Update halo-models.md

nod-ai · Dec 4, 2024 · 06cf14a · 06cf14a
1 parent a2a0c03
commit 06cf14a
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/halo-models.md b/halo-models.md
@@ -107,7 +107,7 @@ iree-compile 405b_f16_tp8_decomposed.mlir --iree-hip-target=gfx942 --iree-hal-ta
 (MI300X GPU, SPX Mode)
 |Item                                      | Generate MLIR | Compile to vmfb | IREE invocation | IREE numeric | Serving numeric |
 |------------------------------------------|---------------|-----------------|-----------------|--------------|-----------------|
-| llama3.1-8B-FP16 (prefill)     |PASS [mlir](https://sharkpublic.blob.core.windows.net/sharkpublic/halo-models/llm-dev/llama3_8b/8b_f16_nondecomposed_prefill_12_3.mlir)   | PASS [compile command](https://gist.github.com/aviator19941/d49e0baca5c6f64a93f8926f99cd111f) | PASS [run command](https://gist.github.com/aviator19941/8cc7b1282e69ec87fc4c5e231d28b80f) [numpy inputs](https://gist.github.com/aviator19941/fe1371e45dc5d28651145884df0b314b) | tbd | tbd
+| llama3.1-8B-FP16 (prefill)     |PASS [mlir](https://sharkpublic.blob.core.windows.net/sharkpublic/halo-models/llm-dev/llama3_8b/8b_f16_nondecomposed_prefill_12_3.mlir) [irpa](https://sharkblobs.blob.core.windows.net/halo-models/llm-dev/llama3_8b/8b_f16.irpa)  | PASS [compile command](https://gist.github.com/aviator19941/d49e0baca5c6f64a93f8926f99cd111f) | PASS [run command](https://gist.github.com/aviator19941/8cc7b1282e69ec87fc4c5e231d28b80f) [numpy inputs](https://gist.github.com/aviator19941/fe1371e45dc5d28651145884df0b314b) | tbd | tbd
 | llama3.1-8B-FP16      |PASS [mlir](https://sharkblobs.blob.core.windows.net/halo-models/llm-dev/llama3_8b/8b_f16_nondecomposed.mlir)   | Fails in iree, [patch](https://github.com/iree-org/iree/pull/18890) | tbd | tbd | tbd
 | llama3.1-70B-FP16      |PASS [mlir](https://sharkblobs.blob.core.windows.net/halo-models/llm-dev/llama3_70b/70b_f16_nondecomposed.mlir)   |Fails in iree, [patch](https://github.com/iree-org/iree/pull/18890) | tbd | tbd | tbd
 | llama3.1-405B-FP16  |  PASS [mlir_tp8](https://sharkpublic.blob.core.windows.net/sharkpublic/halo-models/llm-dev/llama3_405b/405b_f16_tp8_nondecomposed_bs4.mlir) | PASS | FAIL [OOM](https://github.com/iree-org/iree/issues/18864) | tbd | tbd