a-sidorova

# Pull the docker image
docker pull rocm/vllm:rocm7.0.0_vllm_0.11.1_20251103

# Run docker
docker run -it --name=serving-rocm7.0.0_vllm_0.11.1_20251103 --network=host --group-add=video --ipc=host --cap-add=SYS_PTRACE --security-opt seccomp=unconfined --device /dev/kfd --device /dev/dri -v /root/:/workspace rocm7.0.0_vllm_0.11.1_20251103

# If it was started stopped
docker start serving-rocm7.0.0_vllm_0.11.1_20251103 

# Attach to the docker 

a-sidorova

iree-compile conv_wgrad_asm_emitter_dy_nhwc_x_nhwc_grouped_strided.mlir -iree-hal-target-backends=rocm --iree-hip-target=gfx942 --iree-opt-level=O3 --iree-preprocessing-pass-pipeline="builtin.module(util.func(iree-preprocessing-sink-transpose-through-pad))" --iree-dispatch-creation-enable-fuse-padding-into-linalg-consumer-ops -o iree-compile-output.vmfb --mlir-print-ir-after-all --mlir-disable-threading 2> dump.log

a-sidorova

module @module {
  func.func @main(%dw_: !torch.tensor<[32,4,1,1],f32>, %dy: !torch.vtensor<[4,32,8,8],f32>, %x: !torch.vtensor<[4,16,8,8],f32>) attributes {torch.assume_strict_symbolic_shapes} {
  
    %bias_conv_wgrad = torch.constant.none
    %transposed_conv_wgrad = torch.constant.bool false
    %output_padding_conv_wgrad = torch.prim.ListConstruct  : () -> !torch.list<int>
    %groups_conv_wgrad = torch.constant.int 4
    %stride_val_0_conv_wgrad = torch.constant.int 1
    %stride_val_1_conv_wgrad = torch.constant.int 1
    %stride_conv_wgrad = torch.prim.ListConstruct %stride_val_0_conv_wgrad, %stride_val_1_conv_wgrad : (!torch.int, !torch.int) -> !torch.list<int>

a-sidorova

Model list (all of them can be found on HuggingFace).

BERT models:
- ahotrod/electra_large_discriminator_squad2_512
- batterydata/batterybert-cased-squad-v1
- cardiffnlp/twitter-roberta-base-emotion
- cardiffnlp/twitter-roberta-base-sentiment
- cardiffnlp/twitter-roberta-base-sentiment-latest
- cardiffnlp/twitter-xlm-roberta-base-sentiment
- cross-encoder/quora-distilroberta-base