aklein4 commited on
Commit
2eb2b06
·
verified ·
1 Parent(s): 9e39d14

Upload checkpoint at step 1800

Browse files
.gitattributes CHANGED
@@ -218,3 +218,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
218
  000000001750/state/ocdbt.process_0/d/2522278f5115701056a79b44db16218d filter=lfs diff=lfs merge=lfs -text
219
  000000001750/state/ocdbt.process_0/d/b82fe11305d0ebedd50cab5d2132916e filter=lfs diff=lfs merge=lfs -text
220
  000000001750/state/ocdbt.process_0/d/ccfae79b93c617d74de8682caec45446 filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
218
  000000001750/state/ocdbt.process_0/d/2522278f5115701056a79b44db16218d filter=lfs diff=lfs merge=lfs -text
219
  000000001750/state/ocdbt.process_0/d/b82fe11305d0ebedd50cab5d2132916e filter=lfs diff=lfs merge=lfs -text
220
  000000001750/state/ocdbt.process_0/d/ccfae79b93c617d74de8682caec45446 filter=lfs diff=lfs merge=lfs -text
221
+ 000000001800/state/ocdbt.process_0/d/0407656078a034dc088c9c7d951b7b96 filter=lfs diff=lfs merge=lfs -text
222
+ 000000001800/state/ocdbt.process_0/d/1a5f5b02644a6285308b2bf9c9106e75 filter=lfs diff=lfs merge=lfs -text
223
+ 000000001800/state/ocdbt.process_0/d/753391fa175aa418c3bf741f2596a397 filter=lfs diff=lfs merge=lfs -text
224
+ 000000001800/state/ocdbt.process_0/d/9f34a5f225464d1d978284db7379c2ea filter=lfs diff=lfs merge=lfs -text
225
+ 000000001800/state/ocdbt.process_0/d/a68ccd88c8fc4fe88c0d7fbffc28ec60 filter=lfs diff=lfs merge=lfs -text
000000001800/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": 1800,
112
+ "checkpoint_items": [
113
+ "params"
114
+ ]
115
+ }
000000001800/state/_CHECKPOINT_METADATA ADDED
@@ -0,0 +1 @@
 
 
1
+ {"item_handlers": "orbax.checkpoint._src.handlers.standard_checkpoint_handler.StandardCheckpointHandler", "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1789710182549814878, "commit_timestamp_nsecs": 1789710221525631745, "custom_metadata": {}}
000000001800/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}
000000001800/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/000000001800/state
000000001800/state/d/ef3751d2ec1786a80862681b0283dae6 ADDED
Binary file (713 Bytes). View file
 
000000001800/state/manifest.ocdbt ADDED
Binary file (117 Bytes). View file
 
000000001800/state/ocdbt.process_0/d/0407656078a034dc088c9c7d951b7b96 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e2c23ef0d20ca0fe61ec11bdbca6f2cb85408cf52dde1e7a8168de4a10f55c0
3
+ size 560234496
000000001800/state/ocdbt.process_0/d/1a5f5b02644a6285308b2bf9c9106e75 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32d75b675af3a9c991611fbf55be1e1bf34eeaeaf8e206828da591742603e77f
3
+ size 234237
000000001800/state/ocdbt.process_0/d/34ae433a5f28589a9d73b63bf84f0760 ADDED
Binary file (504 Bytes). View file
 
000000001800/state/ocdbt.process_0/d/753391fa175aa418c3bf741f2596a397 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c03f323255ca87fb8ff812fa75aae4f3dfd9225901f8a6ee66755b0471afd1f8
3
+ size 62337024
000000001800/state/ocdbt.process_0/d/9f34a5f225464d1d978284db7379c2ea ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94dc3caf3039d6afed5cbb3cb33c09d053f5620053d8b4e621c85e8a8da395d6
3
+ size 2982514688
000000001800/state/ocdbt.process_0/d/a18668adb7fda3c2c9cb079d2a5be860 ADDED
Binary file (430 Bytes). View file
 
000000001800/state/ocdbt.process_0/d/a68ccd88c8fc4fe88c0d7fbffc28ec60 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:132c8024a5d7e372bcf42af5c58f22b99f1e0aacf1c74a575bd7f942ecbcd02c
3
+ size 998309888
000000001800/state/ocdbt.process_0/d/cf1a33e04b614c79a9af9e45e8ee91a7 ADDED
Binary file (201 Bytes). View file
 
000000001800/state/ocdbt.process_0/manifest.ocdbt ADDED
Binary file (401 Bytes). View file