You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Triton Information
GPU Nvidia A10G
Cuda version 12.3
Driver version 535.183.01
TensorRT-LLM v0.8.0
Image nvcr.io/nvidia/tritonserver:24.02-trtllm-python-py3 (was used to build the tensorrt engine and start triton inference server)
Model meta-llm/Meta-Llama-Guard-2-8B
OS: Ubuntu
Are you using the Triton container or did you build it yourself?
To Reproduce
convert model checkpoint to tensorrt format: python3 convert_checkpoint.py ...
build tensorrt engine: trtllm-build ...
run inference with tensorrt engine: python3 run.py ...
create trt model repo based on .../tensorrtllm_backend_/all_models/inflight_batcher_llm/
# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
# * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# * Neither the name of NVIDIA CORPORATION nor the names of its
# contributors may be used to endorse or promote products derived
# from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
import json
import numpy as np
import triton_python_backend_utils as pb_utils
from transformers import AutoTokenizer, LlamaTokenizer, T5Tokenizer
class TritonPythonModel:
"""Your Python model must use the same class name. Every Python model
that is created must have "TritonPythonModel" as the class name.
"""
def initialize(self, args):
"""`initialize` is called only once when the model is being loaded.
Implementing `initialize` function is optional. This function allows
the model to initialize any state associated with this model.
Parameters
----------
args : dict
Both keys and values are strings. The dictionary keys and values are:
* model_config: A JSON string containing the model configuration
* model_instance_kind: A string containing model instance kind
* model_instance_device_id: A string containing model instance device ID
* model_repository: Model repository path
* model_version: Model version
* model_name: Model name
"""
# Parse model configs
model_config = json.loads(args['model_config'])
tokenizer_dir = model_config['parameters']['tokenizer_dir'][
'string_value']
tokenizer_type = model_config['parameters']['tokenizer_type'][
'string_value']
self.skip_special_tokens = model_config['parameters'].get(
'skip_special_tokens',
{'string_value': "true"})['string_value'].lower() in [
'true', '1', 't', 'y', 'yes'
]
if tokenizer_type == 't5':
self.tokenizer = T5Tokenizer(vocab_file=tokenizer_dir,
padding_side='left')
elif tokenizer_type == 'auto':
self.tokenizer = AutoTokenizer.from_pretrained(
tokenizer_dir, padding_side='left', trust_remote_code=True)
elif tokenizer_type == 'llama':
self.tokenizer = LlamaTokenizer.from_pretrained(
tokenizer_dir, legacy=False, padding_side='left')
else:
raise AttributeError(
f'Unexpected tokenizer type: {tokenizer_type}')
self.tokenizer.pad_token = self.tokenizer.eos_token
# Parse model output configs
output_config = pb_utils.get_output_config_by_name(
model_config, "OUTPUT")
# Convert Triton types to numpy types
self.output_dtype = pb_utils.triton_string_to_numpy(
output_config['data_type'])
def execute(self, requests):
"""`execute` must be implemented in every Python model. `execute`
function receives a list of pb_utils.InferenceRequest as the only
argument. This function is called when an inference is requested
for this model. Depending on the batching configuration (e.g. Dynamic
Batching) used, `requests` may contain multiple requests. Every
Python model, must create one pb_utils.InferenceResponse for every
pb_utils.InferenceRequest in `requests`. If there is an error, you can
set the error argument when creating a pb_utils.InferenceResponse.
Parameters
----------
requests : list
A list of pb_utils.InferenceRequest
Returns
-------
list
A list of pb_utils.InferenceResponse. The length of this list must
be the same as `requests`
"""
responses = []
# Every Python backend must iterate over everyone of the requests
# and create a pb_utils.InferenceResponse for each of them.
for idx, request in enumerate(requests):
# Get input tensors
tokens_batch = pb_utils.get_input_tensor_by_name(
request, 'TOKENS_BATCH').as_numpy()
# Get sequence length
sequence_lengths = pb_utils.get_input_tensor_by_name(
request, 'SEQUENCE_LENGTH').as_numpy()
# Get cum log probs
cum_log_probs = pb_utils.get_input_tensor_by_name(
request, 'CUM_LOG_PROBS').as_numpy()
# Get sequence length
output_log_probs = pb_utils.get_input_tensor_by_name(
request, 'OUTPUT_LOG_PROBS').as_numpy()
# Get context logits
context_logits = pb_utils.get_input_tensor_by_name(
request, 'CONTEXT_LOGITS').as_numpy()
# Get generation logits
generation_logits = pb_utils.get_input_tensor_by_name(
request, 'GENERATION_LOGITS').as_numpy()
# Reshape Input
# tokens_batch = tokens_batch.reshape([-1, tokens_batch.shape[0]])
# tokens_batch = tokens_batch.T
# Postprocessing output data.
outputs = self._postprocessing(tokens_batch, sequence_lengths)
# Create output tensors. You need pb_utils.Tensor
# objects to create pb_utils.InferenceResponse.
output_tensor = pb_utils.Tensor(
'OUTPUT',
np.array(outputs).astype(self.output_dtype))
out_cum_log_probs = pb_utils.Tensor('OUT_CUM_LOG_PROBS',
cum_log_probs)
out_output_log_probs = pb_utils.Tensor('OUT_OUTPUT_LOG_PROBS',
output_log_probs)
out_context_logits = pb_utils.Tensor('OUT_CONTEXT_LOGITS',
context_logits)
out_generation_logits = pb_utils.Tensor('OUT_GENERATION_LOGITS',
generation_logits)
# Create InferenceResponse. You can set an error here in case
# there was a problem with handling this inference request.
# Below is an example of how you can set errors in inference
# response:
#
# pb_utils.InferenceResponse(
# output_tensors=..., TritonError("An error occurred"))
inference_response = pb_utils.InferenceResponse(output_tensors=[
output_tensor, out_cum_log_probs, out_output_log_probs,
out_context_logits, out_generation_logits
])
responses.append(inference_response)
# You should return a list of pb_utils.InferenceResponse. Length
# of this list must match the length of `requests` list.
return responses
def finalize(self):
"""`finalize` is called only once when the model is being unloaded.
Implementing `finalize` function is optional. This function allows
the model to perform any necessary clean ups before exit.
"""
print('Cleaning up...')
def _postprocessing(self, tokens_batch, sequence_lengths):
outputs = []
for batch_idx, beam_tokens in enumerate(tokens_batch):
for beam_idx, tokens in enumerate(beam_tokens):
seq_len = sequence_lengths[batch_idx][beam_idx]
output = self.tokenizer.decode(
tokens[:seq_len],
skip_special_tokens=self.skip_special_tokens)
outputs.append(output.encode('utf8'))
return outputs
# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
# * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# * Neither the name of NVIDIA CORPORATION nor the names of its
# contributors may be used to endorse or promote products derived
# from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
name: "postprocessing"
backend: "python"
max_batch_size: 64
input [
{
name: "TOKENS_BATCH"
data_type: TYPE_INT32
dims: [ -1, -1 ]
},
{
name: "SEQUENCE_LENGTH"
data_type: TYPE_INT32
dims: [ -1 ]
},
{
name: "CUM_LOG_PROBS"
data_type: TYPE_FP32
dims: [ -1 ]
},
{
name: "OUTPUT_LOG_PROBS"
data_type: TYPE_FP32
dims: [ -1, -1 ]
},
{
name: "CONTEXT_LOGITS"
data_type: TYPE_FP32
dims: [ -1, -1 ]
optional: true
},
{
name: "GENERATION_LOGITS"
data_type: TYPE_FP32
dims: [ -1, -1, -1 ]
optional: true
}
]
output [
{
name: "OUTPUT"
data_type: TYPE_STRING
dims: [ -1 ]
},
{
name: "OUT_CUM_LOG_PROBS"
data_type: TYPE_FP32
dims: [ -1 ]
},
{
name: "OUT_OUTPUT_LOG_PROBS"
data_type: TYPE_FP32
dims: [ -1, -1 ]
},
{
name: "OUT_CONTEXT_LOGITS"
data_type: TYPE_FP32
dims: [ -1, -1 ]
},
{
name: "OUT_GENERATION_LOGITS"
data_type: TYPE_FP32
dims: [ -1, -1, -1 ]
}
]
parameters {
key: "tokenizer_dir"
value: {
string_value: "/tensorrt/models/Meta-Llama-Guard-2-8B"
}
}
parameters {
key: "tokenizer_type"
value: {
string_value: "llama"
}
}
parameters {
key: "skip_special_tokens"
value: {
string_value: "True"
}
}
instance_group [
{
count: 1
kind: KIND_CPU
}
]
Expected behavior
no error
The text was updated successfully, but these errors were encountered:
Description
I currently use model: https://huggingface.co/meta-llm/Meta-Llama-Guard-2-8B and get error
UNAVAILABLE: Internal: TypeError: expected str, bytes or os.PathLike object, not NoneType
- please see debug_triton_inference_server.txtwhen i use a different model like https://huggingface.co/Trendyol/Trendyol-LLM-7b-chat-v1.0 i dont see the error...
Triton Information
GPU Nvidia A10G
Cuda version 12.3
Driver version 535.183.01
TensorRT-LLM v0.8.0
Image nvcr.io/nvidia/tritonserver:24.02-trtllm-python-py3 (was used to build the tensorrt engine and start triton inference server)
Model meta-llm/Meta-Llama-Guard-2-8B
OS: Ubuntu
Are you using the Triton container or did you build it yourself?
To Reproduce
logs:
debug_triton_inference_server.txt
cat /tensorrt/triton-repos/trt-Meta-Llama-Guard-2-8B/postprocessing/1/model.py
cat /tensorrt/triton-repos/trt-Meta-Llama-Guard-2-8B/postprocessing/config.pbtxt
Expected behavior
no error
The text was updated successfully, but these errors were encountered: