Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/results.json filter=lfs diff=lfs merge=lfs -text
+outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/results.json filter=lfs diff=lfs merge=lfs -text
+outputs/2025-11-18/exp1_len1024/dream-inst/no_cache/gsm8k/results.json filter=lfs diff=lfs merge=lfs -text
+outputs/2025-11-18/exp1_len1024/dream-inst/prefix/gsm8k/results.json filter=lfs diff=lfs merge=lfs -text
+outputs/2025-11-18/exp2_random_gen/dream-inst/gsm8k_len256/results.json filter=lfs diff=lfs merge=lfs -text

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/.hydra/config.yaml

@@ -0,0 +1,53 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: vanilla
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: 32
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.dLLMCache
+  kr: 1
+  kp: 50
+  rou: 0.25
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: gsm8k
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/.hydra/hydra.yaml

@@ -0,0 +1,167 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k
+    - hydra.mode=RUN
+    task:
+    - dataset.name=gsm8k
+    - model=dream-inst
+    - cache=dllm
+    - generation=vanilla
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - generation.block_length=32
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache=dllm,dataset.name=gsm8k,generation.block_length=32,generation.gen_length=1024,generation.steps=1024,generation=vanilla,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k
+    choices:
+      cache: dllm
+      generation: vanilla
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/.hydra/overrides.yaml

@@ -0,0 +1,9 @@
+- dataset.name=gsm8k
+- model=dream-inst
+- cache=dllm
+- generation=vanilla
+- generation.gen_length=1024
+- generation.steps=1024
+- generation.block_length=32
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/eval.log

@@ -0,0 +1,41 @@
+[2025-11-18 21:17:27,700][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-18 21:17:35,559][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:35,559][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:35,663][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:35,664][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:35,918][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:35,918][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:36,015][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:36,015][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:37,961][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:37,961][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:38,027][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:38,027][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:38,231][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:38,231][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:39,335][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 21:17:39,335][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 21:17:46,329][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:46,330][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 5...
+[2025-11-18 21:17:46,906][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:46,907][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 7...
+[2025-11-18 21:17:47,086][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:47,087][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 3...
+[2025-11-18 21:17:48,048][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:48,048][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 6...
+[2025-11-18 21:17:51,594][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:51,595][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 0...
+[2025-11-18 21:17:51,886][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:51,886][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 4...
+[2025-11-18 21:17:53,012][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:53,012][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 1...
+[2025-11-18 21:17:53,340][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-18 21:17:53,341][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 2...
+[2025-11-18 21:17:57,676][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 21:17:57,676][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 21:17:57,676][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 21:17:57,676][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 21:17:57,676][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 21:17:57,677][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 21:17:57,677][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 21:17:57,677][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/results.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c31ab00416ae76d358c5c97124fbbca1c6b43cf1b1d325f07b3ed2e163bb17e
+size 13676970

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/stderr.log

@@ -0,0 +1,63 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1118 21:17:11.095728867 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-18 21:17:19.569 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'vanilla', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 32, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': None}
+2025-11-18 21:17:19.569 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'kp': 25, 'kr': 2}
+[W1118 21:17:24.959332404 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 21:17:24.316010842 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 21:17:24.483325854 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 21:17:24.510713611 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 21:17:24.551847406 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 21:17:24.622360312 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 21:17:24.658556962 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 21:17:25.013811805 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
  0%|          | 0/165 [00:00<?, ?it/s]
 17%|█▋        | 28/165 [00:00<00:00, 271.20it/s]
 34%|███▍      | 56/165 [00:00<00:00, 271.85it/s]
 53%|█████▎    | 87/165 [00:00<00:00, 286.62it/s]
 77%|███████▋  | 127/165 [00:00<00:00, 327.90it/s]
+
  0%|          | 0/164 [00:00<?, ?it/s]
 13%|█▎        | 22/164 [00:00<00:00, 216.66it/s]
  0%|          | 0/165 [00:00<?, ?it/s]
 37%|███▋      | 61/164 [00:00<00:00, 315.68it/s]
 24%|██▎       | 39/165 [00:00<00:00, 386.72it/s]
 61%|██████    | 100/164 [00:00<00:00, 346.89it/s]
 47%|████▋     | 78/165 [00:00<00:00, 387.03it/s]
 85%|████████▍ | 139/164 [00:00<00:00, 363.02it/s]
 71%|███████   | 117/165 [00:00<00:00, 387.94it/s]
+
 95%|█████████▍| 156/165 [00:00<00:00, 387.82it/s]
+
  0%|          | 0/165 [00:00<?, ?it/s]
 24%|██▎       | 39/165 [00:00<00:00, 385.35it/s]
 47%|████▋     | 78/165 [00:00<00:00, 387.84it/s]
 72%|███████▏  | 118/165 [00:00<00:00, 388.85it/s]
 96%|█████████▌| 158/165 [00:00<00:00, 389.70it/s]
+
  0%|          | 0/165 [00:00<?, ?it/s]
 21%|██        | 35/165 [00:00<00:00, 348.57it/s]
 45%|████▍     | 74/165 [00:00<00:00, 368.97it/s]
  0%|          | 0/165 [00:00<?, ?it/s]
 68%|██████▊   | 112/165 [00:00<00:00, 373.57it/s]
 24%|██▎       | 39/165 [00:00<00:00, 385.26it/s]
 92%|█████████▏| 151/165 [00:00<00:00, 377.34it/s]
+
 47%|████▋     | 78/165 [00:00<00:00, 386.80it/s]
 72%|███████▏  | 118/165 [00:00<00:00, 388.67it/s]
 96%|█████████▌| 158/165 [00:00<00:00, 390.94it/s]
+
  0%|          | 0/165 [00:00<?, ?it/s]
 24%|██▎       | 39/165 [00:00<00:00, 385.52it/s]
 47%|████▋     | 78/165 [00:00<00:00, 388.01it/s]
 72%|███████▏  | 118/165 [00:00<00:00, 390.90it/s]
  0%|          | 0/165 [00:00<?, ?it/s]
 96%|█████████▌| 158/165 [00:00<00:00, 392.38it/s]
+
 24%|██▎       | 39/165 [00:00<00:00, 386.65it/s]
 47%|████▋     | 78/165 [00:00<00:00, 386.51it/s]
 72%|███████▏  | 118/165 [00:00<00:00, 388.81it/s]
 96%|█████████▌| 158/165 [00:00<00:00, 390.15it/s]
+
+2025-11-18 21:17:57.690 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+2025-11-18 21:17:57.690 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-18 21:17:57.691 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-18 21:17:57.692 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'sigma', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+2025-11-18 21:17:57.694 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-18 21:17:57.697 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-18 21:17:57.700 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'sigma') are not supported by the generation strategy 'vanilla'.
+
+2025-11-19 01:01:00.021 | INFO     | __main__:main:81 - Throughput: 1.63 tokens/sec, Tokens per step: 0.96 tokens/step (full: 12.98 tokens/sec, 1.00 tokens/step), Latency: 79.11 s, Total time: 12965.69 s
+2025-11-19 01:01:00.135 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/gsm8k/results.json
+2025-11-19 01:01:00.135 | INFO     | __main__:main:101 - eval time: 12965.69 seconds
+[rank0]:[W1119 01:01:00.284901363 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval/.hydra/config.yaml

@@ -0,0 +1,53 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: vanilla
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: 32
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.dLLMCache
+  kr: 1
+  kp: 50
+  rou: 0.25
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: humaneval
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval/.hydra/hydra.yaml

@@ -0,0 +1,167 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval
+    - hydra.mode=RUN
+    task:
+    - dataset.name=humaneval
+    - model=dream-inst
+    - cache=dllm
+    - generation=vanilla
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - generation.block_length=32
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache=dllm,dataset.name=humaneval,generation.block_length=32,generation.gen_length=1024,generation.steps=1024,generation=vanilla,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval
+    choices:
+      cache: dllm
+      generation: vanilla
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval/.hydra/overrides.yaml

@@ -0,0 +1,9 @@
+- dataset.name=humaneval
+- model=dream-inst
+- cache=dllm
+- generation=vanilla
+- generation.gen_length=1024
+- generation.steps=1024
+- generation.block_length=32
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval/eval.log

@@ -0,0 +1,41 @@
+[2025-11-18 11:28:45,856][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-18 11:29:09,035][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:09,035][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:09,290][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:09,290][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:09,542][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:09,542][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:09,655][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:09,655][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:10,420][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:10,420][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:11,324][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:11,325][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:11,664][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:11,665][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 5...
+[2025-11-18 11:29:11,944][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:11,946][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 4...
+[2025-11-18 11:29:12,224][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:12,226][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 6...
+[2025-11-18 11:29:12,299][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:12,300][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 2...
+[2025-11-18 11:29:12,987][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:12,988][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 1...
+[2025-11-18 11:29:13,185][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:13,185][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:13,388][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:29:13,388][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:29:13,802][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:13,803][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 0...
+[2025-11-18 11:29:15,749][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:15,750][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 7...
+[2025-11-18 11:29:16,000][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:29:16,001][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 3...
+[2025-11-18 11:29:19,932][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:29:19,932][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:29:19,932][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:29:19,932][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:29:19,933][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:29:19,933][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:29:19,933][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:29:19,934][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval/stderr.log

@@ -0,0 +1,63 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1118 11:28:30.873002893 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:28:39.149288617 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-18 11:28:39.480 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'vanilla', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 32, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': None}
+2025-11-18 11:28:39.481 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'kp': 50, 'kr': 1}
+[W1118 11:28:40.236613715 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:28:42.690415311 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:28:43.286219076 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:28:43.303679181 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:28:43.346367847 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:28:43.441263228 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:28:43.473438887 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
  0%|          | 0/20 [00:00<?, ?it/s]
+
  0%|          | 0/20 [00:00<?, ?it/s]
+
  0%|          | 0/20 [00:00<?, ?it/s]
+
  0%|          | 0/21 [00:00<?, ?it/s]
+
  0%|          | 0/21 [00:00<?, ?it/s]
+
  0%|          | 0/21 [00:00<?, ?it/s]
+
  0%|          | 0/20 [00:00<?, ?it/s]
+
  0%|          | 0/21 [00:00<?, ?it/s]
+
+2025-11-18 11:29:19.940 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+2025-11-18 11:29:19.940 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'add_bos_token', 'eot_token_id') are not supported by the generation strategy 'vanilla'.
+2025-11-18 11:29:19.941 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'add_bos_token', 'eot_token_id') are not supported by the generation strategy 'vanilla'.
+2025-11-18 11:29:19.942 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-18 11:29:19.944 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'add_bos_token', 'eot_token_id') are not supported by the generation strategy 'vanilla'.
+2025-11-18 11:29:19.949 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'sigma', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+2025-11-18 11:29:19.950 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+
+2025-11-18 11:56:59.412 | INFO     | __main__:main:81 - Throughput: 2.40 tokens/sec, Tokens per step: 0.94 tokens/step (full: 13.58 tokens/sec, 1.00 tokens/step), Latency: 75.51 s, Total time: 1593.89 s
+2025-11-18 11:56:59.424 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/humaneval/results.json
+2025-11-18 11:56:59.424 | INFO     | __main__:main:101 - eval time: 1593.89 seconds
+[rank0]:[W1118 11:56:59.614479582 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500/.hydra/config.yaml

@@ -0,0 +1,53 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: vanilla
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: 32
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.dLLMCache
+  kr: 1
+  kp: 50
+  rou: 0.25
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: math-500
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500/.hydra/hydra.yaml

@@ -0,0 +1,167 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500
+    - hydra.mode=RUN
+    task:
+    - dataset.name=math-500
+    - model=dream-inst
+    - cache=dllm
+    - generation=vanilla
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - generation.block_length=32
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache=dllm,dataset.name=math-500,generation.block_length=32,generation.gen_length=1024,generation.steps=1024,generation=vanilla,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500
+    choices:
+      cache: dllm
+      generation: vanilla
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500/.hydra/overrides.yaml

@@ -0,0 +1,9 @@
+- dataset.name=math-500
+- model=dream-inst
+- cache=dllm
+- generation=vanilla
+- generation.gen_length=1024
+- generation.steps=1024
+- generation.block_length=32
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500/eval.log

@@ -0,0 +1,41 @@
+[2025-11-19 11:50:26,990][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-19 11:50:38,980][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:38,980][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:39,186][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:39,186][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:39,639][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:39,640][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:39,771][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:39,771][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:39,881][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:39,881][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:39,888][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:39,888][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:40,118][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:40,118][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:40,213][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 11:50:40,214][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 11:50:44,328][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:44,331][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 1...
+[2025-11-19 11:50:44,654][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:44,655][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 0...
+[2025-11-19 11:50:44,984][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:44,988][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 7...
+[2025-11-19 11:50:45,168][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:45,170][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 3...
+[2025-11-19 11:50:45,328][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:45,330][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 2...
+[2025-11-19 11:50:45,344][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:45,345][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 6...
+[2025-11-19 11:50:45,439][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:45,442][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 5...
+[2025-11-19 11:50:45,451][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 11:50:45,452][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 4...
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 11:50:49,775][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500/stderr.log

@@ -0,0 +1,103 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1119 11:50:12.876749115 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-19 11:50:20.128 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'vanilla', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 32, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': None}
+2025-11-19 11:50:20.129 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'kp': 50, 'kr': 1}
+[W1119 11:50:20.738238144 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 11:50:24.879620258 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 11:50:24.888070672 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 11:50:24.180661413 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 11:50:24.384845106 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 11:50:24.449183109 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 11:50:24.576892757 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 11:50:25.843752097 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+2025-11-19 11:50:34.709 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:34.714 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_2s57y1bk...
+2025-11-19 11:50:34.877 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:34.890 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_aos902xl...
+2025-11-19 11:50:34.944 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:34.950 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_9dpjelay...
+2025-11-19 11:50:35.294 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:35.299 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_tljtvj6v...
+2025-11-19 11:50:35.406 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:35.414 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:35.418 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_vn27d0lc...
+2025-11-19 11:50:35.422 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_ozxomwwo...
+2025-11-19 11:50:35.471 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:35.476 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_98204mei...
+2025-11-19 11:50:35.631 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 11:50:35.772 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 11:50:35.846 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 11:50:35.916 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 11:50:35.920 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_456oqdwt...
+2025-11-19 11:50:36.233 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 11:50:36.303 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 11:50:36.343 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 11:50:36.364 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 11:50:36.783 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+
  0%|          | 0/63 [00:00<?, ?it/s]
 79%|███████▉  | 50/63 [00:00<00:00, 498.82it/s]
+
  0%|          | 0/63 [00:00<?, ?it/s]
 81%|████████  | 51/63 [00:00<00:00, 494.69it/s]
+
  0%|          | 0/62 [00:00<?, ?it/s]
 53%|█████▎    | 33/62 [00:00<00:00, 319.86it/s]
  0%|          | 0/63 [00:00<?, ?it/s]
+
 68%|██████▊   | 43/63 [00:00<00:00, 414.51it/s]
  0%|          | 0/63 [00:00<?, ?it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
+
 86%|████████▌ | 54/63 [00:00<00:00, 537.73it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
 90%|█████████ | 56/62 [00:00<00:00, 550.59it/s]
+
  0%|          | 0/62 [00:00<?, ?it/s]
+
 89%|████████▊ | 55/62 [00:00<00:00, 548.88it/s]
 90%|█████████ | 56/62 [00:00<00:00, 549.94it/s]
+
+
+2025-11-19 11:50:49.788 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-19 11:50:49.789 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+2025-11-19 11:50:49.789 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+2025-11-19 11:50:49.789 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-19 11:50:49.790 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+2025-11-19 11:50:49.790 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-19 11:50:49.792 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+
+2025-11-19 13:22:16.866 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:16.866 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:16.868 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:16.870 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:16.870 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:16.870 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:16.872 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:16.874 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:16.884 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:16.885 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:16.887 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:16.889 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:16.891 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:16.892 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:22.355 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 13:22:22.361 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 13:22:22.361 | INFO     | __main__:main:81 - Throughput: 11.82 tokens/sec, Tokens per step: 1.00 tokens/step (full: 12.03 tokens/sec, 1.00 tokens/step), Latency: 85.26 s, Total time: 5443.58 s
+2025-11-19 13:22:22.390 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/math-500/results.json
+2025-11-19 13:22:22.390 | INFO     | __main__:main:101 - eval time: 5443.58 seconds
+[rank0]:[W1119 13:22:22.537894604 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp/.hydra/config.yaml

@@ -0,0 +1,53 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: vanilla
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: 32
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.dLLMCache
+  kr: 1
+  kp: 50
+  rou: 0.25
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: mbpp
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp/.hydra/hydra.yaml

@@ -0,0 +1,167 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp
+    - hydra.mode=RUN
+    task:
+    - dataset.name=mbpp
+    - model=dream-inst
+    - cache=dllm
+    - generation=vanilla
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - generation.block_length=32
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache=dllm,dataset.name=mbpp,generation.block_length=32,generation.gen_length=1024,generation.steps=1024,generation=vanilla,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp
+    choices:
+      cache: dllm
+      generation: vanilla
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp/.hydra/overrides.yaml

@@ -0,0 +1,9 @@
+- dataset.name=mbpp
+- model=dream-inst
+- cache=dllm
+- generation=vanilla
+- generation.gen_length=1024
+- generation.steps=1024
+- generation.block_length=32
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp/eval.log

@@ -0,0 +1,41 @@
+[2025-11-19 06:39:17,131][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-19 06:39:24,192][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:24,192][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:24,598][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:24,598][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:24,635][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:24,636][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:24,675][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:24,675][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:25,170][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:25,171][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:25,174][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:25,174][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:25,182][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:25,183][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:25,398][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 06:39:25,398][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 06:39:39,093][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:39,094][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 4...
+[2025-11-19 06:39:39,887][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:39,888][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 1...
+[2025-11-19 06:39:40,208][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:40,209][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 5...
+[2025-11-19 06:39:40,371][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:40,372][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 7...
+[2025-11-19 06:39:40,544][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:40,546][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 6...
+[2025-11-19 06:39:40,705][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:40,707][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 3...
+[2025-11-19 06:39:41,241][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:41,242][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 0...
+[2025-11-19 06:39:42,146][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 06:39:42,147][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 2...
+[2025-11-19 06:39:46,593][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 06:39:46,593][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 06:39:46,593][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 06:39:46,594][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 06:39:46,594][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 06:39:46,594][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 06:39:46,594][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 06:39:46,594][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp/stderr.log

@@ -0,0 +1,63 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1119 06:39:00.544153052 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-19 06:39:08.594 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'vanilla', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 32, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': None}
+2025-11-19 06:39:08.594 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'kp': 10, 'kr': 8}
+[W1119 06:39:09.079402470 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 06:39:13.471812944 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 06:39:13.518657843 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 06:39:14.961188753 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 06:39:14.176243946 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 06:39:14.237022163 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 06:39:14.289035287 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 06:39:14.549959783 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
  0%|          | 0/62 [00:00<?, ?it/s]
 32%|███▏      | 20/62 [00:00<00:00, 184.79it/s]
 63%|██████▎   | 39/62 [00:00<00:00, 133.65it/s]
 90%|█████████ | 56/62 [00:00<00:00, 144.26it/s]
+
  0%|          | 0/63 [00:00<?, ?it/s]
 24%|██▍       | 15/63 [00:00<00:00, 132.04it/s]
 46%|████▌     | 29/63 [00:00<00:00, 119.63it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
 67%|██████▋   | 42/63 [00:00<00:00, 122.31it/s]
 32%|███▏      | 20/62 [00:00<00:00, 196.39it/s]
 87%|████████▋ | 55/63 [00:00<00:00, 115.72it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
+
 65%|██████▍   | 40/62 [00:00<00:00, 197.85it/s]
 32%|███▏      | 20/62 [00:00<00:00, 193.20it/s]
 97%|█████████▋| 60/62 [00:00<00:00, 198.37it/s]
+
  0%|          | 0/62 [00:00<?, ?it/s]
 65%|██████▍   | 40/62 [00:00<00:00, 193.98it/s]
 32%|███▏      | 20/62 [00:00<00:00, 197.83it/s]
 97%|█████████▋| 60/62 [00:00<00:00, 194.85it/s]
+
  0%|          | 0/63 [00:00<?, ?it/s]
 66%|██████▌   | 41/62 [00:00<00:00, 199.22it/s]
 32%|███▏      | 20/63 [00:00<00:00, 196.64it/s]
+
 63%|██████▎   | 40/63 [00:00<00:00, 197.73it/s]
 95%|█████████▌| 60/63 [00:00<00:00, 198.02it/s]
+
  0%|          | 0/63 [00:00<?, ?it/s]
 29%|██▊       | 18/63 [00:00<00:00, 176.35it/s]
 60%|██████    | 38/63 [00:00<00:00, 188.52it/s]
 92%|█████████▏| 58/63 [00:00<00:00, 192.50it/s]
+
  0%|          | 0/63 [00:00<?, ?it/s]
 32%|███▏      | 20/63 [00:00<00:00, 193.28it/s]
 63%|██████▎   | 40/63 [00:00<00:00, 195.90it/s]
 95%|█████████▌| 60/63 [00:00<00:00, 197.29it/s]
+
+2025-11-19 06:39:46.604 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-19 06:39:46.604 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-19 06:39:46.605 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-19 06:39:46.606 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'sigma', 'eot_token_id') are not supported by the generation strategy 'vanilla'.
+2025-11-19 06:39:46.607 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'add_bos_token', 'eot_token_id') are not supported by the generation strategy 'vanilla'.
+2025-11-19 06:39:46.613 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma') are not supported by the generation strategy 'vanilla'.
+2025-11-19 06:39:46.614 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'vanilla'.
+
+2025-11-19 07:50:55.778 | INFO     | __main__:main:81 - Throughput: 6.68 tokens/sec, Tokens per step: 0.87 tokens/step (full: 15.81 tokens/sec, 1.00 tokens/step), Latency: 64.90 s, Total time: 3990.63 s
+2025-11-19 07:50:55.803 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/dllm/mbpp/results.json
+2025-11-19 07:50:55.803 | INFO     | __main__:main:101 - eval time: 3990.63 seconds
+[rank0]:[W1119 07:50:56.058581021 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/.hydra/config.yaml

@@ -0,0 +1,54 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: dyna
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: null
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.HeatCache
+  rollout_p: 0.1
+  current_k: 32
+  num_rollout_layers: null
+  sigma: 10.0
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: gsm8k
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/.hydra/hydra.yaml

@@ -0,0 +1,168 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k
+    - hydra.mode=RUN
+    task:
+    - dataset.name=gsm8k
+    - model=dream-inst
+    - cache=heat
+    - cache.current_k=32
+    - cache.rollout_p=0.1
+    - generation=dyna
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache.current_k=32,cache.rollout_p=0.1,cache=heat,dataset.name=gsm8k,generation.gen_length=1024,generation.steps=1024,generation=dyna,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k
+    choices:
+      cache: heat
+      generation: dyna
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/.hydra/overrides.yaml

@@ -0,0 +1,10 @@
+- dataset.name=gsm8k
+- model=dream-inst
+- cache=heat
+- cache.current_k=32
+- cache.rollout_p=0.1
+- generation=dyna
+- generation.gen_length=1024
+- generation.steps=1024
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/eval.log

@@ -0,0 +1,41 @@
+[2025-11-19 01:01:25,028][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-19 01:01:34,671][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:34,672][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:34,736][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:34,736][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:34,745][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:34,745][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:35,036][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:35,036][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:35,146][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:35,146][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:35,507][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:35,508][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:36,429][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:36,429][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:37,173][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 01:01:37,173][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 01:01:45,968][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:45,968][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 5...
+[2025-11-19 01:01:46,129][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:46,130][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 6...
+[2025-11-19 01:01:46,432][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:46,433][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 2...
+[2025-11-19 01:01:47,228][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:47,229][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 7...
+[2025-11-19 01:01:47,349][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:47,349][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 4...
+[2025-11-19 01:01:50,407][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:50,407][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 0...
+[2025-11-19 01:01:52,592][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:52,593][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 3...
+[2025-11-19 01:01:53,960][lm_eval.evaluator][INFO] - gsm8k: Using gen_kwargs: {'until': ['Question:', '</s>', '<|im_end|>'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 01:01:53,961][lm_eval.api.task][INFO] - Building contexts for gsm8k on rank 1...
+[2025-11-19 01:01:57,903][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 01:01:57,903][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 01:01:57,903][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 01:01:57,903][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 01:01:57,903][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 01:01:57,903][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 01:01:57,903][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 01:01:57,904][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/results.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c65ec61f9d9cdf56e7b11fde503c84051a0e4952b2546e7350a3993df34c34a3
+size 13622720

outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/stderr.log

@@ -0,0 +1,63 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1119 01:01:11.003348596 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-19 01:01:19.008 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'dyna', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 256, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': 10.0}
+2025-11-19 01:01:19.009 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'rollout_p': 0.1, 'current_k': 32, 'sigma': 10.0}
+[W1119 01:01:21.010138023 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 01:01:21.683502785 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 01:01:23.239356235 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 01:01:23.367433167 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 01:01:24.077399338 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 01:01:24.115102464 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 01:01:24.135317722 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 01:01:24.152975229 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
  0%|          | 0/165 [00:00<?, ?it/s]
 15%|█▍        | 24/165 [00:00<00:00, 236.64it/s]
  0%|          | 0/165 [00:00<?, ?it/s]
 33%|███▎      | 55/165 [00:00<00:00, 276.52it/s]
 23%|██▎       | 38/165 [00:00<00:00, 372.06it/s]
 50%|█████     | 83/165 [00:00<00:00, 271.90it/s]
 46%|████▌     | 76/165 [00:00<00:00, 376.30it/s]
 67%|██████▋   | 111/165 [00:00<00:00, 238.36it/s]
  0%|          | 0/165 [00:00<?, ?it/s]
 69%|██████▉   | 114/165 [00:00<00:00, 345.60it/s]
 85%|████████▍ | 140/165 [00:00<00:00, 254.33it/s]
 24%|██▎       | 39/165 [00:00<00:00, 381.69it/s]
 92%|█████████▏| 152/165 [00:00<00:00, 357.59it/s]
+
+
 47%|████▋     | 78/165 [00:00<00:00, 378.82it/s]
 71%|███████   | 117/165 [00:00<00:00, 382.86it/s]
 95%|█████████▍| 156/165 [00:00<00:00, 380.81it/s]
+
  0%|          | 0/164 [00:00<?, ?it/s]
 23%|██▎       | 38/164 [00:00<00:00, 378.96it/s]
  0%|          | 0/165 [00:00<?, ?it/s]
 47%|████▋     | 77/164 [00:00<00:00, 383.29it/s]
 24%|██▎       | 39/165 [00:00<00:00, 381.01it/s]
 71%|███████   | 116/164 [00:00<00:00, 382.83it/s]
 48%|████▊     | 79/165 [00:00<00:00, 386.84it/s]
 95%|█████████▍| 155/164 [00:00<00:00, 383.20it/s]
 72%|███████▏  | 118/165 [00:00<00:00, 387.08it/s]
+
 95%|█████████▌| 157/165 [00:00<00:00, 386.93it/s]
+
  0%|          | 0/165 [00:00<?, ?it/s]
 12%|█▏        | 20/165 [00:00<00:00, 192.79it/s]
 34%|███▍      | 56/165 [00:00<00:00, 286.45it/s]
 58%|█████▊    | 95/165 [00:00<00:00, 330.03it/s]
 81%|████████  | 134/165 [00:00<00:00, 351.06it/s]
+
  0%|          | 0/165 [00:00<?, ?it/s]
 23%|██▎       | 38/165 [00:00<00:00, 371.37it/s]
 47%|████▋     | 77/165 [00:00<00:00, 381.96it/s]
 71%|███████   | 117/165 [00:00<00:00, 385.84it/s]
 95%|█████████▌| 157/165 [00:00<00:00, 388.45it/s]
+
  0%|          | 0/165 [00:00<?, ?it/s]
 24%|██▎       | 39/165 [00:00<00:00, 383.02it/s]
 47%|████▋     | 78/165 [00:00<00:00, 378.08it/s]
 71%|███████   | 117/165 [00:00<00:00, 381.32it/s]
 95%|█████████▍| 156/165 [00:00<00:00, 379.33it/s]
+
+2025-11-19 01:01:57.918 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'sigma', 'eot_token_id', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 01:01:57.919 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'block_length', 'sigma', 'eot_token_id') are not supported by the generation strategy 'dyna'.
+2025-11-19 01:01:57.920 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'sigma', 'block_length', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+2025-11-19 01:01:57.921 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'sigma', 'add_bos_token', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 01:01:57.922 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'add_bos_token', 'eot_token_id', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 01:01:57.922 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 01:01:57.927 | WARNING  | src.generation:generate:53 - The arguments ('block_length', 'sigma', 'add_bos_token', 'eot_token_id') are not supported by the generation strategy 'dyna'.
+
+2025-11-19 03:35:49.710 | INFO     | __main__:main:81 - Throughput: 2.28 tokens/sec, Tokens per step: 0.99 tokens/step (full: 19.12 tokens/sec, 1.00 tokens/step), Latency: 53.66 s, Total time: 8646.34 s
+2025-11-19 03:35:49.824 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/gsm8k/results.json
+2025-11-19 03:35:49.824 | INFO     | __main__:main:101 - eval time: 8646.34 seconds
+[rank0]:[W1119 03:35:50.020992224 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval/.hydra/config.yaml

@@ -0,0 +1,54 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: dyna
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: null
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.HeatCache
+  rollout_p: 0.1
+  current_k: 32
+  num_rollout_layers: null
+  sigma: 10.0
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: humaneval
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval/.hydra/hydra.yaml

@@ -0,0 +1,168 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval
+    - hydra.mode=RUN
+    task:
+    - dataset.name=humaneval
+    - model=dream-inst
+    - cache=heat
+    - cache.current_k=32
+    - cache.rollout_p=0.1
+    - generation=dyna
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache.current_k=32,cache.rollout_p=0.1,cache=heat,dataset.name=humaneval,generation.gen_length=1024,generation.steps=1024,generation=dyna,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval
+    choices:
+      cache: heat
+      generation: dyna
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval/.hydra/overrides.yaml

@@ -0,0 +1,10 @@
+- dataset.name=humaneval
+- model=dream-inst
+- cache=heat
+- cache.current_k=32
+- cache.rollout_p=0.1
+- generation=dyna
+- generation.gen_length=1024
+- generation.steps=1024
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval/eval.log

@@ -0,0 +1,41 @@
+[2025-11-18 11:57:27,511][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-18 11:57:50,836][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:50,836][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:51,428][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:51,428][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:51,507][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:51,507][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:51,531][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:51,531][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:51,599][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:51,599][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:51,849][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:51,849][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:52,077][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:52,077][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:52,094][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-18 11:57:52,094][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-18 11:57:54,019][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:54,021][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 0...
+[2025-11-18 11:57:55,009][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:55,011][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 5...
+[2025-11-18 11:57:55,040][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:55,041][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 6...
+[2025-11-18 11:57:55,060][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:55,061][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 1...
+[2025-11-18 11:57:55,341][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:55,343][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 2...
+[2025-11-18 11:57:55,498][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:55,499][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 4...
+[2025-11-18 11:57:55,593][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:55,594][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 3...
+[2025-11-18 11:57:55,597][lm_eval.evaluator][INFO] - humaneval: Using gen_kwargs: {'until': ['\nclass', '\ndef', '\n#', '\nif', '\nprint'], 'max_gen_toks': 1024, 'do_sample': False}
+[2025-11-18 11:57:55,598][lm_eval.api.task][INFO] - Building contexts for humaneval on rank 7...
+[2025-11-18 11:57:59,590][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:57:59,590][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:57:59,590][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:57:59,590][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:57:59,590][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:57:59,590][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:57:59,590][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-18 11:57:59,591][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval/stderr.log

@@ -0,0 +1,63 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1118 11:57:09.587421033 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-18 11:57:19.084 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'dyna', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 512, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': 10.0}
+2025-11-18 11:57:19.084 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'rollout_p': 0.1, 'current_k': 32, 'sigma': 10.0}
+[W1118 11:57:24.909665331 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:57:24.996962462 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:57:24.158513485 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:57:24.202768949 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:57:24.221986689 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:57:24.233998212 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:57:25.783812119 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1118 11:57:25.854676461 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
  0%|          | 0/21 [00:00<?, ?it/s]
+
  0%|          | 0/20 [00:00<?, ?it/s]
+
  0%|          | 0/20 [00:00<?, ?it/s]
+
  0%|          | 0/21 [00:00<?, ?it/s]
+
  0%|          | 0/21 [00:00<?, ?it/s]
+
  0%|          | 0/20 [00:00<?, ?it/s]
+
  0%|          | 0/21 [00:00<?, ?it/s]
  0%|          | 0/20 [00:00<?, ?it/s]
+
+
+2025-11-18 11:57:59.597 | WARNING  | src.generation:generate:53 - The arguments ('block_length', 'eot_token_id', 'add_bos_token', 'sigma') are not supported by the generation strategy 'dyna'.
+2025-11-18 11:57:59.599 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'sigma', 'eot_token_id', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-18 11:57:59.601 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'eot_token_id', 'sigma', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-18 11:57:59.602 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'block_length', 'sigma', 'eot_token_id') are not supported by the generation strategy 'dyna'.
+2025-11-18 11:57:59.603 | WARNING  | src.generation:generate:53 - The arguments ('block_length', 'add_bos_token', 'sigma', 'eot_token_id') are not supported by the generation strategy 'dyna'.
+2025-11-18 11:57:59.603 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'block_length', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+2025-11-18 11:57:59.605 | WARNING  | src.generation:generate:53 - The arguments ('block_length', 'eot_token_id', 'sigma', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+
+2025-11-18 12:16:38.734 | INFO     | __main__:main:81 - Throughput: 2.86 tokens/sec, Tokens per step: 0.99 tokens/step (full: 20.39 tokens/sec, 1.00 tokens/step), Latency: 50.24 s, Total time: 1055.85 s
+2025-11-18 12:16:38.745 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/humaneval/results.json
+2025-11-18 12:16:38.745 | INFO     | __main__:main:101 - eval time: 1055.85 seconds
+[rank0]:[W1118 12:16:39.004768498 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500/.hydra/config.yaml

@@ -0,0 +1,54 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: dyna
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: null
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.HeatCache
+  rollout_p: 0.1
+  current_k: 32
+  num_rollout_layers: null
+  sigma: 10.0
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: math-500
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500/.hydra/hydra.yaml

@@ -0,0 +1,168 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500
+    - hydra.mode=RUN
+    task:
+    - dataset.name=math-500
+    - model=dream-inst
+    - cache=heat
+    - cache.current_k=32
+    - cache.rollout_p=0.1
+    - generation=dyna
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache.current_k=32,cache.rollout_p=0.1,cache=heat,dataset.name=math-500,generation.gen_length=1024,generation.steps=1024,generation=dyna,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500
+    choices:
+      cache: heat
+      generation: dyna
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500/.hydra/overrides.yaml

@@ -0,0 +1,10 @@
+- dataset.name=math-500
+- model=dream-inst
+- cache=heat
+- cache.current_k=32
+- cache.rollout_p=0.1
+- generation=dyna
+- generation.gen_length=1024
+- generation.steps=1024
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500/eval.log

@@ -0,0 +1,41 @@
+[2025-11-19 13:22:56,217][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-19 13:23:06,661][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:06,661][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:07,528][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:07,529][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:07,673][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:07,674][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:07,776][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:07,776][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:07,851][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:07,851][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:07,914][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:07,914][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:07,964][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:07,964][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:09,917][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 13:23:09,917][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 13:23:11,870][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:11,871][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 2...
+[2025-11-19 13:23:13,115][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:13,118][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 1...
+[2025-11-19 13:23:13,142][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:13,144][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 4...
+[2025-11-19 13:23:13,160][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:13,162][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 0...
+[2025-11-19 13:23:13,450][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:13,451][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 6...
+[2025-11-19 13:23:13,458][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:13,459][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 3...
+[2025-11-19 13:23:13,552][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:13,553][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 5...
+[2025-11-19 13:23:16,828][lm_eval.evaluator][INFO] - math-500: Using gen_kwargs: {'until': ['Problem:'], 'do_sample': False, 'temperature': 0.0}
+[2025-11-19 13:23:16,830][lm_eval.api.task][INFO] - Building contexts for math-500 on rank 7...
+[2025-11-19 13:23:20,968][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 13:23:20,968][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 13:23:20,968][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 13:23:20,968][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 13:23:20,968][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 13:23:20,968][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 13:23:20,969][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 13:23:20,969][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500/stderr.log

@@ -0,0 +1,103 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1119 13:22:41.887213418 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-19 13:22:49.453 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'dyna', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 256, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': 10.0}
+2025-11-19 13:22:49.454 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'rollout_p': 0.1, 'current_k': 32, 'sigma': 10.0}
+[W1119 13:22:50.209196182 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 13:22:51.118344731 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 13:22:52.247364390 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 13:22:53.481626799 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 13:22:54.841878670 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 13:22:54.861399448 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 13:22:54.915315879 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 13:22:54.940491560 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+2025-11-19 13:23:02.814 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:02.819 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_i1m7kavk...
+2025-11-19 13:23:03.396 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:03.400 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_hgrcnp6w...
+2025-11-19 13:23:03.483 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:03.498 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_msyv5ata...
+2025-11-19 13:23:03.613 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:03.617 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_8t27kby9...
+2025-11-19 13:23:03.645 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:03.650 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_yffd15jm...
+2025-11-19 13:23:03.690 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 13:23:03.709 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:03.716 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_ous84by6...
+2025-11-19 13:23:03.764 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:03.768 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_5ko680x0...
+2025-11-19 13:23:04.277 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 13:23:04.330 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 13:23:04.469 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 13:23:04.557 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 13:23:04.598 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 13:23:04.648 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+2025-11-19 13:23:06.596 | INFO     | src.utils:sympy_antlr_patcher:462 - Detected antlr4-python3-runtime version 4.9.3. Temporarily switching to 4.11.0...
+2025-11-19 13:23:06.600 | INFO     | src.utils:sympy_antlr_patcher:473 - Downloading antlr4-python3-runtime==4.11.0 to /tmp/isolated_antlr_snuey55k...
+2025-11-19 13:23:07.558 | INFO     | src.utils:sympy_antlr_patcher:500 - Unpacking antlr4_python3_runtime-4.11.0-py3-none-any.whl...
+
  0%|          | 0/63 [00:00<?, ?it/s]
 89%|████████▉ | 56/63 [00:00<00:00, 550.79it/s]
+
  0%|          | 0/63 [00:00<?, ?it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
  0%|          | 0/63 [00:00<?, ?it/s]
 86%|████████▌ | 54/63 [00:00<00:00, 537.94it/s]
+
 89%|████████▊ | 55/62 [00:00<00:00, 545.71it/s]
+
 75%|███████▍  | 47/63 [00:00<00:00, 469.70it/s]
+
  0%|          | 0/62 [00:00<?, ?it/s]
  0%|          | 0/63 [00:00<?, ?it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
 87%|████████▋ | 54/62 [00:00<00:00, 539.38it/s]
 87%|████████▋ | 55/63 [00:00<00:00, 547.05it/s]
+
+
 87%|████████▋ | 54/62 [00:00<00:00, 539.81it/s]
+
  0%|          | 0/62 [00:00<?, ?it/s]
 90%|█████████ | 56/62 [00:00<00:00, 550.85it/s]
+
+2025-11-19 13:23:20.979 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'add_bos_token', 'eot_token_id', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 13:23:20.982 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 13:23:20.984 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'add_bos_token', 'block_length', 'eot_token_id') are not supported by the generation strategy 'dyna'.
+2025-11-19 13:23:20.986 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'block_length', 'sigma') are not supported by the generation strategy 'dyna'.
+2025-11-19 13:23:20.987 | WARNING  | src.generation:generate:53 - The arguments ('add_bos_token', 'block_length', 'sigma', 'eot_token_id') are not supported by the generation strategy 'dyna'.
+2025-11-19 13:23:20.993 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'block_length', 'add_bos_token', 'eot_token_id') are not supported by the generation strategy 'dyna'.
+2025-11-19 13:23:20.997 | WARNING  | src.generation:generate:53 - The arguments ('block_length', 'eot_token_id', 'sigma', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+
+2025-11-19 14:19:09.786 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:09.788 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:09.788 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:09.788 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:09.788 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:09.789 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:09.793 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:09.798 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:09.801 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:09.801 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:09.802 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:09.802 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:09.804 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:09.809 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:16.088 | INFO     | src.utils:sympy_antlr_patcher:513 - Restoring original environment...
+2025-11-19 14:19:16.101 | INFO     | src.utils:sympy_antlr_patcher:522 - Environment restored.
+2025-11-19 14:19:16.102 | INFO     | __main__:main:81 - Throughput: 19.64 tokens/sec, Tokens per step: 1.00 tokens/step (full: 19.93 tokens/sec, 1.00 tokens/step), Latency: 51.42 s, Total time: 3188.57 s
+2025-11-19 14:19:16.130 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/math-500/results.json
+2025-11-19 14:19:16.130 | INFO     | __main__:main:101 - eval time: 3188.57 seconds
+[rank0]:[W1119 14:19:16.383022793 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp/.hydra/config.yaml

@@ -0,0 +1,54 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: dyna
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: null
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+cache:
+  _target_: src.cache.HeatCache
+  rollout_p: 0.1
+  current_k: 32
+  num_rollout_layers: null
+  sigma: 10.0
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: mbpp
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}

outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp/.hydra/hydra.yaml

@@ -0,0 +1,168 @@
+hydra:
+  run:
+    dir: outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.run.dir=outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp
+    - hydra.mode=RUN
+    task:
+    - dataset.name=mbpp
+    - model=dream-inst
+    - cache=heat
+    - cache.current_k=32
+    - cache.rollout_p=0.1
+    - generation=dyna
+    - generation.gen_length=1024
+    - generation.steps=1024
+    - batch_size=1
+    - seed=1234
+  job:
+    name: eval
+    chdir: null
+    override_dirname: batch_size=1,cache.current_k=32,cache.rollout_p=0.1,cache=heat,dataset.name=mbpp,generation.gen_length=1024,generation.steps=1024,generation=dyna,model=dream-inst,seed=1234
+    id: ???
+    num: ???
+    config_name: eval
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /xfr_ceph_sh/liuchonghan/HEAT/heat
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /xfr_ceph_sh/liuchonghan/HEAT/heat/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp
+    choices:
+      cache: heat
+      generation: dyna
+      model: dream-inst
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp/.hydra/overrides.yaml

@@ -0,0 +1,10 @@
+- dataset.name=mbpp
+- model=dream-inst
+- cache=heat
+- cache.current_k=32
+- cache.rollout_p=0.1
+- generation=dyna
+- generation.gen_length=1024
+- generation.steps=1024
+- batch_size=1
+- seed=1234

outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp/eval.log

@@ -0,0 +1,41 @@
+[2025-11-19 07:51:25,028][accelerate.utils.other][WARNING] - Detected kernel version 5.4.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
+[2025-11-19 07:51:32,868][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:32,868][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:33,237][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:33,237][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:34,154][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:34,154][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:34,224][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:34,224][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:34,331][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:34,331][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:34,343][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:34,343][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:34,407][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:34,407][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:34,533][lm_eval.evaluator][INFO] - Setting random seed to 1234 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+[2025-11-19 07:51:34,533][lm_eval.evaluator][INFO] - Using pre-initialized model
+[2025-11-19 07:51:47,455][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:47,457][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 0...
+[2025-11-19 07:51:48,141][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:48,142][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 5...
+[2025-11-19 07:51:49,468][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:49,469][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 1...
+[2025-11-19 07:51:49,639][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:49,640][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 2...
+[2025-11-19 07:51:49,706][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:49,707][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 4...
+[2025-11-19 07:51:49,709][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:49,710][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 6...
+[2025-11-19 07:51:50,065][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:50,067][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 7...
+[2025-11-19 07:51:50,067][lm_eval.evaluator][INFO] - mbpp: Using gen_kwargs: {'until': ['[DONE]'], 'do_sample': False}
+[2025-11-19 07:51:50,068][lm_eval.api.task][INFO] - Building contexts for mbpp on rank 3...
+[2025-11-19 07:51:54,467][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 07:51:54,467][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 07:51:54,467][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 07:51:54,467][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 07:51:54,467][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 07:51:54,467][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 07:51:54,467][lm_eval.evaluator][INFO] - Running generate_until requests
+[2025-11-19 07:51:54,469][lm_eval.evaluator][INFO] - Running generate_until requests

outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp/stderr.log

@@ -0,0 +1,63 @@
+ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
+The following values were not passed to `accelerate launch` and had defaults used instead:
+		More than one GPU was found, enabling multi-GPU training.
+		If this was unintended please pass in `--num_processes=1`.
+	`--mixed_precision` was set to a value of `'no'`
+	`--dynamo_backend` was set to a value of `'no'`
+To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
+[W1119 07:51:11.683567084 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+2025-11-19 07:51:19.460 | INFO     | src.utils:pre_initialize:603 - {'strategy': 'dyna', 'threshold': None, 'factor': None, 'alg': 'maskgit_plus', 'gen_length': 1024, 'block_length': 512, 'steps': 1024, 'temperature': 0.0, 'top_p': 0.9, 'top_k': None, 'debias': False, 'output_probs': False, 'mask_token_id': 151666, 'eot_token_id': 151643, 'pad_token_id': 151643, 'add_bos_token': True, 'sigma': 10.0}
+2025-11-19 07:51:19.460 | INFO     | src.utils:pre_initialize:618 - Using cache with args: {'rollout_p': 0.1, 'current_k': 32, 'sigma': 10.0}
+[W1119 07:51:21.592036312 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 07:51:21.593016469 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 07:51:21.693916007 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 07:51:24.814455070 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 07:51:24.438362045 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 07:51:24.687729340 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 07:51:25.782100378 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+[W1119 07:51:25.816994034 socket.cpp:755] [c10d] The client socket cannot be initialized to connect to [localhost]:29500 (errno: 97 - Address family not supported by protocol).
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
+
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details.
+
  0%|          | 0/63 [00:00<?, ?it/s]
 32%|███▏      | 20/63 [00:00<00:00, 192.09it/s]
 63%|██████▎   | 40/63 [00:00<00:00, 195.01it/s]
 95%|█████████▌| 60/63 [00:00<00:00, 196.83it/s]
+
  0%|          | 0/62 [00:00<?, ?it/s]
 27%|██▋       | 17/62 [00:00<00:00, 166.68it/s]
 60%|█████▉    | 37/62 [00:00<00:00, 181.32it/s]
 92%|█████████▏| 57/62 [00:00<00:00, 189.04it/s]
+
  0%|          | 0/63 [00:00<?, ?it/s]
 32%|███▏      | 20/63 [00:00<00:00, 198.33it/s]
  0%|          | 0/63 [00:00<?, ?it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
  0%|          | 0/62 [00:00<?, ?it/s]
 63%|██████▎   | 40/63 [00:00<00:00, 152.60it/s]
 32%|███▏      | 20/63 [00:00<00:00, 197.48it/s]
 32%|███▏      | 20/62 [00:00<00:00, 197.64it/s]
 32%|███▏      | 20/62 [00:00<00:00, 188.25it/s]
 90%|█████████ | 57/63 [00:00<00:00, 156.83it/s]
 65%|██████▌   | 41/63 [00:00<00:00, 199.17it/s]
+
 65%|██████▍   | 40/62 [00:00<00:00, 198.45it/s]
 65%|██████▍   | 40/62 [00:00<00:00, 193.49it/s]
 98%|█████████▊| 62/63 [00:00<00:00, 199.65it/s]
+
 97%|█████████▋| 60/62 [00:00<00:00, 199.02it/s]
 97%|█████████▋| 60/62 [00:00<00:00, 195.92it/s]
+
+
  0%|          | 0/62 [00:00<?, ?it/s]
  0%|          | 0/63 [00:00<?, ?it/s]
 32%|███▏      | 20/62 [00:00<00:00, 198.11it/s]
 29%|██▊       | 18/63 [00:00<00:00, 173.29it/s]
 65%|██████▍   | 40/62 [00:00<00:00, 199.09it/s]
 59%|█████▊    | 37/63 [00:00<00:00, 178.84it/s]
 98%|█████████▊| 61/62 [00:00<00:00, 199.65it/s]
+
 90%|█████████ | 57/63 [00:00<00:00, 188.11it/s]
+
+2025-11-19 07:51:54.479 | WARNING  | src.generation:generate:53 - The arguments ('block_length', 'sigma', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+2025-11-19 07:51:54.484 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'block_length', 'sigma') are not supported by the generation strategy 'dyna'.
+2025-11-19 07:51:54.484 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'add_bos_token', 'sigma', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 07:51:54.484 | WARNING  | src.generation:generate:53 - The arguments ('eot_token_id', 'block_length', 'sigma', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+2025-11-19 07:51:54.485 | WARNING  | src.generation:generate:53 - The arguments ('block_length', 'eot_token_id', 'sigma', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+2025-11-19 07:51:54.486 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'eot_token_id', 'add_bos_token', 'block_length') are not supported by the generation strategy 'dyna'.
+2025-11-19 07:51:54.504 | WARNING  | src.generation:generate:53 - The arguments ('sigma', 'block_length', 'eot_token_id', 'add_bos_token') are not supported by the generation strategy 'dyna'.
+
+2025-11-19 08:49:24.011 | INFO     | __main__:main:81 - Throughput: 7.07 tokens/sec, Tokens per step: 0.93 tokens/step (full: 20.06 tokens/sec, 1.00 tokens/step), Latency: 51.12 s, Total time: 3208.85 s
+2025-11-19 08:49:24.040 | INFO     | __main__:main:98 - Results saved to /xfr_ceph_sh/liuchonghan/HEAT/heat/outputs/2025-11-18/exp1_len1024/dream-inst/heat/mbpp/results.json
+2025-11-19 08:49:24.041 | INFO     | __main__:main:101 - eval time: 3208.85 seconds
+[rank0]:[W1119 08:49:24.234766482 ProcessGroupNCCL.cpp:1479] 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())

outputs/2025-11-18/exp1_len1024/dream-inst/no_cache/gsm8k/.hydra/config.yaml

@@ -0,0 +1,48 @@
+model:
+  generation:
+    mask_token_id: 151666
+    eot_token_id: 151643
+    pad_token_id: 151643
+    add_bos_token: true
+    alg: maskgit_plus
+  name: dream-inst
+  path: ${oc.env:DREAM_INST_PATH}
+generation:
+  strategy: vanilla
+  threshold: null
+  factor: null
+  alg: maskgit_plus
+  gen_length: 1024
+  block_length: 32
+  steps: 1024
+  temperature: 0.0
+  top_p: null
+  top_k: null
+  debias: false
+  output_probs: false
+seed: 1234
+batch_size: 1
+attn_implementation: eager
+dataset:
+  name: gsm8k
+  size: null
+  n_shot: null
+  system_prompt: null
+  batch_size: 1
+mc_num: null
+max_length: 4096
+is_check_greedy: true
+add_bos_token: true
+nll_type: mc
+log_type: ftb
+eval_args:
+  log_samples: true
+  tasks: ${..dataset.name}
+  num_fewshot: ${..dataset.n_shot}
+  batch_size: ${..batch_size}
+  limit: ${..dataset.size}
+  confirm_run_unsafe_code: true
+  random_seed: ${..seed}
+  fewshot_random_seed: ${..seed}
+  numpy_random_seed: ${..seed}
+  torch_random_seed: ${..seed}