Issues
- 23
[INT4] Compress model by quantizing weights to int4
#3307 opened by umangyadav - 0
Onnx support: Parse int4 constants in onnx by inserting the packed buffer and an unpack operator.
#3374 opened by lakhinderwalia - 7
Segmentation fault - MobileNetV2
#3355 opened by CharlieL7 - 0
JIRA Ticket: GLIBCXX_3.4.30 not found on Navi48 system
#3405 opened by ahsan-ca - 0
Update MIGraphX FP8 to support OCP types
#3393 opened by CharlieL7 - 0
Improve performance of GQA sub-operators on GPU
#3396 opened by turneram - 0
- 0
Debug end-to-end accuracy of Llama2
#3394 opened by turneram - 1
Integrate codegen API for CK gemm-multiple-d
#3261 opened by turneram - 0
Support microsoft contrib ONNX operator MatMulNBits
#3390 opened by hgaspar - 1
benchmark: No valid tuned compilation for fused_reduce with <no problem key>
#3347 opened by shivadbhavsar - 0
Verification failure with bert_large model
#3388 opened by djramic - 0
Refactor mlir rewrites into single function
#3384 opened by umangyadav - 0
Investigate verification failure of MLIR + ReduceSum test
#3382 opened by umangyadav - 0
Bump rocMLIR SHA with Navi3x Accuracy Fix
#3266 opened by umangyadav - 0
Add gemm_pointwise optimization for hipblaslt
#3372 opened by ahsan-ca - 2
Bump CI to 6.2 once released
#3318 opened by causten - 0
Fix implicit assumption for inputs of the consumer op when fusing MLIR ops
#3306 opened by umangyadav - 11
Reduce compile time by reducing calls to `compute_shape()` for each IR transformation
#3233 opened by umangyadav - 0
Pooling Improvements
#3367 opened by umangyadav - 1
Fix rewrite_reduce pass when running SD3 mmdit
#3303 opened by kahmed10 - 0
Add signed int4 support for pack_int4 operator. Currently only supports unsigned.
#3358 opened by lakhinderwalia - 0
Automate Stable Diffusion 2.1 Model
#3257 opened by kahmed10 - 2
Generate rocm/Onnxruntime for 6.2 release
#3348 opened by TedThemistokleous - 2
[Issue]: Investigate and Fix GPU error with int8 reduced layer models
#3298 opened by TedThemistokleous - 0
Consider `ceil_mode` when doing `rewrite_pooling` pass
#3356 opened by umangyadav - 2
Remove `qlinear_reused` matcher and instead fuse MLIR `quant_dot` with base pointwise operators
#3269 opened by CharlieL7 - 1
Introduce `--int4-weights` option in `migraphx-driver`. This would require changes in MIGraphX's naive quantizer to set range between `[0, 15]` During quantization it should also insert "pack" and "unpack" instructions.
#3341 opened by lakhinderwalia - 0
Enable MultiStream execution on DLM based models
#3339 opened by causten - 0
[BUG] Buffer shape not the same as Output Shape
#3334 opened by richagadgil - 0
Blog Post for Stable Diffusion Models
#3330 opened by richagadgil - 0
`test_add_layernorm_add_gemm_nonstd` fails to compile when LayerNorm fusions are disabled
#3320 opened by umangyadav - 0
Investigate rocMLIR extended PR Check Failures
#3317 opened by umangyadav - 0
- 3
- 0
UInt64 Overflow with Higher Batch Sizes
#3310 opened by eddieliao - 0
Create a static output shape verify of OneHot
#3279 opened by CharlieL7 - 3
- 0
- 1
Optimize Dot + Slice
#3267 opened by umangyadav - 0
Make `enable_splitk_for_tuning` unit attr
#3254 opened by umangyadav - 1
Pooling JIT kernel causes random perf drop and end to end performance is low
#3232 opened by umangyadav - 0
ResNet34 Perf
#3241 opened by shivadbhavsar - 0
[Documentation]: Integration Onboarding
#3252 opened by shivadbhavsar - 0
- 0
- 0
Bug in find_concat_op with predicate func
#3245 opened by shivadbhavsar - 0
Padding as fusion
#3239 opened by hgaspar - 0
Symbolic shapes
#3238 opened by hgaspar - 0
Find a way to test JIT pooling kernel
#3234 opened by umangyadav