Comments (2)
The input/output shape is not correct:
The correct shape is like the following (batch_size=1, hidden_size=2048, num_heads=16, head_dim=128 for this model):
input_embds: (batch_size, seq_len, hidden_size)
past_key_value: (2, batch_size, past_seq_len, num_heads, head_dim)
attention_mask: (batch_size, 1, seq_len, past_seq_len + seq_len)
hideen_state: (batch_size, seq_len, hidden_size)
presents: (2, batch_size, past_seq_len + seq_len, num_heads, head_dim)
from onnxruntime.
The problems seems to be with the past_key_values, it only allocates a static shape of buffer, and next iteration, it tries to reuse it, but found shape mismatch and raise it.
from onnxruntime.
Related Issues (20)
- [DO NOT UNPIN] ORT 1.19 release candidates available for testing HOT 2
- [Mobile] Setting the namespace via the package attribute in the source AndroidManifest.xml is no longer supported
- FAIL : Failed to load library libonnxruntime_providers_cuda.so with error: libcublasLt.so.11: cannot open shared object file: No such file or directory HOT 2
- device_id is not work
- [Performance] Inference time discrepancy when using TorchScript vs ONNX exported model
- onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : CopyTensorAsync is not implemented HOT 3
- [Build] ONNXRuntime QNN execution provider dlls are not copied to target
- [Feature Request] Enable --use-extensions for pre built releases
- Inference session with 'optimized_model_filepath' expands HardSwish operator into atomic operators
- [Performance] SetIntraOpNumThreads not take effect HOT 2
- [Performance] pytorch quantize_qat model export to onnx, insert a transpose layer befor and after the conv layers HOT 5
- [Web] model can run under python env both cpu and cuda, but cannot run under webGPU HOT 6
- 1.18.1 regression: not working with Emgu.CV <= 4.8.0 HOT 2
- [Build] Access Violation error on using 'new InferenceSession(filepath)' in 1.18.1
- Why is CoreMLExecution an Objective-C class HOT 1
- Question
- [Bug]: Onnxruntime.CPU memoty leaks
- onnxruntime-directml import interference with sklearn
- ORTModelForSeq2SeqLM.from_pretrained can not use provider=['CUDAExecutionProvider','CPUExecutionProvider'] HOT 4
- Inferencing FP16 model using onnxruntime
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from onnxruntime.