Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

draft pr about non-streaming output #95

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

draft pr about non-streaming output #95

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
252cf43
only output valid model output
BasicCoder Nov 3, 2023
57de9f5
set non-streaming
BasicCoder Nov 3, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions all_models/inflight_batcher_llm/ensemble/config.pbtxt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -241,6 +241,10 @@ ensemble_scheduling {
key: "TOKENS_BATCH"
value: "_TOKENS_BATCH"
}
input_map {
key: "REQUEST_INPUT_LEN"
value: "_REQUEST_INPUT_LEN"
}
input_map {
key: "SEQUENCE_LENGTH"
value: "_SEQUENCE_LENGTH"
Expand Down
9 changes: 6 additions & 3 deletions all_models/inflight_batcher_llm/postprocessing/1/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -109,6 +109,8 @@ def execute(self, requests):
tokens_batch = pb_utils.get_input_tensor_by_name(
request, 'TOKENS_BATCH').as_numpy()

request_input_lens = pb_utils.get_input_tensor_by_name(
request, 'REQUEST_INPUT_LEN').as_numpy()
# Get sequence length
sequence_lengths = pb_utils.get_input_tensor_by_name(
request, 'SEQUENCE_LENGTH').as_numpy()
Expand All @@ -118,7 +120,7 @@ def execute(self, requests):
# tokens_batch = tokens_batch.T

# Postprocessing output data.
outputs = self._postprocessing(tokens_batch, sequence_lengths)
outputs = self._postprocessing(tokens_batch, request_input_lens, sequence_lengths)

# Create output tensors. You need pb_utils.Tensor
# objects to create pb_utils.InferenceResponse.
Expand Down Expand Up @@ -148,11 +150,12 @@ def finalize(self):
"""
print('Cleaning up...')

def _postprocessing(self, tokens_batch, sequence_lengths):
def _postprocessing(self, tokens_batch, request_input_lens, 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])
request_input_len = request_input_lens[batch_idx].tolist()[0]
output = self.tokenizer.decode(tokens[request_input_len:max(seq_len-1, 1)])
outputs.append(output.encode('utf8'))
return outputs
5 changes: 5 additions & 0 deletions all_models/inflight_batcher_llm/postprocessing/config.pbtxt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,11 @@ input [
data_type: TYPE_INT32
dims: [ -1, -1 ]
},
{
name: "REQUEST_INPUT_LEN"
data_type: TYPE_INT32
dims: [ 1 ]
},
{
name: "SEQUENCE_LENGTH"
data_type: TYPE_INT32
Expand Down
2 changes: 1 addition & 1 deletion all_models/inflight_batcher_llm/tensorrt_llm/config.pbtxt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ backend: "tensorrtllm"
max_batch_size: 128

model_transaction_policy {
decoupled: ${decoupled_mode}
decoupled: False
}

input [
Expand Down
7 changes: 4 additions & 3 deletions tools/inflight_batcher_llm/end_to_end_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,13 +46,13 @@ def test_functionality(client, prompts, output_lens):
]
result = client.infer(model_name, inputs, request_id=str(i))
output0 = result.as_numpy("INPUT_ID")
output1 = result.as_numpy("REQUEST_INPUT_LEN")
request_input_len = result.as_numpy("REQUEST_INPUT_LEN")
output2 = result.as_numpy("REQUEST_OUTPUT_LEN")

model_name = "tensorrt_llm"
inputs = [
utils.prepare_tensor("input_ids", output0, FLAGS.protocol),
utils.prepare_tensor("input_lengths", output1, FLAGS.protocol),
utils.prepare_tensor("input_lengths", request_input_len, FLAGS.protocol),
utils.prepare_tensor("request_output_len", output2,
FLAGS.protocol),
]
Expand All @@ -63,11 +63,12 @@ def test_functionality(client, prompts, output_lens):
model_name = "postprocessing"
inputs = [
utils.prepare_tensor("TOKENS_BATCH", output0, FLAGS.protocol),
utils.prepare_tensor("REQUEST_INPUT_LEN", request_input_len, FLAGS.protocol),
utils.prepare_tensor("SEQUENCE_LENGTH", seq_lengths,
FLAGS.protocol)
]
inputs[0].set_data_from_numpy(output0)
inputs[1].set_data_from_numpy(seq_lengths)
inputs[2].set_data_from_numpy(seq_lengths)

result = client.infer(model_name, inputs, request_id=str(i))
output0 = result.as_numpy("OUTPUT")
Expand Down