Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:23:52+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:23:52+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:24:06+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:24:06+0900 attempt 1/10: start: env /home/devops/Development/pre-epic-4c44e317/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:24:07+0900 sample 1: clean gpu=[]
2026-10-12T10:24:08+0900 sample 2: clean gpu=[17416:mlxcel-bench-de]
2026-10-12T10:24:09+0900 sample 3: clean gpu=[17416:mlxcel-bench-de]
2026-10-12T10:24:11+0900 sample 4: clean gpu=[17416:mlxcel-bench-de]
2026-10-12T10:24:12+0900 sample 5: clean gpu=[17416:mlxcel-bench-de]
2026-10-12T10:24:13+0900 monitor: 5 samples
2026-10-12T10:24:13+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,12 @@
2026-10-12T10:22:31+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:22:31+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:22:45+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:22:45+0900 attempt 1/10: start: env /home/devops/Development/pre-epic-4c44e317/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:22:45+0900 sample 1: clean gpu=[]
2026-10-12T10:22:47+0900 sample 2: clean gpu=[4171159:mlxcel-bench-de]
2026-10-12T10:22:48+0900 sample 3: clean gpu=[4171159:mlxcel-bench-de]
2026-10-12T10:22:50+0900 sample 4: clean gpu=[4171159:mlxcel-bench-de]
2026-10-12T10:22:51+0900 sample 5: clean gpu=[4171159:mlxcel-bench-de]
2026-10-12T10:22:53+0900 sample 6: clean gpu=[4171159:mlxcel-bench-de]
2026-10-12T10:22:54+0900 monitor: 6 samples
2026-10-12T10:22:54+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:23:25+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:23:25+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:23:40+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:23:40+0900 attempt 1/10: start: env MLXCEL_FORCE_SYNC=1 /home/devops/Development/wt-20261006T1212-issue-2239/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:23:40+0900 sample 1: clean gpu=[]
2026-10-12T10:23:41+0900 sample 2: clean gpu=[4211:mlxcel-bench-de]
2026-10-12T10:23:43+0900 sample 3: clean gpu=[4211:mlxcel-bench-de]
2026-10-12T10:23:44+0900 sample 4: clean gpu=[4211:mlxcel-bench-de]
2026-10-12T10:23:46+0900 sample 5: clean gpu=[4211:mlxcel-bench-de]
2026-10-12T10:23:47+0900 monitor: 5 samples
2026-10-12T10:23:47+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:22:59+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:22:59+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:23:13+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:23:13+0900 attempt 1/10: start: env /home/devops/Development/wt-20261006T1212-issue-2239/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:23:13+0900 sample 1: clean gpu=[]
2026-10-12T10:23:15+0900 sample 2: clean gpu=[4184888:mlxcel-bench-de]
2026-10-12T10:23:16+0900 sample 3: clean gpu=[4184888:mlxcel-bench-de]
2026-10-12T10:23:18+0900 sample 4: clean gpu=[4184888:mlxcel-bench-de]
2026-10-12T10:23:19+0900 sample 5: clean gpu=[4184888:mlxcel-bench-de]
2026-10-12T10:23:20+0900 monitor: 5 samples
2026-10-12T10:23:20+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:25:12+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:25:12+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:25:26+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:25:26+0900 attempt 1/10: start: env /home/devops/Development/pre-epic-4c44e317/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:25:27+0900 sample 1: clean gpu=[]
2026-10-12T10:25:28+0900 sample 2: clean gpu=[57157:mlxcel-bench-de]
2026-10-12T10:25:30+0900 sample 3: clean gpu=[57157:mlxcel-bench-de]
2026-10-12T10:25:31+0900 sample 4: clean gpu=[57157:mlxcel-bench-de]
2026-10-12T10:25:32+0900 sample 5: clean gpu=[57157:mlxcel-bench-de]
2026-10-12T10:25:33+0900 monitor: 5 samples
2026-10-12T10:25:33+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:25:38+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:25:38+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:25:53+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:25:53+0900 attempt 1/10: start: env /home/devops/Development/pre-epic-4c44e317/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:25:53+0900 sample 1: clean gpu=[]
2026-10-12T10:25:55+0900 sample 2: clean gpu=[70408:mlxcel-bench-de]
2026-10-12T10:25:56+0900 sample 3: clean gpu=[70408:mlxcel-bench-de]
2026-10-12T10:25:58+0900 sample 4: clean gpu=[70408:mlxcel-bench-de]
2026-10-12T10:25:59+0900 sample 5: clean gpu=[70408:mlxcel-bench-de]
2026-10-12T10:26:00+0900 monitor: 5 samples
2026-10-12T10:26:00+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:24:45+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:24:45+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:24:59+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:24:59+0900 attempt 1/10: start: env MLXCEL_FORCE_SYNC=1 /home/devops/Development/wt-20261006T1212-issue-2239/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:25:00+0900 sample 1: clean gpu=[]
2026-10-12T10:25:01+0900 sample 2: clean gpu=[43911:mlxcel-bench-de]
2026-10-12T10:25:03+0900 sample 3: clean gpu=[43911:mlxcel-bench-de]
2026-10-12T10:25:04+0900 sample 4: clean gpu=[43911:mlxcel-bench-de]
2026-10-12T10:25:06+0900 sample 5: clean gpu=[43911:mlxcel-bench-de]
2026-10-12T10:25:07+0900 monitor: 5 samples
2026-10-12T10:25:07+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:24:18+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:24:18+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:24:33+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:24:33+0900 attempt 1/10: start: env /home/devops/Development/wt-20261006T1212-issue-2239/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:24:33+0900 sample 1: clean gpu=[]
2026-10-12T10:24:35+0900 sample 2: clean gpu=[30667:mlxcel-bench-de]
2026-10-12T10:24:36+0900 sample 3: clean gpu=[30667:mlxcel-bench-de]
2026-10-12T10:24:38+0900 sample 4: clean gpu=[30667:mlxcel-bench-de]
2026-10-12T10:24:39+0900 sample 5: clean gpu=[30667:mlxcel-bench-de]
2026-10-12T10:24:40+0900 monitor: 5 samples
2026-10-12T10:24:40+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:26:32+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:26:32+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:26:46+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:26:46+0900 attempt 1/10: start: env /home/devops/Development/pre-epic-4c44e317/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:26:47+0900 sample 1: clean gpu=[]
2026-10-12T10:26:48+0900 sample 2: clean gpu=[96908:mlxcel-bench-de]
2026-10-12T10:26:50+0900 sample 3: clean gpu=[96908:mlxcel-bench-de]
2026-10-12T10:26:51+0900 sample 4: clean gpu=[96908:mlxcel-bench-de]
2026-10-12T10:26:53+0900 sample 5: clean gpu=[96908:mlxcel-bench-de]
2026-10-12T10:26:54+0900 monitor: 5 samples
2026-10-12T10:26:54+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:26:59+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:26:59+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:27:13+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:27:13+0900 attempt 1/10: start: env /home/devops/Development/pre-epic-4c44e317/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:27:13+0900 sample 1: clean gpu=[]
2026-10-12T10:27:15+0900 sample 2: clean gpu=[110159:mlxcel-bench-de]
2026-10-12T10:27:16+0900 sample 3: clean gpu=[110159:mlxcel-bench-de]
2026-10-12T10:27:18+0900 sample 4: clean gpu=[110159:mlxcel-bench-de]
2026-10-12T10:27:19+0900 sample 5: clean gpu=[110159:mlxcel-bench-de]
2026-10-12T10:27:20+0900 monitor: 5 samples
2026-10-12T10:27:20+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:26:05+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:26:05+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:26:20+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:26:20+0900 attempt 1/10: start: env MLXCEL_FORCE_SYNC=1 /home/devops/Development/wt-20261006T1212-issue-2239/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:26:20+0900 sample 1: clean gpu=[]
2026-10-12T10:26:22+0900 sample 2: clean gpu=[83655:mlxcel-bench-de]
2026-10-12T10:26:23+0900 sample 3: clean gpu=[83655:mlxcel-bench-de]
2026-10-12T10:26:24+0900 sample 4: clean gpu=[83655:mlxcel-bench-de]
2026-10-12T10:26:26+0900 sample 5: clean gpu=[83655:mlxcel-bench-de]
2026-10-12T10:26:27+0900 monitor: 5 samples
2026-10-12T10:26:27+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:27:25+0900 lock held by outer guard 4162902: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:27:25+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:27:40+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:27:40+0900 attempt 1/10: start: env /home/devops/Development/wt-20261006T1212-issue-2239/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:27:40+0900 sample 1: clean gpu=[]
2026-10-12T10:27:42+0900 sample 2: clean gpu=[123429:mlxcel-bench-de]
2026-10-12T10:27:43+0900 sample 3: clean gpu=[123429:mlxcel-bench-de]
2026-10-12T10:27:45+0900 sample 4: clean gpu=[123429:mlxcel-bench-de]
2026-10-12T10:27:46+0900 sample 5: clean gpu=[123429:mlxcel-bench-de]
2026-10-12T10:27:47+0900 monitor: 5 samples
2026-10-12T10:27:47+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:39:49+0900 lock held by outer guard 138159: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:39:49+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:40:03+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:40:03+0900 attempt 1/10: start: env MLXCEL_FORCE_SYNC=1 /home/devops/Development/pre-fix-e24fb7b2/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:40:03+0900 sample 1: clean gpu=[]
2026-10-12T10:40:05+0900 sample 2: clean gpu=[159154:mlxcel-bench-de]
2026-10-12T10:40:06+0900 sample 3: clean gpu=[159154:mlxcel-bench-de]
2026-10-12T10:40:08+0900 sample 4: clean gpu=[159154:mlxcel-bench-de]
2026-10-12T10:40:09+0900 sample 5: clean gpu=[159154:mlxcel-bench-de]
2026-10-12T10:40:10+0900 monitor: 5 samples
2026-10-12T10:40:10+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:39:22+0900 lock held by outer guard 138159: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:39:22+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:39:36+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:39:36+0900 attempt 1/10: start: env /home/devops/Development/pre-fix-e24fb7b2/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:39:37+0900 sample 1: clean gpu=[]
2026-10-12T10:39:38+0900 sample 2: clean gpu=[146337:mlxcel-bench-de]
2026-10-12T10:39:40+0900 sample 3: clean gpu=[146337:mlxcel-bench-de]
2026-10-12T10:39:41+0900 sample 4: clean gpu=[146337:mlxcel-bench-de]
2026-10-12T10:39:43+0900 sample 5: clean gpu=[146337:mlxcel-bench-de]
2026-10-12T10:39:44+0900 monitor: 5 samples
2026-10-12T10:39:44+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:40:15+0900 lock held by outer guard 138159: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:40:15+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:40:29+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:40:29+0900 attempt 1/10: start: env MLXCEL_FORCE_SYNC=1 /home/devops/Development/pre-fix-e24fb7b2/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:40:30+0900 sample 1: clean gpu=[]
2026-10-12T10:40:31+0900 sample 2: clean gpu=[172055:mlxcel-bench-de]
2026-10-12T10:40:33+0900 sample 3: clean gpu=[172055:mlxcel-bench-de]
2026-10-12T10:40:34+0900 sample 4: clean gpu=[172055:mlxcel-bench-de]
2026-10-12T10:40:36+0900 sample 5: clean gpu=[172055:mlxcel-bench-de]
2026-10-12T10:40:37+0900 monitor: 5 samples
2026-10-12T10:40:37+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:40:42+0900 lock held by outer guard 138159: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:40:42+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:40:56+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:40:56+0900 attempt 1/10: start: env /home/devops/Development/pre-fix-e24fb7b2/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:40:56+0900 sample 1: clean gpu=[]
2026-10-12T10:40:58+0900 sample 2: clean gpu=[185046:mlxcel-bench-de]
2026-10-12T10:40:59+0900 sample 3: clean gpu=[185046:mlxcel-bench-de]
2026-10-12T10:41:01+0900 sample 4: clean gpu=[185046:mlxcel-bench-de]
2026-10-12T10:41:02+0900 sample 5: clean gpu=[185046:mlxcel-bench-de]
2026-10-12T10:41:03+0900 monitor: 5 samples
2026-10-12T10:41:03+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:41:35+0900 lock held by outer guard 138159: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:41:35+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:41:49+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:41:49+0900 attempt 1/10: start: env MLXCEL_FORCE_SYNC=1 /home/devops/Development/pre-fix-e24fb7b2/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:41:50+0900 sample 1: clean gpu=[]
2026-10-12T10:41:51+0900 sample 2: clean gpu=[211152:mlxcel-bench-de]
2026-10-12T10:41:53+0900 sample 3: clean gpu=[211152:mlxcel-bench-de]
2026-10-12T10:41:54+0900 sample 4: clean gpu=[211152:mlxcel-bench-de]
2026-10-12T10:41:56+0900 sample 5: clean gpu=[211152:mlxcel-bench-de]
2026-10-12T10:41:57+0900 monitor: 5 samples
2026-10-12T10:41:57+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
2026-10-12T10:41:08+0900 lock held by outer guard 138159: not taking /tmp/mlxcel-rocm-gpu-guard.lock
2026-10-12T10:41:08+0900 waiting for 10s of idle GPU and no compiler
2026-10-12T10:41:22+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
2026-10-12T10:41:22+0900 attempt 1/10: start: env /home/devops/Development/pre-fix-e24fb7b2/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
2026-10-12T10:41:23+0900 sample 1: clean gpu=[]
2026-10-12T10:41:24+0900 sample 2: clean gpu=[198087:mlxcel-bench-de]
2026-10-12T10:41:26+0900 sample 3: clean gpu=[198087:mlxcel-bench-de]
2026-10-12T10:41:27+0900 sample 4: clean gpu=[198087:mlxcel-bench-de]
2026-10-12T10:41:29+0900 sample 5: clean gpu=[198087:mlxcel-bench-de]
2026-10-12T10:41:30+0900 monitor: 5 samples
2026-10-12T10:41:30+0900 attempt 1: CLEAN, exit 0
Original file line number Diff line number Diff line change
@@ -0,0 +1,24 @@
rocm_gpu_guard: 2026-10-12T10:39:49+0900 lock held by outer guard 138159: not taking /tmp/mlxcel-rocm-gpu-guard.lock
rocm_gpu_guard: 2026-10-12T10:39:49+0900 waiting for 10s of idle GPU and no compiler
rocm_gpu_guard: 2026-10-12T10:40:03+0900 idle for 10s: kfd proc empty and no compiler in every 1 Hz sample
rocm_gpu_guard: 2026-10-12T10:40:03+0900 attempt 1/10: start: env MLXCEL_FORCE_SYNC=1 /home/devops/Development/pre-fix-e24fb7b2/target/release/mlxcel-bench-decode -m /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit -p x -n 128 --warmup-tokens 20 --ignore-eos --prompt-tokens 2048
[mlx-rocm] bound HIP device 0: gfx1151 (AMD Radeon 8060S Graphics) cus=20 warp=32 lds=64KB
[Memory] cache limit: 2.15 GB
Warning: model.safetensors.index.json in /home/devops/Development/wt-20261006T1212-issue-2239/models/mlx/gemma-3-4b-it-4bit references shards that don't match the on-disk files (likely a repackaged mlx-community quant). Falling back to all *.safetensors files in the directory.
[Load] wall: 0.654 s MLX peak: 0.00 GB
[Memory] after load: active 0.00 GB, cache 0.00 GB, active+cache 0.00 GB, phase peak 0.00 GB
[long-prompt] target=2048 tokens -> using 2048 tokens (max_context=None, reserved 128 for generation)
[mlx-rocm] gfx1151 native_wmma=1 cus=20 warp=32 low_cu=0 host_WARP_SIZE=32
[hipBLASLt caps] device 0 (gfx1151): bf16=1 fp8_e4m3=0 fp8_e5m2=0 int8=1
[Memory] after warmup (prefill + warmup tokens): active 2.56 GB, cache 1.02 GB, active+cache 3.58 GB, phase peak 4.27 GB
[long-prompt] target=2048 tokens -> using 2048 tokens (max_context=None, reserved 128 for generation)
[Memory] after measured pass: active 2.56 GB, cache 1.19 GB, active+cache 3.76 GB, phase peak 4.52 GB
[Profile Results]
Prompt tokens: 2048
Generated tokens: 128
Prefill: 711.43 ms (2878.72 tok/s)
Decode: 2161.19 ms (59.23 tok/s)
Decode path: engine
MLX peak memory: 4.52 GB
rocm_gpu_guard: 2026-10-12T10:40:10+0900 monitor: 5 samples
rocm_gpu_guard: 2026-10-12T10:40:10+0900 attempt 1: CLEAN, exit 0
Loading
Loading