[frontend][tflite] float16 quant support #7736
Conversation
merge upstream main
| ) | ||
| for fp16_quantized in [False, True]: | ||
| if quantized and fp16_quantized: | ||
| continue |
There was a problem hiding this comment.
Why are we skipping the fp16_quantized test
There was a problem hiding this comment.
fp16 quantized path doesn't go through quantized path.
quantized=True and fp16_quantzied=True goes through same code as quantized=False and fp16_quantize=True. So I just wanted to reduce the test time here.
There was a problem hiding this comment.
Should I remove the skipping code?
There was a problem hiding this comment.
Sorry for the delay. Yes, please. Can you restructure the test such that there is no need of skipping code. One might be confused while reading the code otherwise.
There was a problem hiding this comment.
Sorry for the delay. Yes, please. Can you restructure the test such that there is no need of skipping code. One might be confused while reading the code otherwise.
There was a problem hiding this comment.
I have addressed your request.
|
@euntaik What are the input and output dtypes for the float16 quantized graph? I do not see any |
Cast(float16 to float32) is already done in the Dequantize node and the resulting expression is passed to the FullyConnected op as weight and bias. |
|
Thanks @euntaik This is merged! |
* [frontend][tflite] float16 quant support * remove skip conditions in tests
* [frontend][tflite] float16 quant support * remove skip conditions in tests
* [frontend][tflite] float16 quant support * remove skip conditions in tests
* [frontend][tflite] float16 quant support * remove skip conditions in tests
* [frontend][tflite] float16 quant support * remove skip conditions in tests
add float16 quant support for fc and transpose_conv