[quantization] Decoder output quantization

[stamalakhov]

This PR ensures output of decoder layer is quantized.

Right now output of llama decoder layer is quantized to float.

./ccex test -k quantization.wrapq.wrappers.llama.test_quant_decoder_layer

RUN unit tests with -k quantization.wrapq.wrappers.llama.test_quant_decoder_layer ...
test_dtype_override (quantization.wrapq.wrappers.llama.test_quant_decoder_layer.TestQuantLlamaDecoderLayer) ... ok
test_forward_diff (quantization.wrapq.wrappers.llama.test_quant_decoder_layer.TestQuantLlamaDecoderLayer) ... ok
test_mode_transitions (quantization.wrapq.wrappers.llama.test_quant_decoder_layer.TestQuantLlamaDecoderLayer) ... ok

----------------------------------------------------------------------
Ran 3 tests in 0.054s

OK

Draft: #436
TICO-DCO-1.0-Signed-off-by: s.malakhov s.malakhov@partner.samsung.com

[mhs4670go]

I missed that part when I implement it. Thanks.

[stamalakhov]

I missed that part when I implement it. Thanks.

@mhs4670go
Glad to help.
Shall i provide tests for it?

[mhs4670go]

Shall i provide tests for it?

Ah, how about add a simple test for observers like other wrappers?

[stamalakhov]

Shall i provide tests for it?

Ah, how about add a simple test for observers like other wrappers?

@mhs4670go Tnank you.
I've added test_dtype_override for the only observer created in QuantLlamaDecoderLayer.

[mhs4670go]

LGTM