-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathslurm_8867047.out
More file actions
375 lines (359 loc) · 33.8 KB
/
Copy pathslurm_8867047.out
File metadata and controls
375 lines (359 loc) · 33.8 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
The token has not been saved to the git credentials helper. Pass `add_to_git_credential=True` in this function directly or `--add-to-git-credential` if using via `huggingface-cli` if you want to set the git credential as well.
Token is valid (permission: fineGrained).
The token `llama3` has been saved to /dev/shm/hf-home/stored_tokens
Your token has been saved to /dev/shm/hf-home/token
Login successful.
The current active token is: `llama3`
wandb: Appending key for api.wandb.ai to your netrc file: /home/users/ap794/.netrc
wandb: W&B API key is configured. Use `wandb login --relogin` to force relogin
/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/utils/hub.py:105: FutureWarning: Using `TRANSFORMERS_CACHE` is deprecated and will be removed in v5 of Transformers. Use `HF_HOME` instead.
warnings.warn(
🦥 Unsloth: Will patch your computer to enable 2x faster free finetuning.
Unsloth: Failed to patch Gemma3ForConditionalGeneration.
🦥 Unsloth Zoo will now patch everything to make training faster!
INFO 04-20 16:23:06 [__init__.py:239] Automatically detected platform cuda.
Running GRPO script
Sun Apr 20 16:23:11 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.144.03 Driver Version: 550.144.03 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX A5000 Off | 00000000:17:00.0 Off | Off |
| 30% 23C P2 57W / 230W | 209MiB / 24564MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 2463496 C python 202MiB |
+-----------------------------------------------------------------------------------------+
==((====))== Unsloth 2025.3.19: Fast Qwen2 patching. Transformers: 4.51.3. vLLM: 0.8.2.
\\ /| NVIDIA RTX A5000. Num GPUs = 1. Max memory: 23.673 GB. Platform: Linux.
O^O/ \_/ \ Torch: 2.6.0+cu124. CUDA: 8.6. CUDA Toolkit: 12.4. Triton: 3.2.0
\ / Bfloat16 = TRUE. FA [Xformers = 0.0.29.post2. FA2 = False]
"-____-" Free license: http://github.com/unslothai/unsloth
Unsloth: Fast downloading is enabled - ignore downloading bars which are red colored!
Unsloth: vLLM loading /home/users/ap794/final_project_distillLLM/minillm/results/qwen2.5/train/sft/qwen2.5-1.5B-Instruct/e10-bs1-lr1e-05-G2-N4-NN1/8000 with actual GPU utilization = 59.38%
Unsloth: Your GPU has CUDA compute capability 8.6 with VRAM = 23.67 GB.
Unsloth: Using conservativeness = 1.0. Chunked prefill tokens = 2042. Num Sequences = 224.
Unsloth: vLLM's KV Cache can use up to 11.11 GB. Also swap space = 6 GB.
WARNING 04-20 16:23:12 [config.py:2614] Casting torch.float16 to torch.bfloat16.
INFO 04-20 16:23:25 [config.py:585] This model supports multiple tasks: {'generate', 'score', 'classify', 'reward', 'embed'}. Defaulting to 'generate'.
WARNING 04-20 16:23:25 [arg_utils.py:1854] --quantization bitsandbytes is not supported by the V1 Engine. Falling back to V0.
Unsloth: vLLM Bitsandbytes config using kwargs = {'load_in_8bit': False, 'load_in_4bit': True, 'bnb_4bit_compute_dtype': 'bfloat16', 'bnb_4bit_quant_storage': 'uint8', 'bnb_4bit_quant_type': 'fp4', 'bnb_4bit_use_double_quant': False, 'llm_int8_enable_fp32_cpu_offload': False, 'llm_int8_has_fp16_weight': False, 'llm_int8_skip_modules': [], 'llm_int8_threshold': 6.0}
INFO 04-20 16:23:26 [llm_engine.py:241] Initializing a V0 LLM engine (v0.8.2) with config: model='/home/users/ap794/final_project_distillLLM/minillm/results/qwen2.5/train/sft/qwen2.5-1.5B-Instruct/e10-bs1-lr1e-05-G2-N4-NN1/8000', speculative_config=None, tokenizer='/home/users/ap794/final_project_distillLLM/minillm/results/qwen2.5/train/sft/qwen2.5-1.5B-Instruct/e10-bs1-lr1e-05-G2-N4-NN1/8000', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=2042, download_dir=None, load_format=bitsandbytes, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=bitsandbytes, enforce_eager=False, kv_cache_dtype=auto, device_config=cuda:0, decoding_config=DecodingConfig(guided_decoding_backend='xgrammar', reasoning_backend=None), observability_config=ObservabilityConfig(show_hidden_metrics=False, otlp_traces_endpoint=None, collect_model_forward_time=False, collect_model_execute_time=False), seed=0, served_model_name=/home/users/ap794/final_project_distillLLM/minillm/results/qwen2.5/train/sft/qwen2.5-1.5B-Instruct/e10-bs1-lr1e-05-G2-N4-NN1/8000, num_scheduler_steps=1, multi_step_stream_outputs=True, enable_prefix_caching=True, chunked_prefill_enabled=False, use_async_output_proc=True, disable_mm_preprocessor_cache=False, mm_processor_kwargs=None, pooler_config=None, compilation_config={"level":0,"splitting_ops":[],"compile_sizes":[],"cudagraph_capture_sizes":[224,216,208,200,192,184,176,168,160,152,144,136,128,120,112,104,96,88,80,72,64,56,48,40,32,24,16,8,4,2,1],"max_capture_size":224}, use_cached_outputs=False,
INFO 04-20 16:23:27 [cuda.py:291] Using Flash Attention backend.
INFO 04-20 16:23:27 [parallel_state.py:954] rank 0 in world size 1 is assigned as DP rank 0, PP rank 0, TP rank 0
INFO 04-20 16:23:27 [model_runner.py:1110] Starting to load model /home/users/ap794/final_project_distillLLM/minillm/results/qwen2.5/train/sft/qwen2.5-1.5B-Instruct/e10-bs1-lr1e-05-G2-N4-NN1/8000...
INFO 04-20 16:23:27 [loader.py:1155] Loading weights with BitsAndBytes quantization. May take a while ...
Loading pt checkpoint shards: 0% Completed | 0/1 [00:00<?, ?it/s]
Loading pt checkpoint shards: 100% Completed | 1/1 [00:14<00:00, 14.20s/it]
Loading pt checkpoint shards: 100% Completed | 1/1 [00:14<00:00, 14.20s/it]
INFO 04-20 16:23:42 [punica_selector.py:18] Using PunicaWrapperGPU.
INFO 04-20 16:23:42 [model_runner.py:1146] Model loading took 1.1443 GB and 14.618849 seconds
INFO 04-20 16:23:45 [worker.py:267] Memory profiling takes 2.19 seconds
INFO 04-20 16:23:45 [worker.py:267] the current vLLM instance can use total_gpu_memory (23.67GiB) x gpu_memory_utilization (0.59) = 14.06GiB
INFO 04-20 16:23:45 [worker.py:267] model weights take 1.14GiB; non_torch_memory takes 0.05GiB; PyTorch activation peak memory takes 1.23GiB; the rest of the memory reserved for KV Cache is 11.63GiB.
INFO 04-20 16:23:45 [executor_base.py:111] # cuda blocks: 27219, # CPU blocks: 14043
INFO 04-20 16:23:45 [executor_base.py:116] Maximum concurrency for 2042 tokens per request: 213.27x
INFO 04-20 16:23:48 [model_runner.py:1442] Capturing cudagraphs for decoding. This may lead to unexpected consequences if the model is not static. To run the model in eager mode, set 'enforce_eager=True' or use '--enforce-eager' in the CLI. If out-of-memory error occurs during cudagraph capture, consider decreasing `gpu_memory_utilization` or switching to eager mode. You can also reduce the `max_num_seqs` as needed to decrease memory usage.
Capturing CUDA graph shapes: 0%| | 0/31 [00:00<?, ?it/s]Capturing CUDA graph shapes: 3%|▎ | 1/31 [00:00<00:23, 1.28it/s]Capturing CUDA graph shapes: 6%|▋ | 2/31 [00:01<00:20, 1.44it/s]Capturing CUDA graph shapes: 10%|▉ | 3/31 [00:02<00:18, 1.49it/s]Capturing CUDA graph shapes: 13%|█▎ | 4/31 [00:02<00:17, 1.51it/s]Capturing CUDA graph shapes: 16%|█▌ | 5/31 [00:03<00:16, 1.53it/s]Capturing CUDA graph shapes: 19%|█▉ | 6/31 [00:03<00:16, 1.53it/s]Capturing CUDA graph shapes: 23%|██▎ | 7/31 [00:04<00:15, 1.54it/s]Capturing CUDA graph shapes: 26%|██▌ | 8/31 [00:05<00:14, 1.55it/s]Capturing CUDA graph shapes: 29%|██▉ | 9/31 [00:05<00:14, 1.55it/s]Capturing CUDA graph shapes: 32%|███▏ | 10/31 [00:06<00:13, 1.55it/s]Capturing CUDA graph shapes: 35%|███▌ | 11/31 [00:07<00:12, 1.54it/s]Capturing CUDA graph shapes: 39%|███▊ | 12/31 [00:07<00:12, 1.54it/s]Capturing CUDA graph shapes: 42%|████▏ | 13/31 [00:08<00:11, 1.55it/s]Capturing CUDA graph shapes: 45%|████▌ | 14/31 [00:09<00:11, 1.53it/s]Capturing CUDA graph shapes: 48%|████▊ | 15/31 [00:09<00:10, 1.51it/s]Capturing CUDA graph shapes: 52%|█████▏ | 16/31 [00:10<00:09, 1.51it/s]Capturing CUDA graph shapes: 55%|█████▍ | 17/31 [00:11<00:09, 1.52it/s]Capturing CUDA graph shapes: 58%|█████▊ | 18/31 [00:11<00:08, 1.50it/s]Capturing CUDA graph shapes: 61%|██████▏ | 19/31 [00:12<00:07, 1.51it/s]Capturing CUDA graph shapes: 65%|██████▍ | 20/31 [00:13<00:07, 1.52it/s]Capturing CUDA graph shapes: 68%|██████▊ | 21/31 [00:13<00:06, 1.52it/s]Capturing CUDA graph shapes: 71%|███████ | 22/31 [00:14<00:05, 1.52it/s]Capturing CUDA graph shapes: 74%|███████▍ | 23/31 [00:15<00:05, 1.52it/s]Capturing CUDA graph shapes: 77%|███████▋ | 24/31 [00:15<00:04, 1.52it/s]Capturing CUDA graph shapes: 81%|████████ | 25/31 [00:16<00:03, 1.53it/s]Capturing CUDA graph shapes: 84%|████████▍ | 26/31 [00:17<00:03, 1.52it/s]Capturing CUDA graph shapes: 87%|████████▋ | 27/31 [00:17<00:02, 1.52it/s]Capturing CUDA graph shapes: 90%|█████████ | 28/31 [00:18<00:01, 1.52it/s]Capturing CUDA graph shapes: 94%|█████████▎| 29/31 [00:19<00:01, 1.53it/s]Capturing CUDA graph shapes: 97%|█████████▋| 30/31 [00:19<00:00, 1.53it/s]Capturing CUDA graph shapes: 100%|██████████| 31/31 [00:20<00:00, 1.50it/s]Capturing CUDA graph shapes: 100%|██████████| 31/31 [00:20<00:00, 1.52it/s]
INFO 04-20 16:24:08 [model_runner.py:1570] Graph capturing finished in 20 secs, took 0.60 GiB
INFO 04-20 16:24:08 [llm_engine.py:447] init engine (profile, create kv cache, warmup model) took 26.39 seconds
[WARNING|logging.py:328] 2025-04-20 16:24:17,143 >> Unsloth 2025.3.19 patched 28 layers with 28 QKV layers, 28 O layers and 28 MLP layers.
/home/users/ap794/final_project_distillLLM/minillm/results/qwen2.5/train/sft/qwen2.5-1.5B-Instruct/e10-bs1-lr1e-05-G2-N4-NN1/8000 does not have a padding token! Will use pad_token = <|vision_pad|>.
Sun Apr 20 16:24:14 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.144.03 Driver Version: 550.144.03 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX A5000 Off | 00000000:17:00.0 Off | Off |
| 30% 26C P2 57W / 230W | 14241MiB / 24564MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 2463496 C python 14220MiB |
+-----------------------------------------------------------------------------------------+
Sun Apr 20 16:24:19 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.144.03 Driver Version: 550.144.03 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX A5000 Off | 00000000:17:00.0 Off | Off |
| 30% 26C P2 57W / 230W | 14351MiB / 24564MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 2463496 C python 14330MiB |
+-----------------------------------------------------------------------------------------+
Unsloth: We now expect `per_device_train_batch_size` to be a multiple of `num_generations`.
We will change the batch size of 1 to the `num_generations` of 6
Generating train split: 0%| | 0/47780 [00:00<?, ? examples/s]Generating train split: 2%|▏ | 1000/47780 [00:00<00:25, 1860.54 examples/s]Generating train split: 4%|▍ | 2000/47780 [00:00<00:17, 2611.89 examples/s]Generating train split: 6%|▋ | 3000/47780 [00:01<00:15, 2835.02 examples/s]Generating train split: 8%|▊ | 4000/47780 [00:01<00:12, 3609.71 examples/s]Generating train split: 12%|█▏ | 5778/47780 [00:01<00:12, 3376.94 examples/s]Generating train split: 14%|█▍ | 6778/47780 [00:02<00:11, 3519.52 examples/s]Generating train split: 16%|█▋ | 7778/47780 [00:02<00:11, 3609.14 examples/s]Generating train split: 18%|█▊ | 8778/47780 [00:02<00:10, 3845.90 examples/s]Generating train split: 22%|██▏ | 10556/47780 [00:03<00:10, 3515.96 examples/s]Generating train split: 24%|██▍ | 11556/47780 [00:03<00:09, 3722.24 examples/s]Generating train split: 26%|██▋ | 12556/47780 [00:03<00:08, 3977.47 examples/s]Generating train split: 28%|██▊ | 13556/47780 [00:03<00:08, 4156.89 examples/s]Generating train split: 32%|███▏ | 15334/47780 [00:04<00:08, 3729.59 examples/s]Generating train split: 34%|███▍ | 16334/47780 [00:04<00:07, 4099.50 examples/s]Generating train split: 36%|███▋ | 17334/47780 [00:04<00:07, 3891.62 examples/s]Generating train split: 38%|███▊ | 18334/47780 [00:04<00:06, 4588.07 examples/s]Generating train split: 42%|████▏ | 20112/47780 [00:05<00:06, 4078.05 examples/s]Generating train split: 44%|████▍ | 21112/47780 [00:05<00:06, 3872.34 examples/s]Generating train split: 46%|████▋ | 22112/47780 [00:06<00:07, 3489.94 examples/s]Generating train split: 48%|████▊ | 23112/47780 [00:06<00:05, 4116.88 examples/s]Generating train split: 52%|█████▏ | 24890/47780 [00:06<00:06, 3435.56 examples/s]Generating train split: 54%|█████▍ | 25890/47780 [00:07<00:06, 3317.16 examples/s]Generating train split: 56%|█████▋ | 26890/47780 [00:07<00:06, 3279.93 examples/s]Generating train split: 58%|█████▊ | 27890/47780 [00:07<00:05, 3870.58 examples/s]Generating train split: 62%|██████▏ | 29668/47780 [00:08<00:05, 3292.74 examples/s]Generating train split: 64%|██████▍ | 30668/47780 [00:08<00:04, 3692.76 examples/s]Generating train split: 66%|██████▋ | 31668/47780 [00:08<00:04, 3438.20 examples/s]Generating train split: 68%|██████▊ | 32668/47780 [00:08<00:03, 3902.76 examples/s]Generating train split: 72%|███████▏ | 34446/47780 [00:09<00:04, 3049.01 examples/s]Generating train split: 74%|███████▍ | 35446/47780 [00:10<00:03, 3135.77 examples/s]Generating train split: 76%|███████▋ | 36446/47780 [00:10<00:03, 3421.26 examples/s]Generating train split: 78%|███████▊ | 37446/47780 [00:10<00:02, 3784.07 examples/s]Generating train split: 82%|████████▏ | 39224/47780 [00:11<00:02, 3264.60 examples/s]Generating train split: 84%|████████▍ | 40224/47780 [00:11<00:02, 3245.07 examples/s]Generating train split: 86%|████████▋ | 41224/47780 [00:11<00:02, 3157.03 examples/s]Generating train split: 88%|████████▊ | 42224/47780 [00:11<00:01, 3563.89 examples/s]Generating train split: 92%|█████████▏| 44002/47780 [00:12<00:01, 3044.01 examples/s]Generating train split: 94%|█████████▍| 45002/47780 [00:12<00:00, 3092.42 examples/s]Generating train split: 96%|█████████▋| 46002/47780 [00:13<00:00, 2945.84 examples/s]Generating train split: 98%|█████████▊| 47002/47780 [00:13<00:00, 3518.46 examples/s]Generating train split: 100%|██████████| 47780/47780 [00:13<00:00, 3536.57 examples/s]
Map: 0%| | 0/47780 [00:00<?, ? examples/s]Map: 2%|▏ | 746/47780 [00:00<00:06, 7356.28 examples/s]Map: 4%|▎ | 1738/47780 [00:00<00:10, 4212.32 examples/s]Map: 5%|▍ | 2330/47780 [00:00<00:14, 3093.02 examples/s]Map: 6%|▋ | 2996/47780 [00:00<00:11, 3844.86 examples/s]Map: 8%|▊ | 3616/47780 [00:01<00:14, 3130.86 examples/s]Map: 9%|▉ | 4341/47780 [00:01<00:18, 2364.68 examples/s]Map: 10%|█ | 5000/47780 [00:01<00:17, 2460.32 examples/s]Map: 12%|█▏ | 5706/47780 [00:01<00:13, 3120.44 examples/s]Map: 13%|█▎ | 6318/47780 [00:02<00:14, 2948.63 examples/s]Map: 15%|█▍ | 6936/47780 [00:02<00:11, 3480.36 examples/s]Map: 16%|█▋ | 7875/47780 [00:02<00:11, 3516.34 examples/s]Map: 18%|█▊ | 8394/47780 [00:02<00:12, 3241.36 examples/s]Map: 19%|█▉ | 9000/47780 [00:02<00:11, 3326.85 examples/s]Map: 21%|██ | 9909/47780 [00:02<00:08, 4366.49 examples/s]Map: 23%|██▎ | 10827/47780 [00:03<00:08, 4152.46 examples/s]Map: 24%|██▍ | 11427/47780 [00:03<00:09, 3686.60 examples/s]Map: 25%|██▌ | 12000/47780 [00:03<00:09, 3668.40 examples/s]Map: 27%|██▋ | 12852/47780 [00:03<00:07, 4577.10 examples/s]Map: 28%|██▊ | 13464/47780 [00:03<00:08, 4273.10 examples/s]Map: 29%|██▉ | 14000/47780 [00:03<00:08, 3999.15 examples/s]Map: 31%|███ | 14880/47780 [00:04<00:06, 5000.43 examples/s]Map: 33%|███▎ | 15887/47780 [00:04<00:06, 4797.59 examples/s]Map: 35%|███▌ | 16840/47780 [00:04<00:06, 4617.88 examples/s]Map: 37%|███▋ | 17480/47780 [00:04<00:07, 4070.14 examples/s]Map: 38%|███▊ | 18000/47780 [00:04<00:07, 3894.54 examples/s]Map: 40%|███▉ | 18927/47780 [00:04<00:05, 4920.13 examples/s]Map: 42%|████▏ | 19917/47780 [00:05<00:06, 4616.22 examples/s]Map: 44%|████▎ | 20890/47780 [00:05<00:05, 4695.29 examples/s]Map: 45%|████▍ | 21449/47780 [00:05<00:05, 4407.41 examples/s]Map: 46%|████▌ | 22000/47780 [00:05<00:06, 4051.04 examples/s]Map: 48%|████▊ | 22879/47780 [00:05<00:05, 4978.15 examples/s]Map: 49%|████▉ | 23447/47780 [00:06<00:05, 4242.68 examples/s]Map: 50%|█████ | 24000/47780 [00:06<00:06, 3842.87 examples/s]Map: 52%|█████▏ | 24755/47780 [00:06<00:05, 4583.48 examples/s]Map: 53%|█████▎ | 25420/47780 [00:06<00:05, 3862.60 examples/s]Map: 54%|█████▍ | 26000/47780 [00:06<00:05, 3647.36 examples/s]Map: 56%|█████▌ | 26808/47780 [00:06<00:04, 4504.10 examples/s]Map: 57%|█████▋ | 27410/47780 [00:07<00:05, 3793.88 examples/s]Map: 59%|█████▊ | 28000/47780 [00:07<00:05, 3590.12 examples/s]Map: 60%|██████ | 28845/47780 [00:07<00:04, 4525.06 examples/s]Map: 62%|██████▏ | 29431/47780 [00:07<00:04, 3839.19 examples/s]Map: 63%|██████▎ | 30000/47780 [00:07<00:04, 3590.03 examples/s]Map: 65%|██████▍ | 30930/47780 [00:07<00:03, 4696.56 examples/s]Map: 67%|██████▋ | 31873/47780 [00:08<00:03, 4659.10 examples/s]Map: 68%|██████▊ | 32421/47780 [00:08<00:03, 4240.41 examples/s]Map: 69%|██████▉ | 33000/47780 [00:08<00:03, 3759.21 examples/s]Map: 71%|███████ | 33807/47780 [00:08<00:03, 4587.29 examples/s]Map: 72%|███████▏ | 34411/47780 [00:08<00:03, 4006.48 examples/s]Map: 73%|███████▎ | 35000/47780 [00:08<00:03, 3579.15 examples/s]Map: 75%|███████▌ | 35859/47780 [00:09<00:02, 4529.03 examples/s]Map: 76%|███████▋ | 36440/47780 [00:09<00:03, 3756.11 examples/s]Map: 77%|███████▋ | 37000/47780 [00:09<00:02, 3705.01 examples/s]Map: 79%|███████▉ | 37892/47780 [00:09<00:02, 4739.97 examples/s]Map: 81%|████████ | 38820/47780 [00:09<00:01, 4542.93 examples/s]Map: 83%|████████▎ | 39465/47780 [00:09<00:02, 3941.03 examples/s]Map: 84%|████████▎ | 40000/47780 [00:10<00:02, 3556.07 examples/s]Map: 86%|████████▌ | 40914/47780 [00:10<00:01, 4579.50 examples/s]Map: 88%|████████▊ | 41819/47780 [00:10<00:01, 4250.75 examples/s]Map: 89%|████████▉ | 42440/47780 [00:10<00:01, 3754.23 examples/s]Map: 90%|████████▉ | 43000/47780 [00:10<00:01, 3464.97 examples/s]Map: 92%|█████████▏| 43924/47780 [00:11<00:00, 4489.08 examples/s]Map: 93%|█████████▎| 44473/47780 [00:11<00:00, 4088.10 examples/s]Map: 94%|█████████▍| 45000/47780 [00:11<00:00, 3683.38 examples/s]Map: 96%|█████████▌| 45810/47780 [00:11<00:00, 4549.80 examples/s]Map: 98%|█████████▊| 46705/47780 [00:11<00:00, 3974.32 examples/s]Map: 99%|█████████▉| 47535/47780 [00:12<00:00, 3565.96 examples/s]Map: 100%|██████████| 47780/47780 [00:12<00:00, 3923.81 examples/s]
[WARNING|<string>:173] 2025-04-20 16:24:56,506 >> ==((====))== Unsloth - 2x faster free finetuning | Num GPUs used = 1
\\ /| Num examples = 47,780 | Num Epochs = 1 | Total steps = 250
O^O/ \_/ \ Batch size per device = 6 | Gradient accumulation steps = 1
\ / Data Parallel GPUs = 1 | Total batch size (6 x 1 x 1) = 6
"-____-" Trainable parameters = 36,929,536/5,000,000,000 (0.74% trained)
wandb: WARNING The `run_name` is currently set to the same value as `TrainingArguments.output_dir`. If this was not intended, please specify a different run name by setting the `TrainingArguments.run_name` parameter.
wandb: Using wandb-core as the SDK backend. Please refer to https://wandb.me/wandb-core for more information.
wandb: Currently logged in as: alejandro-paredeslatorre to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
[{'content': '\n Respond in the following format:\n <think>\n ...\n </think>\n <answer>\n ...\n </answer>\n ', 'role': 'system'}, {'content': 'You will be given a competitive programming problem. Please reason step by step about the solution, then provide a complete implementation in C++17.\n\nYour solution must read input from standard input (cin), write output to standard output (cout).\nDo not include any debug prints or additional output.\n\nPut your final solution within a single code block:\n```cpp\n<your code here>\n```\n\n# Problem\n\nYou are given an array $$$a$$$ of $$$n$$$ integers, where $$$n$$$ is odd.\n\nIn one operation, you will remove two adjacent elements from the array $$$a$$$, and then concatenate the remaining parts of the array. For example, given the array $$$[4,7,4,2,9]$$$, we can obtain the arrays $$$[4,2,9]$$$ and $$$[4,7,9]$$$ by the operations $$$[\\underline{4,7}, 4,2,9] \\to [4,2,9]$$$ and $$$[4,7,\\underline{4,2},9] \\to [4,7,9]$$$ respectively. However, we cannot obtain the array $$$[7,2,9]$$$ as it requires deleting non-adjacent elements $$$[\\underline{4},7,\\underline{4},2,9]$$$.\n\nYou will repeatedly perform this operation until exact', 'role': 'user'}]
```cpp
#include <bits/stdc++.h>
using namespace std;
int main() {
int t;
cin >> t;
while (t--) {
int n;
cin >> n;
vector<int> a(n);
for (int i = 0; i < n; ++i) {
cin >> a[i];
}
int max_val = a[0];
for (int i = 2; i < n; i += 2) {
if (a[i] > max_val) {
max_val = a[i];
}
}
cout << max_val << '\n';
}
return 0;
}
```
Sun Apr 20 16:24:55 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.144.03 Driver Version: 550.144.03 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX A5000 Off | 00000000:17:00.0 Off | Off |
| 30% 21C P8 16W / 230W | 14351MiB / 24564MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 2463496 C python 14330MiB |
+-----------------------------------------------------------------------------------------+
wandb: ERROR failed to upsert bucket: returned error 403: {"data":{"upsertBucket":null},"errors":[{"message":"permission denied","path":["upsertBucket"],"extensions":{"code":"PERMISSION_ERROR"}}]}
Traceback (most recent call last):
File "/home/users/ap794/final_project_distillLLM/aleGRPO/src/main.py", line 236, in <module>
main()
File "/home/users/ap794/final_project_distillLLM/aleGRPO/src/main.py", line 188, in main
trainer.train()
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/trainer.py", line 2245, in train
return inner_training_loop(
File "<string>", line 223, in _fast_inner_training_loop
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/trainer_callback.py", line 506, in on_train_begin
return self.call_event("on_train_begin", args, state, control)
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/trainer_callback.py", line 556, in call_event
result = getattr(callback, event)(
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/integrations/integration_utils.py", line 930, in on_train_begin
self.setup(args, state, model, **kwargs)
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/integrations/integration_utils.py", line 857, in setup
self._wandb.init(
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1544, in init
wandb._sentry.reraise(e)
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/analytics/sentry.py", line 156, in reraise
raise exc.with_traceback(sys.exc_info()[2])
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1530, in init
return wi.init(run_settings, run_config)
File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 987, in init
raise error
wandb.errors.errors.CommError: failed to upsert bucket: returned error 403: {"data":{"upsertBucket":null},"errors":[{"message":"permission denied","path":["upsertBucket"],"extensions":{"code":"PERMISSION_ERROR"}}]}
[rank0]: Traceback (most recent call last):
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/src/main.py", line 236, in <module>
[rank0]: main()
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/src/main.py", line 188, in main
[rank0]: trainer.train()
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/trainer.py", line 2245, in train
[rank0]: return inner_training_loop(
[rank0]: File "<string>", line 223, in _fast_inner_training_loop
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/trainer_callback.py", line 506, in on_train_begin
[rank0]: return self.call_event("on_train_begin", args, state, control)
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/trainer_callback.py", line 556, in call_event
[rank0]: result = getattr(callback, event)(
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/integrations/integration_utils.py", line 930, in on_train_begin
[rank0]: self.setup(args, state, model, **kwargs)
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/transformers/integrations/integration_utils.py", line 857, in setup
[rank0]: self._wandb.init(
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1544, in init
[rank0]: wandb._sentry.reraise(e)
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/analytics/sentry.py", line 156, in reraise
[rank0]: raise exc.with_traceback(sys.exc_info()[2])
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1530, in init
[rank0]: return wi.init(run_settings, run_config)
[rank0]: File "/home/users/ap794/final_project_distillLLM/aleGRPO/grpo_venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 987, in init
[rank0]: raise error
[rank0]: wandb.errors.errors.CommError: failed to upsert bucket: returned error 403: {"data":{"upsertBucket":null},"errors":[{"message":"permission denied","path":["upsertBucket"],"extensions":{"code":"PERMISSION_ERROR"}}]}
[1;34mwandb[0m:
[1;34mwandb[0m: 🚀 View run [33moutputs[0m at: [34mhttps://wandb.ai/alejandro-paredeslatorre/qwen-cot-training/runs/82oh4fux[0m
[1;34mwandb[0m: Find logs at: [1;35mwandb/run-20250420_162457-82oh4fux/logs[0m
[rank0]:[W420 16:25:00.557612253 ProcessGroupNCCL.cpp:1496] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
srun: error: compsci-cluster-fitz-14: task 0: Exited with exit code 1