aklein4 commited on
Commit
22e95c0
·
verified ·
1 Parent(s): d852ef2

Upload checkpoint at step 2300

Browse files
.gitattributes CHANGED
@@ -301,3 +301,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
301
  000000002250/state/ocdbt.process_0/d/30f98da6483abaadd2e80d7365e91cca filter=lfs diff=lfs merge=lfs -text
302
  000000002250/state/ocdbt.process_0/d/692c6b1e09bd6452886f892c7d46c9d7 filter=lfs diff=lfs merge=lfs -text
303
  000000002250/state/ocdbt.process_0/d/f49179960ced98a8d728451e50e0bc93 filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
301
  000000002250/state/ocdbt.process_0/d/30f98da6483abaadd2e80d7365e91cca filter=lfs diff=lfs merge=lfs -text
302
  000000002250/state/ocdbt.process_0/d/692c6b1e09bd6452886f892c7d46c9d7 filter=lfs diff=lfs merge=lfs -text
303
  000000002250/state/ocdbt.process_0/d/f49179960ced98a8d728451e50e0bc93 filter=lfs diff=lfs merge=lfs -text
304
+ 000000002300/state/ocdbt.process_0/d/138c56b579fc151b0e7684a45769b0f7 filter=lfs diff=lfs merge=lfs -text
305
+ 000000002300/state/ocdbt.process_0/d/2e3e360869da677c7afe70dbb2ef932b filter=lfs diff=lfs merge=lfs -text
306
+ 000000002300/state/ocdbt.process_0/d/3bbbe853fbd8f80409336dc574aac55b filter=lfs diff=lfs merge=lfs -text
307
+ 000000002300/state/ocdbt.process_0/d/55ad30d69da1045eb52acc3d6d24ced2 filter=lfs diff=lfs merge=lfs -text
308
+ 000000002300/state/ocdbt.process_0/d/6fa2b02bf6d9aad684b62bc2a1e48b86 filter=lfs diff=lfs merge=lfs -text
000000002300/metadata.json ADDED
@@ -0,0 +1,136 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "git_commit": "b1fea531a1bb067c5975282a447929a416793831",
3
+ "wandb_run_id": "ef3f1rcw",
4
+ "config": {
5
+ "seed": 42,
6
+ "notes": null,
7
+ "trainer": {
8
+ "run_dir": "local_data/runs/ablate-scales",
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
+ "fast": {
42
+ "_target_": "optax.contrib.muon",
43
+ "lr": 0.0005,
44
+ "beta": 0.9,
45
+ "adam_b1": 0.9,
46
+ "adam_b2": 0.95,
47
+ "eps": 1e-06,
48
+ "weight_decay": 0.01,
49
+ "ns_steps": 5,
50
+ "consistent_rms": 0.2,
51
+ "nesterov": true,
52
+ "warmup_steps": 100,
53
+ "stable_steps": 899900,
54
+ "decay_steps": 100000,
55
+ "min_lr_ratio": 0.0
56
+ }
57
+ },
58
+ "_target_": "trainers.forte_trainer.ForteTrainer"
59
+ },
60
+ "data": {
61
+ "skip_batches": 0,
62
+ "dataset": {
63
+ "path": "aklein4/300k-horizons",
64
+ "split": "train",
65
+ "streaming": true
66
+ },
67
+ "collator": {
68
+ "_target_": "collators.horizon.HorizonCollator",
69
+ "tokenizer_url": "meta-llama/Llama-3.2-1B-Instruct",
70
+ "max_length": 1024,
71
+ "cluster_length": 64
72
+ }
73
+ },
74
+ "distributed": {
75
+ "initialize": "auto"
76
+ },
77
+ "mesh": {
78
+ "fsdp_axis_size": -1
79
+ },
80
+ "name": "ablate-scales",
81
+ "initialization": "aklein4/forte-init-instruct",
82
+ "initialization_step": 0,
83
+ "model": {
84
+ "sharding": {
85
+ "default": [],
86
+ "rules": [
87
+ {
88
+ "pattern": "^embed_tokens/embedding$",
89
+ "spec": [
90
+ "fsdp",
91
+ null
92
+ ]
93
+ },
94
+ {
95
+ "pattern": "layers/mlp/(gate_proj|up_proj|down_proj)",
96
+ "spec": [
97
+ null,
98
+ null,
99
+ "fsdp"
100
+ ]
101
+ }
102
+ ]
103
+ },
104
+ "_target_": "models.forte.ForteModel",
105
+ "vocab_size": 128256,
106
+ "pad_token_id": -100,
107
+ "hidden_size": 2048,
108
+ "intermediate_size": 8192,
109
+ "num_hidden_layers": 16,
110
+ "num_attention_heads": 32,
111
+ "num_key_value_heads": 8,
112
+ "head_dim": 64,
113
+ "max_position_embeddings": 131072,
114
+ "rope": {
115
+ "theta": 500000.0,
116
+ "factor": 32.0,
117
+ "high_freq_factor": 4.0,
118
+ "low_freq_factor": 1.0,
119
+ "original_context_len": 8192
120
+ },
121
+ "rms_norm_eps": 1e-05,
122
+ "compute_dtype": "bfloat16",
123
+ "gradient_checkpointing": true,
124
+ "attention_backend": "auto",
125
+ "num_fast_vectors": 512,
126
+ "num_fast_heads": 16,
127
+ "base_lr": 0.0001,
128
+ "pool_size": 512,
129
+ "grad_rms_eps": 1e-12
130
+ }
131
+ },
132
+ "step": 2300,
133
+ "checkpoint_items": [
134
+ "params"
135
+ ]
136
+ }
000000002300/state/_CHECKPOINT_METADATA ADDED
@@ -0,0 +1 @@
 
 
1
+ {"item_handlers": "orbax.checkpoint._src.handlers.standard_checkpoint_handler.StandardCheckpointHandler", "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1791462472868297743, "commit_timestamp_nsecs": 1791462521953901285, "custom_metadata": {}}
000000002300/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', 'fast_cartridge', 'forward_gate_proj', 'bias')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "forward_gate_proj", "key_type": 2}, {"key": "bias", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'forward_gate_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "forward_gate_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'k_gate')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "k_gate", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'k_up')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "k_up", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'log_lr')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "log_lr", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'o_proj')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "o_proj", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'q_gate_proj', 'bias')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "q_gate_proj", "key_type": 2}, {"key": "bias", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'q_gate_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "q_gate_proj", "key_type": 2}, {"key": "kernel", "key_type": 2}], "value_metadata": {"value_type": "np.ndarray", "skip_deserialize": false}}, "('params', 'layers', 'mlp', 'fast_cartridge', 'q_up_proj', 'kernel')": {"key_metadata": [{"key": "params", "key_type": 2}, {"key": "layers", "key_type": 2}, {"key": "mlp", "key_type": 2}, {"key": "fast_cartridge", "key_type": 2}, {"key": "q_up_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}
000000002300/state/commit_success.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ Checkpoint commit was successful to /home/aklein4/outer-loop-jax/local_data/runs/ablate-scales/checkpoints/000000002300/state
000000002300/state/d/162ceb3002af54f9d0f8a630ad7ae63e ADDED
Binary file (2.08 kB). View file
 
000000002300/state/manifest.ocdbt ADDED
Binary file (117 Bytes). View file
 
000000002300/state/ocdbt.process_0/d/138c56b579fc151b0e7684a45769b0f7 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:639180ea90458692ba2840e1e6da55da4f5490eeb1ffbd3852ff7c1b11e082b7
3
+ size 61542400
000000002300/state/ocdbt.process_0/d/2e3e360869da677c7afe70dbb2ef932b ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee2dede0afcf76c6e8b17e8b0cef6f1ec6c51da8fe04a24cf2c6b124944d0777
3
+ size 2982510592
000000002300/state/ocdbt.process_0/d/3bbbe853fbd8f80409336dc574aac55b ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4962158d14430f58614f7e7892b78e17e15836eb7e59b042ba26b127c4b4f069
3
+ size 2299068
000000002300/state/ocdbt.process_0/d/55ad30d69da1045eb52acc3d6d24ced2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:262835d1db505699d5c690c07eb3a2cf264e99f7ecd4bf9f47fd4af8abd7d846
3
+ size 998309888
000000002300/state/ocdbt.process_0/d/6fa2b02bf6d9aad684b62bc2a1e48b86 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fc0861d9f4780812567ffc744fbc8e2527478732959473d3f64e225ab45de67
3
+ size 1493037056
000000002300/state/ocdbt.process_0/d/96c169d4ac88552e8553bfa31ce52960 ADDED
Binary file (1.76 kB). View file
 
000000002300/state/ocdbt.process_0/d/9eb7100be5469ffde99df139017e93ff ADDED
Binary file (577 Bytes). View file
 
000000002300/state/ocdbt.process_0/d/d558dbf7b7886a87e596b60fe0de5e2c ADDED
Binary file (201 Bytes). View file
 
000000002300/state/ocdbt.process_0/manifest.ocdbt ADDED
Binary file (404 Bytes). View file