aklein4 commited on
Commit
81cd439
·
verified ·
1 Parent(s): ce874a9

Upload checkpoint at step 400

Browse files
.gitattributes CHANGED
@@ -69,3 +69,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
69
  000000000350/state/ocdbt.process_0/d/acd4bebf1a92c47844135976c34993c8 filter=lfs diff=lfs merge=lfs -text
70
  000000000350/state/ocdbt.process_0/d/b3be547cbc2570d67b5a83ed9b1d87f9 filter=lfs diff=lfs merge=lfs -text
71
  000000000350/state/ocdbt.process_0/d/f768cee1da17b2ab7cab2a813b2b657d filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
69
  000000000350/state/ocdbt.process_0/d/acd4bebf1a92c47844135976c34993c8 filter=lfs diff=lfs merge=lfs -text
70
  000000000350/state/ocdbt.process_0/d/b3be547cbc2570d67b5a83ed9b1d87f9 filter=lfs diff=lfs merge=lfs -text
71
  000000000350/state/ocdbt.process_0/d/f768cee1da17b2ab7cab2a813b2b657d filter=lfs diff=lfs merge=lfs -text
72
+ 000000000400/state/ocdbt.process_0/d/406899825d7cde4cb47e915524763267 filter=lfs diff=lfs merge=lfs -text
73
+ 000000000400/state/ocdbt.process_0/d/6b93745c511738eb0d6cb21229d978f4 filter=lfs diff=lfs merge=lfs -text
74
+ 000000000400/state/ocdbt.process_0/d/873106f34058a157f4e76fa520ea3b43 filter=lfs diff=lfs merge=lfs -text
75
+ 000000000400/state/ocdbt.process_0/d/93f05bf6d599e3a8d8dec5f3e1dcccc4 filter=lfs diff=lfs merge=lfs -text
76
+ 000000000400/state/ocdbt.process_0/d/a905f0ed0c8fc38efc73f389553fc57a filter=lfs diff=lfs merge=lfs -text
000000000400/metadata.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "git_commit": "a34fdf8bdceb4b644a9106819b9d6f7d6d0020e0",
3
+ "wandb_run_id": "fnn7p0ul",
4
+ "config": {
5
+ "seed": 42,
6
+ "notes": null,
7
+ "trainer": {
8
+ "run_dir": "local_data/runs/baseline-fixed",
9
+ "checkpoint": {
10
+ "load_state": false,
11
+ "save_optimizer": false,
12
+ "interval": 50
13
+ },
14
+ "wandb": {
15
+ "project": "oloop-instruct",
16
+ "enabled": true
17
+ },
18
+ "huggingface": {
19
+ "private": false,
20
+ "enabled": true
21
+ },
22
+ "global_batch_size": 128,
23
+ "max_steps": 1000000,
24
+ "num_logit_iterations": 4,
25
+ "aux_loss_weight": 0.1,
26
+ "max_grad_norm": 1.0,
27
+ "debug": false,
28
+ "groups": {
29
+ "slow": {
30
+ "_target_": "optax.adamw",
31
+ "lr": 3e-05,
32
+ "b1": 0.9,
33
+ "b2": 0.95,
34
+ "eps": 1e-06,
35
+ "weight_decay": 0.0,
36
+ "warmup_steps": 100,
37
+ "stable_steps": 899900,
38
+ "decay_steps": 100000,
39
+ "min_lr_ratio": 0.0
40
+ }
41
+ },
42
+ "_target_": "trainers.horizon_lm_trainer.HorizonLMTrainer"
43
+ },
44
+ "data": {
45
+ "skip_batches": 0,
46
+ "dataset": {
47
+ "path": "aklein4/300k-horizons",
48
+ "split": "train",
49
+ "streaming": true
50
+ },
51
+ "collator": {
52
+ "_target_": "collators.horizon.HorizonCollator",
53
+ "tokenizer_url": "meta-llama/Llama-3.2-1B-Instruct",
54
+ "max_length": 1024,
55
+ "cluster_length": 64
56
+ }
57
+ },
58
+ "distributed": {
59
+ "initialize": "auto"
60
+ },
61
+ "mesh": {
62
+ "fsdp_axis_size": -1
63
+ },
64
+ "name": "baseline-fixed",
65
+ "initialization": "meta-llama/Llama-3.2-1B-Instruct",
66
+ "initialization_step": null,
67
+ "model": {
68
+ "sharding": {
69
+ "default": [],
70
+ "rules": [
71
+ {
72
+ "pattern": "^embed_tokens/embedding$",
73
+ "spec": [
74
+ "fsdp",
75
+ null
76
+ ]
77
+ },
78
+ {
79
+ "pattern": "layers/mlp/(gate_proj|up_proj|down_proj)",
80
+ "spec": [
81
+ null,
82
+ null,
83
+ "fsdp"
84
+ ]
85
+ }
86
+ ]
87
+ },
88
+ "_target_": "models.llama.LlamaForCausalLM",
89
+ "vocab_size": 128256,
90
+ "pad_token_id": -100,
91
+ "hidden_size": 2048,
92
+ "intermediate_size": 8192,
93
+ "num_hidden_layers": 16,
94
+ "num_attention_heads": 32,
95
+ "num_key_value_heads": 8,
96
+ "head_dim": 64,
97
+ "max_position_embeddings": 131072,
98
+ "rope": {
99
+ "theta": 500000.0,
100
+ "factor": 32.0,
101
+ "high_freq_factor": 4.0,
102
+ "low_freq_factor": 1.0,
103
+ "original_context_len": 8192
104
+ },
105
+ "rms_norm_eps": 1e-05,
106
+ "compute_dtype": "bfloat16",
107
+ "gradient_checkpointing": true,
108
+ "attention_backend": "auto"
109
+ }
110
+ },
111
+ "step": 400,
112
+ "checkpoint_items": [
113
+ "params"
114
+ ]
115
+ }
000000000400/state/_CHECKPOINT_METADATA ADDED
@@ -0,0 +1 @@
 
 
1
+ {"item_handlers": "orbax.checkpoint._src.handlers.standard_checkpoint_handler.StandardCheckpointHandler", "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1789646860536295059, "commit_timestamp_nsecs": 1789646899388642732, "custom_metadata": {}}
000000000400/state/_METADATA ADDED
@@ -0,0 +1 @@
 
 
1
+ {"tree_metadata": {"('params', 'embed_tokens', 'embedding')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "embed_tokens", "key_type": 2}, {"key": "embedding", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'input_layernorm', 'scale')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "input_layernorm", "key_type": 2}, {"key": "scale", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'down_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "down_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'gate_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "gate_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'up_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "up_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'post_attention_layernorm', 'scale')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "post_attention_layernorm", "key_type": 2}, {"key": "scale", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'self_attn', 'k_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "self_attn", "key_type": 2}, {"key": "k_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'self_attn', 'o_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "self_attn", "key_type": 2}, {"key": "o_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'self_attn', 'q_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "self_attn", "key_type": 2}, {"key": "q_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'self_attn', 'v_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "self_attn", "key_type": 2}, {"key": "v_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'lm_head', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "lm_head", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'norm', 'scale')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "norm", "key_type": 2}, {"key": "scale", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}}, "use_ocdbt": true, "use_zarr3": false, "store_array_data_equal_to_fill_value": true, "custom_metadata": null}
000000000400/state/commit_success.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ Checkpoint commit was successful to /home/aklein4/outer-loop-jax/local_data/runs/baseline-fixed/checkpoints/000000000400/state
000000000400/state/d/1011a6a40c64aa6f3330932ad0e66bd5 ADDED
Binary file (734 Bytes). View file
 
000000000400/state/manifest.ocdbt ADDED
Binary file (117 Bytes). View file
 
000000000400/state/ocdbt.process_0/d/1001a1cf64d7ad20b033c76be079eb19 ADDED
Binary file (7.02 kB). View file
 
000000000400/state/ocdbt.process_0/d/147c4a954484c426eaa193c6a5303c76 ADDED
Binary file (201 Bytes). View file
 
000000000400/state/ocdbt.process_0/d/406899825d7cde4cb47e915524763267 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0121767fad8ec3797e2cbb76dad391a01b63102e1a44bf559439d9bc99995ead
3
+ size 62226432
000000000400/state/ocdbt.process_0/d/6b93745c511738eb0d6cb21229d978f4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06f4f397e15d22a188e743ec06904197221aa7f0d914a29b1299c93524b1c4c2
3
+ size 227037
000000000400/state/ocdbt.process_0/d/7fae062b9475eb572560609d76984ffd ADDED
Binary file (430 Bytes). View file
 
000000000400/state/ocdbt.process_0/d/873106f34058a157f4e76fa520ea3b43 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:469bf496946df19dd396b9afa2369ba18a9e7eccbaf7fd2d2d722380d20dcf07
3
+ size 560353280
000000000400/state/ocdbt.process_0/d/93f05bf6d599e3a8d8dec5f3e1dcccc4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83d9a8812b7abd1a7cea2b62fc297b162e3bfee806c484fd173b89a5dbf83aec
3
+ size 998309888
000000000400/state/ocdbt.process_0/d/a905f0ed0c8fc38efc73f389553fc57a ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2837cc8cd04ce4e1aed4c9f6613f8caebab71c25d735f7e3e8184915c98adc6a
3
+ size 2982498304
000000000400/state/ocdbt.process_0/manifest.ocdbt ADDED
Binary file (399 Bytes). View file