Upload folder using huggingface_hub

model_config.yaml

@@ -0,0 +1,315 @@
+mcore_gpt: true
+micro_batch_size: 1
+global_batch_size: 128
+tensor_model_parallel_size: 8
+pipeline_model_parallel_size: 8
+virtual_pipeline_model_parallel_size: null
+encoder_seq_length: 8192
+max_position_embeddings: 8192
+num_layers: 80
+hidden_size: 8192
+ffn_hidden_size: 28672
+num_attention_heads: 64
+init_method_std: 0.02
+use_scaled_init_method: true
+hidden_dropout: 0.0
+attention_dropout: 0.0
+ffn_dropout: 0.0
+kv_channels: null
+apply_query_key_layer_scaling: true
+normalization: rmsnorm
+layernorm_epsilon: 1.0e-05
+do_layer_norm_weight_decay: false
+make_vocab_size_divisible_by: 128
+pre_process: true
+post_process: true
+persist_layer_norm: true
+bias: false
+activation: fast-swiglu
+headscale: false
+transformer_block_type: pre_ln
+openai_gelu: false
+normalize_attention_scores: true
+position_embedding_type: rope
+rotary_percentage: 1.0
+attention_type: multihead
+share_embeddings_and_output_weights: false
+overlap_p2p_comm: false
+batch_p2p_comm: true
+num_query_groups: 8
+tokenizer:
+  library: huggingface
+  type: meta-llama/Meta-Llama-3-70B
+  model: /dataset/models/llama2-13b/llama-tokenizer.model
+  vocab_file: null
+  merge_file: null
+  tokenizer_model: /dataset/models/llama2-13b/llama-tokenizer.model
+  sentencepiece_legacy: false
+  use_fast: true
+native_amp_init_scale: 4294967296
+native_amp_growth_interval: 1000
+hysteresis: 2
+fp32_residual_connection: false
+fp16_lm_cross_entropy: false
+megatron_amp_O2: true
+grad_allreduce_chunk_size_mb: 125
+grad_div_ar_fusion: true
+gradient_accumulation_fusion: false
+bias_activation_fusion: false
+bias_dropout_add_fusion: false
+masked_softmax_fusion: true
+get_attention_mask_from_fusion: true
+apply_rope_fusion: false
+seed: 1234
+resume_from_checkpoint: null
+use_cpu_initialization: false
+onnx_safe: false
+apex_transformer_log_level: 30
+gradient_as_bucket_view: false
+sync_batch_comm: false
+activations_checkpoint_granularity: full
+activations_checkpoint_method: uniform
+activations_checkpoint_num_layers: 1
+num_micro_batches_with_partial_activation_checkpoints: null
+activations_checkpoint_layers_per_pipeline: null
+sequence_parallel: false
+transformer_engine: true
+fp8: false
+fp8_e4m3: false
+fp8_hybrid: true
+fp8_margin: 0
+fp8_interval: 1
+fp8_amax_history_len: 1024
+fp8_amax_compute_algo: max
+reduce_amax: true
+use_emha: false
+data:
+  chat: true
+  chat_prompt_tokens:
+    system_turn_start: <extra_id_0>
+    turn_start: <extra_id_1>
+    label_start: <extra_id_2>
+    end_of_turn: '
+
+      '
+    end_of_name: '
+
+      '
+  sample: true
+  num_workers: 2
+  dataloader_type: single
+  train_ds:
+    file_path: /dataset/train.jsonl
+    global_batch_size: 384
+    micro_batch_size: 1
+    shuffle: true
+    memmap_workers: null
+    max_seq_length: 4096
+    min_seq_length: 1
+    drop_last: true
+    concat_sampling_probabilities: null
+    label_key: output
+    add_eos: false
+    add_sep: false
+    add_bos: false
+    truncation_field: input
+    index_mapping_dir: /indexmap_dir
+    prompt_template: '<extra_id_0>System
+
+      {system message}
+
+      <extra_id_1>User
+
+      {turn 1 user message}
+
+      <extra_id_1>Assistant
+
+      <extra_id_2>{turn 1 assistant label}
+
+      {turn 1 assistant message}
+
+      <extra_id_1>User
+
+      {turn 2 user message}
+
+      <extra_id_1>Assistant
+
+      <extra_id_2>{turn 2 assistant label}
+
+      {turn 2 assistant message}
+
+      <extra_id_1>'
+    hf_dataset: true
+    truncation_method: right
+  validation_ds:
+    file_path: /dataset/train.jsonl
+    names: null
+    global_batch_size: 384
+    micro_batch_size: 1
+    shuffle: false
+    memmap_workers: null
+    max_seq_length: 4096
+    min_seq_length: 1
+    drop_last: false
+    label_key: output
+    add_eos: false
+    add_sep: false
+    add_bos: false
+    write_predictions_to_file: false
+    output_file_path_prefix: null
+    truncation_field: input
+    index_mapping_dir: /indexmap_dir
+    prompt_template: '<extra_id_0>System
+
+      {system message}
+
+      <extra_id_1>User
+
+      {turn 1 user message}
+
+      <extra_id_1>Assistant
+
+      <extra_id_2>{turn 1 assistant label}
+
+      {turn 1 assistant message}
+
+      <extra_id_1>User
+
+      {turn 2 user message}
+
+      <extra_id_1>Assistant
+
+      <extra_id_2>{turn 2 assistant label}
+
+      {turn 2 assistant message}
+
+      <extra_id_1>'
+    tokens_to_generate: 32
+    hf_dataset: true
+    truncation_method: right
+    metric:
+      name: loss
+      average: null
+      num_classes: null
+  test_ds:
+    prompt_template: '<extra_id_0>System
+
+      {system message}
+
+      <extra_id_1>User
+
+      {turn 1 user message}
+
+      <extra_id_1>Assistant
+
+      <extra_id_2>{turn 1 assistant label}
+
+      {turn 1 assistant message}
+
+      <extra_id_1>User
+
+      {turn 2 user message}
+
+      <extra_id_1>Assistant
+
+      <extra_id_2>{turn 2 assistant label}
+
+      {turn 2 assistant message}
+
+      <extra_id_1>'
+  data_impl: jsonl
+  splits_string: null
+  seq_length: 8192
+  skip_warmup: true
+  reset_position_ids: false
+  reset_attention_mask: false
+  eod_mask_loss: false
+  index_mapping_dir: null
+  data_prefix:
+    train:
+    - /lustre/fsw/coreai_dlalgo_llm/geshen/trt_llm/helpsteer_data/helpsteer_11_train_prompts.jsonl
+    validation:
+    - /lustre/fsw/coreai_dlalgo_llm/geshen/trt_llm/helpsteer_data/helpsteer_11_val_prompts.jsonl
+    test:
+    - /lustre/fsw/coreai_dlalgo_llm/geshen/trt_llm/helpsteer_data/helpsteer_11_val_prompts.jsonl
+nsys_profile:
+  enabled: false
+  start_step: 10
+  end_step: 10
+  ranks:
+  - 0
+  gen_shape: false
+optim:
+  name: distributed_fused_adam
+  lr: 2.0e-07
+  weight_decay: 0.1
+  betas:
+  - 0.9
+  - 0.98
+  sched:
+    name: CosineAnnealing
+    warmup_steps: 10
+    constant_steps: 1000
+    min_lr: 1.9999e-07
+    max_steps: 88
+  bucket_cap_mb: 200
+  overlap_grad_sync: false
+  contiguous_grad_buffer: true
+rotary_base: 500000.0
+precision: bf16
+answer_only_loss: true
+restore_from_path: /models/unpacked_llama3_70b_base
+save_nemo_on_validation_end: true
+use_flash_attention: null
+pipeline_model_parallel_split_rank: 0
+ppo:
+  trt_llm:
+    enable: true
+    reshard: true
+    max_input_len: 4096
+    max_input_tokens: 32768
+    model_type: LLaMAForCausalLM
+    unload_engine_train: false
+  rollout_micro_batch_size: 8
+  num_rollout_samples: 128
+  forward_micro_batch_size: 4
+  val_rollout_micro_batch_size: 8
+  num_val_samples: 128
+  offload_adam_states: true
+  entropy_bonus: 0.0
+  ratio_eps: 0.2
+  sampling_params:
+    use_greedy: false
+    temperature: 1.0
+    top_k: 0
+    top_p: 1.0
+    repetition_penalty: 1.0
+    add_BOS: false
+    all_probs: false
+    compute_logprob: false
+    end_strings:
+    - <|endoftext|>
+    - <extra_id_1>
+  length_params:
+    max_length: 2048
+    min_length: 1
+peft:
+  peft_scheme: none
+  restore_from_path: null
+  restore_from_ckpt:
+    checkpoint_dir: null
+    checkpoint_name: null
+  lora_tuning:
+    target_modules:
+    - attention_qkv
+    adapter_dim: 32
+    adapter_dropout: 0.0
+    column_init_method: xavier
+    row_init_method: zero
+    layer_selection: null
+    weight_tying: false
+    position_embedding_strategy: null
+offload_adam_states: true
+enable_nge: true
+target: nemo_aligner.models.nlp.gpt.megatron_gpt_ppo_actor.MegatronGPTActorModel
+nemo_version: 2.0.0rc0

model_weights/common.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d508deac1f3085ed140e04c6620939154ae96f07278be1363e143c0a083e3a3f
+size 35715

model_weights/metadata.json

@@ -0,0 +1 @@
+{"sharded_backend": "zarr", "sharded_backend_version": 1, "common_backend": "torch", "common_backend_version": 1}

model_weights/model.decoder.final_layernorm.weight/.zarray

@@ -0,0 +1,14 @@
+{
+    "chunks": [
+        8192
+    ],
+    "compressor": null,
+    "dtype": "bfloat16",
+    "fill_value": null,
+    "filters": null,
+    "order": "C",
+    "shape": [
+        8192
+    ],
+    "zarr_format": 2
+}

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_0_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e9a729975a47eb62643f3ac6a8bf145c719b877b2ce5b967be11c2b18801913
+size 1836

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_10_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:faac9f0f6c1329624b1cad1daa596a7833bf7475949da3e7e4ec86709dca4952
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_11_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:158179c28c9e196e80f0eb646eb219fce5fff77cd50a03ab445e028d2ac8bf76
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_12_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87fe00dd30d5e94f76b7c60e50c4fac2b19594743a26da4d5081a1d56fa54c47
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_13_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9a51a0705943fe66ca2db9ac36923dee52b1dadb1b32efcf1c50c1e62eb54f0
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_14_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d96cc54ddbca63e537764caffb8b3602a65f652827354e19907c5980de821b9
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_15_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41cb45e944859eb6dcedcd8e7e56c7332ae31753f6daa3f61289ca04368db067
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_16_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34e4d66af34f6390b2bc1b83973f1ebe6ad6b32b391e0f1795f426ddcbdda783
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_17_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5732c2a0c0f07c0d4d11616be40ad87ac2a80986400cdc1f7d55a98f1d094d4
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_18_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66833ba47bee314a4c94098e985aa831ca98533ee2ec794b1c5dd95149f9f82c
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_19_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01b45b3f4ef6b69193217eb1bbdd50781c76efcb176194b72f28b2caeb617799
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_1_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15b6735d1668ebc2574d56235fd254e7c2977a571e8b2aa717f53158adaa76b8
+size 1836

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_20_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27682c2cd23ed3f1a7a3a5d40c5c84f0ac24ff07e1528665e3574d52ea707d71
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_21_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:040aa73c2be92a7ed400cac3319ef89eede969f585b0e12c2cddd9c01800f2cc
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_22_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02368a8b8351735ecb4d892cea21dea8c5cbaee51e9ab6e0f42994c093e42f53
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_23_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b738c1a25114960e407fa1d35e3417e1b9f9e4a5d4d7e45c777f4ddffaa24a2b
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_24_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c72f2341f3ef6ebc07f2d7f409c29e85a27730089b09fd2c30e2cc1c9e30b17b
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_25_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5130873b235257ae1630a2f83dfb3bfb073e49d469ad1e802cbe797fb4ea1bb
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_26_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:249b210a061b725c547cb1855b4afd0feaef02593d6e8c371fc8d96c7ae019f4
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_27_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1266642f65d8d695707e2eaa9631ea2472f16f4f97d3e22e0065da6875daf252
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_28_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e5b09e3f9f634bf19a80bb9abb0b45f49ee4b3c7f007e35cd8e7925daabbe32
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_29_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba77cb7bb7ee730d9e91fe35e1084a89c45eb32ff363a94133b8f30c50c229f1
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_2_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:026af90a314db680105fea6a1aae98989ebe2d6f52eaf062006550d077bda05a
+size 1836

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_30_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78b87e344c6ce15fc3190ed5741823cc913defc8da276302cd704de63ddf3f85
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_31_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ae569fddb6d7b9d201e4e935ea33b736dee2487c533fe330b1f4b20edbf0175
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_32_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:433659103399b2f5ea6533ab08d9cd213c58e13501de3f3f049a84b8d7a58ac0
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_33_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9169d2efa0cc9cc2c46021569a1140179409233c658576499376f5db9d1d8c8
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_34_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abebd91eaa58bf1577d69ce1c0bfb8b91f05002d3b26edf434c0b7eb55f7475d
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_35_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3d95be77323d1eb8c0716e39d9ddb8d06cfa3cda9c5189d82afb91e459a9146
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_36_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e38c7ed5c1127ea72da184b561f154c2393e512769ee59e77512c6c85f7517a7
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_37_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc79248144069a2751c43f604917cca98519c9fe514dd53be67d2a955d48c7b1
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_38_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c894260cb35b9c59a8501e743bee84de53188279a4d31791f63a33f4ad56f229
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_39_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed58173cd24c9b37daef9ef2c3ceac4da6f2927c2320d3eefb3c222b6b68d8f6
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_3_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28ae5501b97631ba9aff2524f117edd3eb863dfebb29395610f689c851282d58
+size 1836

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_40_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01530e6ee0f9dd2007409746855c10469ff5e93da59ccc5eb4e678cd478a9c17
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_41_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3477e4945e17e402f8190e1d9937627097783ba125917bb0df1760ed25844bda
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_42_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7f9dcbf8f93adbcf221e0ae8117118a069ce4eadeb0c11cc44cf661eb1c860f
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_43_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74147b0cf9114f4966c081b94393a9ada2b3fc27eed18e0ac34d580ed8f5d2d2
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_44_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bc4cbaf569529f6b36bfc4ab323c0581d32135a021837ee0e932ff41f1c5e42
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_45_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56f96e00f6542f620b6ee51e64bc9aa76e3907c9e117e27825af4ef10f4d3cd7
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_46_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36515ee57e2e8b16b0ee22dc8b9176a29aa212a901eadb82a6d133c2487939b9
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_47_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce91aff8cb5ebdeeb73dd871ed4aa4b19efb94861f55757cfc2d3ebb037dca78
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_48_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc878f893ca0b2bd6a4556a1b8ef3e13dabbf7c2001e556a4939383886e2f857
+size 1840

model_weights/model.decoder.layers.mlp.linear_fc1._extra_state/shard_49_80.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9187866aa84481d462ec5d43082b88c1c92363c6207de7a1a7b43a7744b2fda6
+size 1840