dacorvo HF Staff commited on
Commit
fc123c5
·
verified ·
1 Parent(s): b1e53a8

Synchronizing local compiler cache.

Browse files
neuronxcc-2.21.33363.0+82129205/MODULE_8508fced3b142be2a2ee+fb4cc044/compile_flags.json ADDED
@@ -0,0 +1 @@
 
 
1
+ ["--target=trn1", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=2 --vectorize-strided-dma ", "-O2", "--lnc=1", "--logfile=/tmp/nxd_model/encoding/_tp0_bk0/log-neuron-cc.txt"]
neuronxcc-2.21.33363.0+82129205/MODULE_8508fced3b142be2a2ee+fb4cc044/model.hlo_module.pb ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2af5485875eb0c496ff7c3ecea5e35812074c14b25db1f04586a6e66b8d50cd7
3
+ size 810056
neuronxcc-2.21.33363.0+82129205/MODULE_8508fced3b142be2a2ee+fb4cc044/model.log ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Failed compilation with ['neuronx-cc', 'compile', '--framework=XLA', '/tmp/nxd_model/encoding/_tp0_bk0/model.MODULE_8508fced3b142be2a2ee+fb4cc044.hlo_module.pb', '--output', '/tmp/nxd_model/encoding/_tp0_bk0/model.MODULE_8508fced3b142be2a2ee+fb4cc044.neff', '--target=trn1', '--auto-cast=none', '--model-type=transformer', '--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=2 --vectorize-strided-dma ', '-O2', '--lnc=1', '--logfile=/tmp/nxd_model/encoding/_tp0_bk0/log-neuron-cc.txt', '--verbose=35']: 2026-02-09T10:25:27Z
2
+ Pre-Partition Pre-Opt Histogram:
3
+ total HLO instructions: 5171
4
+ convert 910 17.60% ################################################################
5
+ reshape 802 15.51% ########################################################
6
+ transpose 723 13.98% ##################################################
7
+ broadcast 548 10.60% ######################################
8
+ slice 543 10.50% ######################################
9
+ multiply 362 7.00% #########################
10
+ parameter 328 6.34% #######################
11
+ constant 221 4.27% ###############
12
+ call 217 4.20% ###############
13
+ dot 181 3.50% ############
14
+ add 144 2.78% ##########
15
+ concatenate 74 1.43% #####
16
+ negate 72 1.39% #####
17
+ get-tuple-element 37 0.72% ##
18
+ iota 3 0.06%
19
+ gather 2 0.04%
20
+ tuple 1 0.02%
21
+ cosine 1 0.02%
22
+ reduce 1 0.02%
23
+ sine 1 0.02%
24
+
25
+
26
+ Pre-Partition Post-Op Histogram:
27
+ total HLO instructions: 4140
28
+ convert 909 21.96% ################################################################
29
+ reshape 650 15.70% #############################################
30
+ transpose 540 13.04% ######################################
31
+ parameter 328 7.92% #######################
32
+ constant 256 6.18% ##################
33
+ broadcast 255 6.16% #################
34
+ slice 252 6.09% #################
35
+ custom-call 217 5.24% ###############
36
+ multiply 217 5.24% ###############
37
+ dot 180 4.35% ############
38
+ add 144 3.48% ##########
39
+ concatenate 74 1.79% #####
40
+ negate 72 1.74% #####
41
+ get-tuple-element 37 0.89% ##
42
+ iota 3 0.07%
43
+ gather 2 0.05%
44
+ cosine 1 0.02%
45
+ tuple 1 0.02%
46
+ reduce 1 0.02%
47
+ sine 1 0.02%
48
+
49
+ Potential split-points stats: #CC 0 #AR 0 #AG 0 #BN 0 nClamp 0
50
+ WARNING: Insufficient number of potential split points found. Entire model will be compiled as a single module.
51
+ No partitions found. Compiling as flat model
52
+ 2026-02-09 10:25:27.146344: F hilo/hlo_passes/NeuronHloVerifier.cc:504] [ERROR] [NCC_VRF007] Tiled instruction count 3784872448 exceeds 5000000. TIP: Input HLO might be too big, please consider using smaller batches, applying model parallelism or compile under --optlevel=1 to create smaller subgraphs
53
+