Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,4 @@ 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
+train_hnet_with_docstring_18_04/wandb/run-20260417_085757-sa79g3yl/run-sa79g3yl.wandb filter=lfs diff=lfs merge=lfs -text

train_hnet_with_docstring_18_04/.hydra/config.yaml

@@ -0,0 +1,54 @@
+model:
+  config_path: ${oc.env:PROJECT_ROOT}/hnet_project/configs/hnet_2stage_XL_code.json
+  checkpoint_path: ${oc.env:PROJECT_ROOT}/hnet_project/checkpoints/hnet_2stage_XL_code.pt
+training:
+  epochs: 3
+  batch_size: 4
+  eval_batch_size: 24
+  gradient_accumulation_steps: 4
+  lr: 0.0001
+  weight_decay: 0.1
+  betas:
+  - 0.9
+  - 0.95
+  eps: 1.0e-08
+  lr_scheduler: wsd
+  warmup_ratio: 0.1
+  decay_ratio: 0.2
+  warmup_steps: 100
+  min_lr_ratio: 0.1
+  lr_multiplier:
+  - 2.0
+  - 1.5
+  - 1.0
+  load_balancing_weight: 0.01
+  load_balancing_N: 4.0
+  max_grad_norm: 1.0
+  use_amp: true
+  resume: false
+  resume_checkpoint: null
+  warmup_model: true
+data:
+  path: ${oc.env:PROJECT_ROOT}/code_completion_exp/datasets/data_V5_full
+  max_context_len: 4096
+  max_target_len: 256
+  num_workers: 0
+  pin_memory: true
+  max_train_samples: null
+  max_val_samples: null
+logging:
+  log_interval: 10
+  save_interval: 3000
+  eval_interval: 1000
+  save_every_epoch: true
+tracking:
+  enabled: true
+  backend: wandb
+  project: code-completion-full-docstring
+  run_name: hnet_train
+  entity: null
+  base_url: https://wandb.platun0v.ru
+paths:
+  output_dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+seed: 42
+device: cuda

train_hnet_with_docstring_18_04/.hydra/hydra.yaml

@@ -0,0 +1,160 @@
+hydra:
+  run:
+    dir: ${paths.output_dir}
+  sweep:
+    dir: outputs/multirun/${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.mode=RUN
+    task: []
+  job:
+    name: train
+    chdir: false
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: config
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /workspace/byte-llms-code/code_completion_exp/train_hnet
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /workspace/byte-llms-code/code_completion_exp/train_hnet/configs
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /workspace/byte-llms-code/code_completion_exp/train_hnet/outputs/2026-04-17/08-57-56
+    choices:
+      paths: default
+      tracking: wandb
+      logging: default
+      data: default
+      training: default
+      model: hnet_xl_code
+      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

train_hnet_with_docstring_18_04/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

train_hnet_with_docstring_18_04/eval_results/.ipynb_checkpoints/metrics_checkpoint_step_3000-checkpoint.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_3000.pt
+================================================================================
+
+exact_match: 0.3265588423068738
+token_accuracy: 0.437939395023843
+bleu: 17.26934111421602
+bpb: 1.3647181750481843
+num_samples: 37592
+gen_wall_time_s: 1328.2266302730422
+gen_samples_per_s: 28.302398960539023
+gen_time_per_sample_ms: 35.332693931502504
+gen_chars_per_s: 731.4956482993198
+gen_batch_mean_ms: 1130.4056427855678
+gen_batch_p50_ms: 1147.103229071945
+gen_batch_p95_ms: 1343.7763461610302
+gen_batch_max_ms: 1544.4510788656771
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/.ipynb_checkpoints/metrics_initial_checkpoint-checkpoint.txt

@@ -0,0 +1,17 @@
+Checkpoint: hnet_2stage_XL_code.pt
+================================================================================
+
+exact_match: 0.00042562247286656737
+token_accuracy: 0.44197793742674885
+bleu: 4.782346760579283
+bpb: 1.9778437943839007
+num_samples: 37592
+gen_wall_time_s: 1962.1507909195498
+gen_samples_per_s: 19.15856832918674
+gen_time_per_sample_ms: 52.19596698551686
+gen_chars_per_s: 1225.6407668204554
+gen_batch_mean_ms: 1662.1550524910476
+gen_batch_p50_ms: 1646.254621911794
+gen_batch_p95_ms: 1715.8207091037184
+gen_batch_max_ms: 20510.621909983456
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/eval_config.yaml

@@ -0,0 +1,29 @@
+model:
+  config_path: /workspace/byte-llms-code/hnet_project/configs/hnet_2stage_XL_code.json
+  checkpoint_path: /workspace/byte-llms-code/hnet_project/checkpoints/hnet_2stage_XL_code.pt
+data:
+  path: /workspace/byte-llms-code/code_completion_exp/datasets/data_V5_full
+  max_context_len: 4096
+  max_target_len: 256
+  num_workers: 0
+  pin_memory: true
+  max_train_samples: null
+  max_val_samples: null
+paths:
+  checkpoints_dir: outputs/2026-04-17/08-57-56
+  initial_checkpoint: auto
+  output_dir: outputs/2026-04-17/08-57-56/eval_results
+evaluation:
+  batch_size: 16
+  max_samples: null
+  compute_bpb: true
+  bleu_tokenize: none
+  use_amp: true
+  save_predictions: true
+generation:
+  max_length: 64
+  temperature: 0.1
+  top_k: 0
+  top_p: 1.0
+seed: 42
+device: cuda

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_latest.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_latest.pt
+================================================================================
+
+exact_match: 0.3314535007448393
+token_accuracy: 0.4422865987313573
+bleu: 17.390015716696666
+bpb: 1.3799796338028427
+num_samples: 37592
+gen_wall_time_s: 1371.5999331497587
+gen_samples_per_s: 27.407408743213683
+gen_time_per_sample_ms: 36.48648470817618
+gen_chars_per_s: 727.3775507621499
+gen_batch_mean_ms: 1162.65390314459
+gen_batch_p50_ms: 1180.899173952639
+gen_batch_p95_ms: 1378.4097837517038
+gen_batch_max_ms: 1539.6567089483142
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_10591.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_10591.pt
+================================================================================
+
+exact_match: 0.3369067886784422
+token_accuracy: 0.4435060345521727
+bleu: 18.150587601970468
+bpb: 1.3515752053962538
+num_samples: 37592
+gen_wall_time_s: 1318.3035029922612
+gen_samples_per_s: 28.515436630999133
+gen_time_per_sample_ms: 35.06872480826403
+gen_chars_per_s: 729.9859234354543
+gen_batch_mean_ms: 1121.9604280785202
+gen_batch_p50_ms: 1138.8398550916463
+gen_batch_p95_ms: 1339.150638342835
+gen_batch_max_ms: 1520.291404100135
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_12000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_12000.pt
+================================================================================
+
+exact_match: 0.32953819961693975
+token_accuracy: 0.44194036126792696
+bleu: 17.409712977551532
+bpb: 1.3706288980495067
+num_samples: 37592
+gen_wall_time_s: 1358.1111189327203
+gen_samples_per_s: 27.67962022838153
+gen_time_per_sample_ms: 36.127663304232826
+gen_chars_per_s: 737.8254886739063
+gen_batch_mean_ms: 1155.8392501555065
+gen_batch_p50_ms: 1174.8729590326548
+gen_batch_p95_ms: 1358.717922004871
+gen_batch_max_ms: 1544.2728700581938
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_15000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_15000.pt
+================================================================================
+
+exact_match: 0.33560331985528835
+token_accuracy: 0.445342792982205
+bleu: 17.80188542846622
+bpb: 1.3634364831092445
+num_samples: 37592
+gen_wall_time_s: 1354.7123250523582
+gen_samples_per_s: 27.749064731176126
+gen_time_per_sample_ms: 36.03725061322511
+gen_chars_per_s: 733.6996804554685
+gen_batch_mean_ms: 1151.095885965141
+gen_batch_p50_ms: 1165.4436129610986
+gen_batch_p95_ms: 1363.1549226120112
+gen_batch_max_ms: 1552.3370699957013
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_18000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_18000.pt
+================================================================================
+
+exact_match: 0.33656097041923816
+token_accuracy: 0.44532847825503474
+bleu: 17.75754919808768
+bpb: 1.3630499140307977
+num_samples: 37592
+gen_wall_time_s: 1351.9282269885298
+gen_samples_per_s: 27.806209863475942
+gen_time_per_sample_ms: 35.963189694310756
+gen_chars_per_s: 733.1284162975089
+gen_batch_mean_ms: 1150.5772144583232
+gen_batch_p50_ms: 1165.5358579009771
+gen_batch_p95_ms: 1360.9706948278472
+gen_batch_max_ms: 1532.5795689132065
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_21000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_21000.pt
+================================================================================
+
+exact_match: 0.33586933390082996
+token_accuracy: 0.4456639796730874
+bleu: 17.890089292040738
+bpb: 1.3623578011160296
+num_samples: 37592
+gen_wall_time_s: 1358.3291775202379
+gen_samples_per_s: 27.675176696585325
+gen_time_per_sample_ms: 36.133463968935885
+gen_chars_per_s: 732.6714440580971
+gen_batch_mean_ms: 1154.565481636277
+gen_batch_p50_ms: 1167.435775976628
+gen_batch_p95_ms: 1375.0762917567044
+gen_batch_max_ms: 1508.0128121189773
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_21182.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_21182.pt
+================================================================================
+
+exact_match: 0.3359757395190466
+token_accuracy: 0.4453508450162383
+bleu: 17.700113254581588
+bpb: 1.3621293462625146
+num_samples: 37592
+gen_wall_time_s: 1357.277649092488
+gen_samples_per_s: 27.696617582360552
+gen_time_per_sample_ms: 36.10549183582911
+gen_chars_per_s: 731.3654657642988
+gen_batch_mean_ms: 1148.2316075352596
+gen_batch_p50_ms: 1163.495829096064
+gen_batch_p95_ms: 1363.2275794865564
+gen_batch_max_ms: 1515.0637179613113
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_24000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_24000.pt
+================================================================================
+
+exact_match: 0.3315333049585018
+token_accuracy: 0.4435024558703801
+bleu: 17.3520067636136
+bpb: 1.3781568520216898
+num_samples: 37592
+gen_wall_time_s: 1368.3171897311695
+gen_samples_per_s: 27.473162130913245
+gen_time_per_sample_ms: 36.39915912245077
+gen_chars_per_s: 733.2883102909277
+gen_batch_mean_ms: 1161.236023812535
+gen_batch_p50_ms: 1180.4353760089725
+gen_batch_p95_ms: 1372.9352780152112
+gen_batch_max_ms: 1518.2082359679043
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_27000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_27000.pt
+================================================================================
+
+exact_match: 0.3306820600127687
+token_accuracy: 0.44246642749143356
+bleu: 17.28689580869071
+bpb: 1.379878546557371
+num_samples: 37592
+gen_wall_time_s: 1372.4666455930565
+gen_samples_per_s: 27.390100969452796
+gen_time_per_sample_ms: 36.50954047651246
+gen_chars_per_s: 729.034838997957
+gen_batch_mean_ms: 1163.5719769113161
+gen_batch_p50_ms: 1181.47262907587
+gen_batch_p95_ms: 1373.7670538946986
+gen_batch_max_ms: 1530.0602258648723
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_3000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_3000.pt
+================================================================================
+
+exact_match: 0.3265588423068738
+token_accuracy: 0.437939395023843
+bleu: 17.26934111421602
+bpb: 1.3647181750481843
+num_samples: 37592
+gen_wall_time_s: 1328.2266302730422
+gen_samples_per_s: 28.302398960539023
+gen_time_per_sample_ms: 35.332693931502504
+gen_chars_per_s: 731.4956482993198
+gen_batch_mean_ms: 1130.4056427855678
+gen_batch_p50_ms: 1147.103229071945
+gen_batch_p95_ms: 1343.7763461610302
+gen_batch_max_ms: 1544.4510788656771
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_30000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_30000.pt
+================================================================================
+
+exact_match: 0.3316663119812726
+token_accuracy: 0.4430926968051318
+bleu: 17.327071481779033
+bpb: 1.380080428458689
+num_samples: 37592
+gen_wall_time_s: 1370.579366396647
+gen_samples_per_s: 27.427816966800037
+gen_time_per_sample_ms: 36.45933619910212
+gen_chars_per_s: 731.2644744061805
+gen_batch_mean_ms: 1166.450524592891
+gen_batch_p50_ms: 1182.7268451452255
+gen_batch_p95_ms: 1375.3103922354057
+gen_batch_max_ms: 1513.1844920106232
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_31773.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_31773.pt
+================================================================================
+
+exact_match: 0.3321451372632475
+token_accuracy: 0.44243600869619676
+bleu: 17.34043772632883
+bpb: 1.3799796338028427
+num_samples: 37592
+gen_wall_time_s: 1366.8325954594184
+gen_samples_per_s: 27.50300228782927
+gen_time_per_sample_ms: 36.359666829629134
+gen_chars_per_s: 729.8084661675138
+gen_batch_mean_ms: 1163.2617833697177
+gen_batch_p50_ms: 1180.2458260208368
+gen_batch_p95_ms: 1379.442374384962
+gen_batch_max_ms: 1505.9127290733159
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_6000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_6000.pt
+================================================================================
+
+exact_match: 0.3314268993402852
+token_accuracy: 0.44133109069274334
+bleu: 17.71897114578991
+bpb: 1.3570473840195383
+num_samples: 37592
+gen_wall_time_s: 1331.4521436940413
+gen_samples_per_s: 28.233834898266075
+gen_time_per_sample_ms: 35.41849711890938
+gen_chars_per_s: 731.7867221999805
+gen_batch_mean_ms: 1133.1507605906734
+gen_batch_p50_ms: 1147.1740510314703
+gen_batch_p95_ms: 1350.74239235837
+gen_batch_max_ms: 1578.9136730600148
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_checkpoint_step_9000.txt

@@ -0,0 +1,17 @@
+Checkpoint: checkpoint_step_9000.pt
+================================================================================
+
+exact_match: 0.3374388167695254
+token_accuracy: 0.44255768387714384
+bleu: 18.095023606139048
+bpb: 1.3526477022435122
+num_samples: 37592
+gen_wall_time_s: 1328.4889196834993
+gen_samples_per_s: 28.296811093430843
+gen_time_per_sample_ms: 35.339671198220344
+gen_chars_per_s: 728.5706231035733
+gen_batch_mean_ms: 1130.6288678157441
+gen_batch_p50_ms: 1148.1934499461204
+gen_batch_p95_ms: 1352.066579903476
+gen_batch_max_ms: 1580.2085960749537
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_initial_checkpoint.txt

@@ -0,0 +1,17 @@
+Checkpoint: hnet_2stage_XL_code.pt
+================================================================================
+
+exact_match: 0.00042562247286656737
+token_accuracy: 0.44197793742674885
+bleu: 4.782346760579283
+bpb: 1.9778437943839007
+num_samples: 37592
+gen_wall_time_s: 1962.1507909195498
+gen_samples_per_s: 19.15856832918674
+gen_time_per_sample_ms: 52.19596698551686
+gen_chars_per_s: 1225.6407668204554
+gen_batch_mean_ms: 1662.1550524910476
+gen_batch_p50_ms: 1646.254621911794
+gen_batch_p95_ms: 1715.8207091037184
+gen_batch_max_ms: 20510.621909983456
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_model_best.txt

@@ -0,0 +1,17 @@
+Checkpoint: model_best.pt
+================================================================================
+
+exact_match: 0.3375186209831879
+token_accuracy: 0.4429110787041593
+bleu: 18.031857027994704
+bpb: 1.3521020891823154
+num_samples: 37592
+gen_wall_time_s: 1318.3916445451323
+gen_samples_per_s: 28.51353022111262
+gen_time_per_sample_ms: 35.07106949736998
+gen_chars_per_s: 733.5983992326425
+gen_batch_mean_ms: 1122.03544216607
+gen_batch_p50_ms: 1138.4079209528863
+gen_batch_p95_ms: 1340.2311942540107
+gen_batch_max_ms: 1512.969312025234
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/metrics_model_final.txt

@@ -0,0 +1,17 @@
+Checkpoint: model_final.pt
+================================================================================
+
+exact_match: 0.3311342838901894
+token_accuracy: 0.442775088796042
+bleu: 17.367347972897164
+bpb: 1.3799796338028427
+num_samples: 37592
+gen_wall_time_s: 1363.505116418004
+gen_samples_per_s: 27.570120234499786
+gen_time_per_sample_ms: 36.27115121350298
+gen_chars_per_s: 731.4141971244836
+gen_batch_mean_ms: 1159.4888140699093
+gen_batch_p50_ms: 1178.2626849599183
+gen_batch_p95_ms: 1370.1597674982622
+gen_batch_max_ms: 1537.723605055362
+gen_num_batches: 1175

train_hnet_with_docstring_18_04/eval_results/summary.txt

@@ -0,0 +1,22 @@
+EVALUATION SUMMARY
+==================================================================================================
+
+Checkpoint                      Exact Match    Token Acc     BLEU          BPB  ms/sample   samp/s
+--------------------------------------------------------------------------------------------------
+initial_checkpoint                    0.04%       44.20%     4.78         1.98       52.2    19.16
+checkpoint_latest                    33.15%       44.23%    17.39         1.38       36.5    27.41
+checkpoint_step_10591                33.69%       44.35%    18.15         1.35       35.1    28.52
+checkpoint_step_12000                32.95%       44.19%    17.41         1.37       36.1    27.68
+checkpoint_step_15000                33.56%       44.53%    17.80         1.36       36.0    27.75
+checkpoint_step_18000                33.66%       44.53%    17.76         1.36       36.0    27.81
+checkpoint_step_21000                33.59%       44.57%    17.89         1.36       36.1    27.68
+checkpoint_step_21182                33.60%       44.54%    17.70         1.36       36.1    27.70
+checkpoint_step_24000                33.15%       44.35%    17.35         1.38       36.4    27.47
+checkpoint_step_27000                33.07%       44.25%    17.29         1.38       36.5    27.39
+checkpoint_step_3000                 32.66%       43.79%    17.27         1.36       35.3    28.30
+checkpoint_step_30000                33.17%       44.31%    17.33         1.38       36.5    27.43
+checkpoint_step_31773                33.21%       44.24%    17.34         1.38       36.4    27.50
+checkpoint_step_6000                 33.14%       44.13%    17.72         1.36       35.4    28.23
+checkpoint_step_9000                 33.74%       44.26%    18.10         1.35       35.3    28.30
+model_best                           33.75%       44.29%    18.03         1.35       35.1    28.51
+model_final                          33.11%       44.28%    17.37         1.38       36.3    27.57

train_hnet_with_docstring_18_04/model_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e55307aa30def4fd79bd523e9cb1e50e109f3fe99acb2d8651ed405569a999e5
+size 3315165139

train_hnet_with_docstring_18_04/model_final.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cdefe95504a2ed0d743929eb0d1af737bda1be80273a53c6672c687a0197762
+size 3315165484

train_hnet_with_docstring_18_04/wandb/run-20260417_085757-sa79g3yl/files/requirements.txt

@@ -0,0 +1,245 @@
+setuptools==78.1.1
+wheel==0.45.1
+pip==25.2
+webencodings==0.5.1
+triton==3.2.0
+pytz==2025.2
+pydub==0.25.1
+pure_eval==0.2.3
+ptyprocess==0.7.0
+nvidia-ml-py==13.590.48
+nvidia-cusparselt-cu12==0.6.2
+mpmath==1.3.0
+ipython-genutils==0.2.0
+fastjsonschema==2.21.2
+brotli==1.2.0
+antlr4-python3-runtime==4.9.3
+xxhash==3.6.0
+widgetsnbextension==4.0.14
+websocket-client==1.9.0
+webcolors==24.11.1
+wcwidth==0.2.14
+urllib3==2.5.0
+uri-template==1.3.0
+tzdata==2025.2
+typing_extensions==4.15.0
+types-python-dateutil==2.9.0.20251008
+traitlets==5.14.3
+tqdm==4.67.1
+tornado==6.5.2
+tomlkit==0.13.3
+tinycss2==1.4.0
+tabulate==0.9.0
+sympy==1.13.1
+soupsieve==2.8
+sniffio==1.3.1
+smmap==5.0.2
+six==1.17.0
+shellingham==1.5.4
+Send2Trash==1.8.3
+semantic-version==2.10.0
+safetensors==0.6.2
+rpds-py==0.27.1
+rfc3986-validator==0.1.1
+regex==2025.9.18
+pyzmq==27.1.0
+PyYAML==6.0.3
+python-multipart==0.0.22
+python-json-logger==4.0.0
+python-dotenv==1.2.1
+pyparsing==3.2.5
+PyJWT==2.8.0
+Pygments==2.19.2
+pycparser==2.23
+pyarrow==22.0.0
+psutil==7.1.0
+protobuf==6.33.4
+propcache==0.4.1
+prometheus_client==0.23.1
+portalocker==3.2.0
+platformdirs==4.5.0
+pillow==11.3.0
+pexpect==4.9.0
+pathspec==1.0.4
+parso==0.8.5
+pandocfilters==1.5.1
+packaging==25.0
+orjson==3.11.6
+opt_einsum==3.4.0
+nvidia-nvtx-cu12==12.4.127
+nvidia-nvjitlink-cu12==12.4.127
+nvidia-nccl-cu12==2.21.5
+nvidia-curand-cu12==10.3.5.147
+nvidia-cufile-cu12==1.13.1.3
+nvidia-cufft-cu12==11.2.1.3
+nvidia-cuda-runtime-cu12==12.4.127
+nvidia-cuda-nvrtc-cu12==12.4.127
+nvidia-cuda-cupti-cu12==12.4.127
+nvidia-cublas-cu12==12.4.5.8
+numpy==2.3.3
+ninja==1.13.0
+networkx==3.5
+nest-asyncio==1.6.0
+narwhals==2.15.0
+mypy_extensions==1.1.0
+multidict==6.7.0
+mistune==3.1.4
+mdurl==0.1.2
+MarkupSafe==3.0.3
+lxml==6.0.2
+librt==0.8.0
+lark==1.3.0
+kiwisolver==1.4.9
+jupyterlab_widgets==3.0.15
+jupyterlab_pygments==0.3.0
+jsonpointer==3.0.0
+json5==0.12.1
+itsdangerous==2.2.0
+idna==3.10
+hf-xet==1.1.10
+h11==0.16.0
+groovy==0.1.2
+fsspec==2025.9.0
+frozenlist==1.8.0
+fqdn==1.5.1
+fonttools==4.60.1
+filelock==3.19.1
+ffmpy==1.0.0
+executing==2.2.1
+einops==0.8.1
+dill==0.4.0
+defusedxml==0.7.1
+decorator==5.2.1
+debugpy==1.8.17
+dacite==1.9.2
+cycler==0.12.1
+comm==0.2.3
+colorama==0.4.6
+click==8.3.1
+charset-normalizer==3.4.3
+certifi==2025.10.5
+bleach==6.2.0
+babel==2.17.0
+attrs==25.4.0
+async-lru==2.0.5
+asttokens==3.0.0
+annotated-types==0.7.0
+annotated-doc==0.0.4
+aiohappyeyeballs==2.6.1
+aiofiles==24.1.0
+yarl==1.22.0
+uvicorn==0.40.0
+typing-inspection==0.4.2
+terminado==0.18.1
+stack-data==0.6.3
+sentry-sdk==2.50.0
+scipy==1.17.0
+sacrebleu==2.6.0
+rfc3987-syntax==1.1.0
+rfc3339-validator==0.1.4
+requests==2.32.5
+reportlab==4.4.9
+referencing==0.36.2
+python-dateutil==2.9.0.post0
+pydantic_core==2.41.5
+prompt_toolkit==3.0.52
+plotly==6.5.2
+pathlib2==2.3.7.post1
+orderedmultidict==1.0.2
+optree==0.17.0
+omegaconf==2.3.0
+nvidia-cusparse-cu12==12.3.1.170
+nvidia-cudnn-cu12==9.1.0.70
+mypy==1.19.1
+multiprocess==0.70.16
+matplotlib-inline==0.1.7
+markdown-it-py==4.0.0
+jupyter_core==5.8.1
+Jinja2==3.1.6
+jedi==0.19.2
+ipython_pygments_lexers==1.1.1
+httpcore==1.0.9
+gitdb==4.0.12
+ftfy==6.3.1
+contourpy==1.3.3
+cffi==2.0.0
+beautifulsoup4==4.14.2
+anyio==4.11.0
+aiosignal==1.4.0
+starlette==0.50.0
+rich==14.2.0
+pydantic==2.12.5
+pandas==2.3.3
+nvidia-cusolver-cu12==11.6.1.9
+matplotlib==3.10.7
+jupyter_server_terminals==0.5.3
+jupyter_client==8.6.3
+jsonschema-specifications==2025.9.1
+ipython==9.6.0
+hydra-core==1.3.2
+huggingface-hub==0.35.3
+httpx==0.28.1
+GitPython==3.1.46
+furl==2.1.4
+cryptography==46.0.4
+arrow==1.3.0
+argon2-cffi-bindings==25.1.0
+aiohttp==3.13.1
+wandb==0.24.0
+typer==0.21.1
+torch==2.6.0
+tokenizers==0.22.1
+seaborn==0.13.2
+safehttpx==0.1.7
+jsonschema==4.25.1
+joypy==0.2.6
+isoduration==20.11.0
+ipywidgets==8.1.7
+ipykernel==6.30.1
+gradio_client==2.0.3
+fastapi==0.128.0
+Authlib==1.6.6
+argon2-cffi==25.1.0
+transformers==4.57.6
+nbformat==5.10.4
+mlstm_kernels==2.0.2
+jupyter-console==6.6.3
+gradio==6.5.1
+datasets==4.3.0
+clearml==1.16.4
+accelerate==1.10.1
+xlstm==2.0.4
+nbclient==0.10.2
+jupyter-events==0.12.0
+trackio==0.15.0
+nbconvert==7.16.6
+jupyter_server==2.17.0
+notebook_shim==0.2.4
+jupyterlab_server==2.27.3
+jupyter-lsp==2.3.0
+nbclassic==1.3.3
+jupyterlab==4.4.9
+notebook==7.4.7
+jupyter_contrib_core==0.4.2
+jupyter==1.1.1
+jupyter_nbextensions_configurator==0.6.4
+causal-conv1d==1.5.0.post8
+flash_attn==2.7.4.post1
+mamba-ssm==2.2.4
+hnet==0.0.1
+autocommand==2.2.2
+backports.tarfile==1.2.0
+importlib_metadata==8.0.0
+inflect==7.3.1
+jaraco.collections==5.1.0
+jaraco.context==5.3.0
+jaraco.functools==4.0.1
+jaraco.text==3.12.1
+more-itertools==10.3.0
+packaging==24.2
+platformdirs==4.2.2
+tomli==2.0.1
+typeguard==4.3.0
+typing_extensions==4.12.2
+wheel==0.45.1
+zipp==3.19.2

train_hnet_with_docstring_18_04/wandb/run-20260417_085757-sa79g3yl/files/wandb-metadata.json

@@ -0,0 +1 @@
+{"os": "Linux-5.15.0-173-generic-x86_64-with-glibc2.39", "python": "CPython 3.12.0", "started_at": "2026-04-17T08:57:57.464191Z", "program": "/workspace/byte-llms-code/code_completion_exp/train_hnet/train.py", "code_path": "code_completion_exp/train_hnet/train.py", "code_path_local": "train.py", "git": {"remote_url": "https://github.com/naryst/byte-llms-code.git", "commit": "056a135fbb34bc28ed3adfeeb2f4ac97cbf12a89"}, "email": "nikita@local.ru", "root": "/workspace/byte-llms-code/code_completion_exp/train_hnet", "host": "3e675e030992", "executable": "/venv/bytellm/bin/python", "cpu_count": 112, "cpu_count_logical": 224, "gpu_type": "NVIDIA H100 80GB HBM3", "gpu_count": 2, "disk": {"/": {"total": "244813135872", "used": "36382741504"}}, "memory": {"total": "1622968434688"}, "gpu_nvidia": [{"name": "NVIDIA H100 80GB HBM3", "memory_total": "85520809984", "cuda_cores": 16896, "architecture": "Hopper", "uuid": "GPU-3c87d2f8-c595-49bd-bb1d-1ebfd19c6fb0"}, {"name": "NVIDIA H100 80GB HBM3", "memory_total": "85520809984", "cuda_cores": 16896, "architecture": "Hopper", "uuid": "GPU-beb9a6b0-ebef-1f4c-d886-465c96f57ca4"}], "cuda_version": "12.9", "writer_id": "1enfm68bplbg421e1aqnc3guby2j3hk6"}

train_hnet_with_docstring_18_04/wandb/run-20260417_085757-sa79g3yl/logs/debug-internal.log

@@ -0,0 +1,15 @@
+{"time":"2026-04-17T08:57:57.764884416Z","level":"INFO","msg":"stream: starting","core version":"0.24.0"}
+{"time":"2026-04-17T08:57:58.240853232Z","level":"INFO","msg":"stream: created new stream","id":"sa79g3yl"}
+{"time":"2026-04-17T08:57:58.241001941Z","level":"INFO","msg":"handler: started","stream_id":"sa79g3yl"}
+{"time":"2026-04-17T08:57:58.241279701Z","level":"INFO","msg":"stream: started","id":"sa79g3yl"}
+{"time":"2026-04-17T08:57:58.241368206Z","level":"INFO","msg":"writer: started","stream_id":"sa79g3yl"}
+{"time":"2026-04-17T08:57:58.241360163Z","level":"INFO","msg":"sender: started","stream_id":"sa79g3yl"}
+{"time":"2026-04-17T08:57:58.392376272Z","level":"ERROR","msg":"git repo not found","error":"repository does not exist"}
+{"time":"2026-04-17T12:02:07.081755663Z","level":"ERROR","msg":"api: HTTP error","status":403,"method":"POST","url":"https://wandb.platun0v.ru/files/nikita/code-completion-full-docstring/sa79g3yl/file_stream"}
+{"time":"2026-04-17T12:02:07.082162122Z","level":"ERROR+4","msg":"filestream: fatal error: filestream: failed to upload: 403 Forbidden url=https://wandb.platun0v.ru/files/nikita/code-completion-full-docstring/sa79g3yl/file_stream: "}
+{"time":"2026-04-17T15:45:07.499206911Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-04-17T15:45:07.50174734Z","level":"INFO","msg":"handler: operation stats","stats":{}}
+{"time":"2026-04-17T15:45:07.50517879Z","level":"INFO","msg":"stream: closing","id":"sa79g3yl"}
+{"time":"2026-04-17T15:45:07.505192348Z","level":"INFO","msg":"handler: closed","stream_id":"sa79g3yl"}
+{"time":"2026-04-17T15:45:07.505334051Z","level":"INFO","msg":"sender: closed","stream_id":"sa79g3yl"}
+{"time":"2026-04-17T15:45:07.505346596Z","level":"INFO","msg":"stream: closed","id":"sa79g3yl"}

train_hnet_with_docstring_18_04/wandb/run-20260417_085757-sa79g3yl/logs/debug.log

@@ -0,0 +1,24 @@
+2026-04-17 08:57:57,465 INFO    MainThread:14733 [wandb_setup.py:_flush():81] Current SDK version is 0.24.0
+2026-04-17 08:57:57,465 INFO    MainThread:14733 [wandb_setup.py:_flush():81] Configure stats pid to 14733
+2026-04-17 08:57:57,465 INFO    MainThread:14733 [wandb_setup.py:_flush():81] Loading settings from environment variables
+2026-04-17 08:57:57,465 INFO    MainThread:14733 [wandb_init.py:setup_run_log_directory():717] Logging user logs to /workspace/byte-llms-code/code_completion_exp/train_hnet/wandb/run-20260417_085757-sa79g3yl/logs/debug.log
+2026-04-17 08:57:57,466 INFO    MainThread:14733 [wandb_init.py:setup_run_log_directory():718] Logging internal logs to /workspace/byte-llms-code/code_completion_exp/train_hnet/wandb/run-20260417_085757-sa79g3yl/logs/debug-internal.log
+2026-04-17 08:57:57,466 INFO    MainThread:14733 [wandb_init.py:init():844] calling init triggers
+2026-04-17 08:57:57,466 INFO    MainThread:14733 [wandb_init.py:init():849] wandb.init called with sweep_config: {}
+config: {'model': {'config_path': '/workspace/byte-llms-code/hnet_project/configs/hnet_2stage_XL_code.json', 'checkpoint_path': '/workspace/byte-llms-code/hnet_project/checkpoints/hnet_2stage_XL_code.pt'}, 'training': {'epochs': 3, 'batch_size': 4, 'eval_batch_size': 24, 'gradient_accumulation_steps': 4, 'lr': 0.0001, 'weight_decay': 0.1, 'betas': [0.9, 0.95], 'eps': 1e-08, 'lr_scheduler': 'wsd', 'warmup_ratio': 0.1, 'decay_ratio': 0.2, 'warmup_steps': 100, 'min_lr_ratio': 0.1, 'lr_multiplier': [2.0, 1.5, 1.0], 'load_balancing_weight': 0.01, 'load_balancing_N': 4.0, 'max_grad_norm': 1.0, 'use_amp': True, 'resume': False, 'resume_checkpoint': None, 'warmup_model': True}, 'data': {'path': '/workspace/byte-llms-code/code_completion_exp/datasets/data_V5_full', 'max_context_len': 4096, 'max_target_len': 256, 'num_workers': 0, 'pin_memory': True, 'max_train_samples': None, 'max_val_samples': None}, 'logging': {'log_interval': 10, 'save_interval': 3000, 'eval_interval': 1000, 'save_every_epoch': True}, 'tracking': {'enabled': True, 'backend': 'wandb', 'project': 'code-completion-full-docstring', 'run_name': 'hnet_train', 'entity': None, 'base_url': 'https://wandb.platun0v.ru'}, 'paths': {'output_dir': 'outputs/2026-04-17/08-57-56'}, 'seed': 42, 'device': 'cuda', '_wandb': {'code_path': 'code/code_completion_exp/train_hnet/train.py'}}
+2026-04-17 08:57:57,466 INFO    MainThread:14733 [wandb_init.py:init():892] starting backend
+2026-04-17 08:57:57,736 INFO    MainThread:14733 [wandb_init.py:init():895] sending inform_init request
+2026-04-17 08:57:57,761 INFO    MainThread:14733 [wandb_init.py:init():903] backend started and connected
+2026-04-17 08:57:57,767 INFO    MainThread:14733 [wandb_init.py:init():973] updated telemetry
+2026-04-17 08:57:57,799 INFO    MainThread:14733 [wandb_init.py:init():997] communicating run to backend with 90.0 second timeout
+2026-04-17 08:57:58,390 INFO    MainThread:14733 [wandb_init.py:init():1044] starting run threads in backend
+2026-04-17 08:57:58,538 INFO    MainThread:14733 [wandb_run.py:_console_start():2529] atexit reg
+2026-04-17 08:57:58,538 INFO    MainThread:14733 [wandb_run.py:_redirect():2377] redirect: wrap_raw
+2026-04-17 08:57:58,538 INFO    MainThread:14733 [wandb_run.py:_redirect():2446] Wrapping output streams.
+2026-04-17 08:57:58,538 INFO    MainThread:14733 [wandb_run.py:_redirect():2469] Redirects installed.
+2026-04-17 08:57:58,541 INFO    MainThread:14733 [wandb_init.py:init():1084] run started, returning control to user process
+2026-04-17 15:45:06,141 INFO    MainThread:14733 [wandb_run.py:_finish():2295] finishing run nikita/code-completion-full-docstring/sa79g3yl
+2026-04-17 15:45:06,141 INFO    MainThread:14733 [wandb_run.py:_atexit_cleanup():2494] got exitcode: 0
+2026-04-17 15:45:06,141 INFO    MainThread:14733 [wandb_run.py:_restore():2476] restore
+2026-04-17 15:45:06,141 INFO    MainThread:14733 [wandb_run.py:_restore():2482] restore done
+2026-04-17 15:45:07,504 INFO    MainThread:14733 [wandb_run.py:_footer_sync_info():3870] logging synced files

train_hnet_with_docstring_18_04/wandb/run-20260417_085757-sa79g3yl/run-sa79g3yl.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83ddbba5f49bfcc57498f835024667a996817dcfda0a3b48ee53acb818ba76b5
+size 7745921