Bf16 gpu support

[richagadgil]

No description provided.

[migraphx-bot]

Test	Batch	Rate new 79e980	Rate old 4b15b6	Diff	Compare
torchvision-resnet50	64	3,252.69	3,257.89	-0.16%	✅
torchvision-resnet50_fp16	64	6,991.11	6,918.72	1.05%	✅
torchvision-densenet121	32	2,436.59	2,432.91	0.15%	✅
torchvision-densenet121_fp16	32	4,062.08	4,086.33	-0.59%	✅
torchvision-inceptionv3	32	1,629.20	1,628.71	0.03%	✅
torchvision-inceptionv3_fp16	32	2,746.80	2,745.51	0.05%	✅
cadene-inceptionv4	16	765.07	764.62	0.06%	✅
cadene-resnext64x4	16	814.32	814.13	0.02%	✅
slim-mobilenet	64	7,467.03	7,458.12	0.12%	✅
slim-nasnetalarge	64	209.04	209.03	0.00%	✅
slim-resnet50v2	64	3,441.73	3,436.59	0.15%	✅
bert-mrpc-onnx	8	1,148.17	1,150.32	-0.19%	✅
bert-mrpc-tf	1	469.27	449.82	4.32%	🔆
pytorch-examples-wlang-gru	1	425.03	437.44	-2.84%	✅
pytorch-examples-wlang-lstm	1	395.93	383.25	3.31%	🔆
torchvision-resnet50_1	1	808.31	741.45	9.02%	🔆
cadene-dpn92_1	1	399.23	400.49	-0.31%	✅
cadene-resnext101_1	1	382.34	382.70	-0.09%	✅
onnx-taau-downsample	1	345.42	345.15	0.08%	✅
dlrm-criteoterabyte	1	33.32	33.31	0.00%	✅
dlrm-criteoterabyte_fp16	1	52.67	52.72	-0.09%	✅
agentmodel	1	8,335.35	8,463.32	-1.51%	✅
unet_fp16	2	58.78	58.77	0.02%	✅
resnet50v1_fp16	1	959.43	934.37	2.68%	✅
resnet50v1_int8	1	1,012.45	1,030.33	-1.74%	✅
bert_base_cased_fp16	64	1,168.98	1,170.28	-0.11%	✅
bert_large_uncased_fp16	32	363.22	363.03	0.05%	✅
bert_large_fp16	1	200.71	199.90	0.40%	✅
distilgpt2_fp16	16	2,199.43	2,197.64	0.08%	✅
yolov5s	1	527.60	518.78	1.70%	✅
tinyllama	1	43.32	43.68	-0.83%	✅
vicuna-fastchat	1	178.34	173.15	3.00%	🔆
whisper-tiny-encoder	1	417.16	417.60	-0.10%	✅
whisper-tiny-decoder	1	434.79	429.03	1.34%	✅

Check results before merge 🔆

[migraphx-bot]

     ✅ bert-mrpc-onnx: PASSED: MIGraphX meets tolerance

     ✅ bert-mrpc-tf: PASSED: MIGraphX meets tolerance

     ✅ pytorch-examples-wlang-gru: PASSED: MIGraphX meets tolerance

     ✅ pytorch-examples-wlang-lstm: PASSED: MIGraphX meets tolerance

     ✅ torchvision-resnet50_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-dpn92_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-resnext101_1: PASSED: MIGraphX meets tolerance

     ✅ dlrm-criteoterabyte: PASSED: MIGraphX meets tolerance

     ✅ agentmodel: PASSED: MIGraphX meets tolerance

     ✅ unet: PASSED: MIGraphX meets tolerance

     ✅ resnet50v1: PASSED: MIGraphX meets tolerance

     ✅ bert_base_cased_fp16: PASSED: MIGraphX meets tolerance

🔴bert_large_uncased_fp16: FAILED: MIGraphX is not within tolerance - check verbose output

     ✅ bert_large: PASSED: MIGraphX meets tolerance

     ✅ yolov5s: PASSED: MIGraphX meets tolerance

     ✅ tinyllama: PASSED: MIGraphX meets tolerance

     ✅ vicuna-fastchat: PASSED: MIGraphX meets tolerance

     ✅ whisper-tiny-encoder: PASSED: MIGraphX meets tolerance

     ✅ whisper-tiny-decoder: PASSED: MIGraphX meets tolerance

     ✅ distilgpt2_fp16: PASSED: MIGraphX meets tolerance

[shivadbhavsar]

I'll stop touching this now, with the pooling update it now passes verify for resnet50 with no extra flags. I'll try to run a few more models to see if there are glaring issues.