Do you mean if we use Python's multi process, you could get ideal result but
use Python's multi thread, you get bad result? Or what else things you mean?
---
[Visit
Topic](https://discuss.tvm.apache.org/t/multithread-threadpool-performance-degradation-when-running-relay-module-in-multiple-
if we change the thread pool to OpenMP, do we have the same problem?
---
[Visit
Topic](https://discuss.tvm.apache.org/t/multithread-threadpool-performance-degradation-when-running-relay-module-in-multiple-threads/10374/6)
to respond.
You are receiving this because you enabled mailing list
[quote="MinminSun, post:1, topic:10374"]
TVM_NUM_THREADS
[/quote]
How about setting one environment `TVM_THREAD_POOL_SPIN_COUNT = 0` , does it
improve your case?
---
[Visit
Topic](https://discuss.tvm.apache.org/t/multithread-threadpool-performance-degradation-when-running-relay-module-in
TVM doesn't provide it currently.
---
[Visit
Topic](https://discuss.tvm.apache.org/t/how-to-manually-controll-cpu-affinity-in-multithreading-scenario/10348/4)
to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click
here](https:/
I assume you don't have your own NPU accelerate library, so you couldn't go TVM
BYOC.
Firstly, you should implement your own quantization algorithm based on your NPU
(not all operation / data type could be provided on your NPU, like int64)
Secondly, you should consider provide your own relay
I think we don't support it.
---
[Visit
Topic](https://discuss.tvm.apache.org/t/tflite-model-with-dynamic-shapes/10012/2)
to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click
here](https://discuss.tvm.apache.org/email/unsubsc
Yeah, it is unfriendly for Ansor. However, I think it is not contradict. We
could not expect we could generate asm like ACL, but we could expect we could
achieve the same optimization. For example, your example is we can not do
`register blocking` optimization easily, but we could expect we ha
When we want to do some advanced optimization like `register blocking` the goal
you want to achieve , TVM codegen can not handle it very well. My experience is
1. write micro gemm like `4x4` or `8x8` and then tensorize 2. try, try and try
different schedule and find one combination to match yo
For the int8, there are intrinsic like DP4A to accelerate. Currently, auto
scheduler doesn't support it.
---
[Visit
Topic](https://discuss.tvm.apache.org/t/auto-scheduler-seems-slower-on-int8/9585/2)
to respond.
You are receiving this because you enabled mailing list mode.
To unsubscrib
Good question. I think we can not get it directly currently. But the mechanism
should have support it. For example, when you pass the output index, we could
use `nodes_[outputs_[index].node_id].name` to get the name.
Something a bit more, I think the name you want more is the model's name, the
I want this feature too and we should have this. One workaround is to pull the
version of tvm code you want and generate the doc by yourself.
@tqchen @comaniac
---
[Visit Topic](https://discuss.tvm.apache.org/t/about-document-version/8459/2)
to respond.
You are receiving this because yo
I don't know why relay drop it, but i think @tqchen @jroesch maybe know more
information and background.
---
[Visit
Topic](https://discuss.tvm.apache.org/t/do-layer-names-get-imported-into-relay/7986/4)
to respond.
You are receiving this because you enabled mailing list mode.
To unsubsc
Previous IR (NNVM) has one name attribute but relay miss it.
---
[Visit
Topic](https://discuss.tvm.apache.org/t/do-layer-names-get-imported-into-relay/7986/2)
to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from these emails, [click
here](https:/
If i remember correctly, we still not support relay e2e on hexagon, right?
@kparzysz
---
[Visit
Topic](https://discuss.tvm.apache.org/t/how-to-run-module-with-hexagon-simulator-sim-dev/7946/2)
to respond.
You are receiving this because you enabled mailing list mode.
To unsubscribe from
The problem is in
```cpp
void Init(const std::string& type_key, const std::string& device_type,
const std::string& platform_name = "");
virtual void Init() { Init("opencl", "gpu"); } // gpu is the device type
```
Inside `Init`, we will only try to match the `gpu` device type, if
15 matches
Mail list logo
