{
"CUDA_VISIBLE_DEVICES": true, "MODEL_PATH": true, "HF_TOKEN": true, "HUGGING_FACE_HUB_TOKEN": true, "CUDA_VISIBLE_DEVICES": true, "MODEL_PATH": true, "HF_TOKEN": true, "HUGGING_FACE_HUB_TOKEN": true, "HF_ENDPOINT": true,
"TOKENIZERS_PARALLELISM": true, "NCCL_DEBUG": true, "NCCL_IB_DISABLE": true, "NCCL_P2P_DISABLE": true, "TOKENIZERS_PARALLELISM": true, "NCCL_DEBUG": true, "NCCL_IB_DISABLE": true, "NCCL_P2P_DISABLE": true,
"TENSORRT_LLM_CACHE_DIR": true, "HF_HOME": true, "TRANSFORMERS_CACHE": true, "HF_DATASETS_CACHE": true, "TENSORRT_LLM_CACHE_DIR": true, "HF_HOME": true, "TRANSFORMERS_CACHE": true, "HF_DATASETS_CACHE": true,
"PATH": true, "LD_LIBRARY_PATH": true, "PYTHONPATH": true, "HOME": true, "USER": true, "PATH": true, "LD_LIBRARY_PATH": true, "PYTHONPATH": true, "HOME": true, "USER": true,
......
deploy/cloud/operator/internal/dynamo/backend_trtllm_test.go
View file @ 45e38d3d
...@@ -60,7 +60,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) { ...@@ -60,7 +60,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) {
{Name: commonconsts.MpiRunSshSecretName, MountPath: "/ssh-pk", ReadOnly: true}, {Name: commonconsts.MpiRunSshSecretName, MountPath: "/ssh-pk", ReadOnly: true},
}, },
expectedCommand: []string{"/bin/sh", "-c"}, expectedCommand: []string{"/bin/sh", "-c"},
expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 6 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-1.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python3 --model test'"}, expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 6 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-1.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python3 --model test'"},
expectedEnv: []corev1.EnvVar{ expectedEnv: []corev1.EnvVar{
{Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"}, {Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"},
}, },
...@@ -116,7 +116,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) { ...@@ -116,7 +116,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) {
{Name: commonconsts.MpiRunSshSecretName, MountPath: "/ssh-pk", ReadOnly: true}, {Name: commonconsts.MpiRunSshSecretName, MountPath: "/ssh-pk", ReadOnly: true},
}, },
expectedCommand: []string{"/bin/sh", "-c"}, expectedCommand: []string{"/bin/sh", "-c"},
expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${LWS_LEADER_ADDRESS},${LWS_WORKER_1_ADDRESS} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python3 --model test'"}, expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${LWS_LEADER_ADDRESS},${LWS_WORKER_1_ADDRESS} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python3 --model test'"},
expectedEnv: []corev1.EnvVar{ expectedEnv: []corev1.EnvVar{
{Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"}, {Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"},
}, },
...@@ -563,7 +563,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -563,7 +563,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"python3", "--model", "test"}, initialArgs: []string{"python3", "--model", "test"},
initialCommand: []string{}, initialCommand: []string{},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 6 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-1.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python3 --model test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 6 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-service-wkr-1.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python3 --model test'",
}, },
{ {
name: "Leader with command and no GPU resources", name: "Leader with command and no GPU resources",
...@@ -573,7 +573,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -573,7 +573,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
component: &v1alpha1.DynamoComponentDeploymentOverridesSpec{}, component: &v1alpha1.DynamoComponentDeploymentOverridesSpec{},
initialArgs: []string{}, initialArgs: []string{},
initialCommand: []string{"python", "-m", "worker"}, initialCommand: []string{"python", "-m", "worker"},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 0 -H ${LWS_LEADER_ADDRESS},${LWS_WORKER_1_ADDRESS} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python -m worker'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 0 -H ${LWS_LEADER_ADDRESS},${LWS_WORKER_1_ADDRESS} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch python -m worker'",
}, },
{ {
name: "Leader with both command and args (args take precedence)", name: "Leader with both command and args (args take precedence)",
...@@ -591,7 +591,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -591,7 +591,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"launch", "--config", "test.yaml"}, initialArgs: []string{"launch", "--config", "test.yaml"},
initialCommand: []string{"ignored-command"}, initialCommand: []string{"ignored-command"},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-wkr-0.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch launch --config test.yaml'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-wkr-0.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch launch --config test.yaml'",
}, },
{ {
name: "Leader with all environment variables forwarded", name: "Leader with all environment variables forwarded",
...@@ -609,7 +609,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -609,7 +609,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"serve", "--model", "test"}, initialArgs: []string{"serve", "--model", "test"},
initialCommand: []string{}, initialCommand: []string{},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-wkr-0.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch serve --model test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-wkr-0.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch serve --model test'",
}, },
{ {
name: "Leader with overlapping environment variables (deduplication test)", name: "Leader with overlapping environment variables (deduplication test)",
...@@ -627,7 +627,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -627,7 +627,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"serve", "--model", "test"}, initialArgs: []string{"serve", "--model", "test"},
initialCommand: []string{}, initialCommand: []string{},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-wkr-0.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x CUSTOM_VAR -x HF_DATASETS_CACHE -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch serve --model test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --oversubscribe -n 2 -H ${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-ldr-0.${GROVE_HEADLESS_SERVICE},${GROVE_PCSG_NAME}-${GROVE_PCSG_INDEX}-test-wkr-0.${GROVE_HEADLESS_SERVICE} --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x CUSTOM_VAR -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch serve --model test'",
}, },
} }
......
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