Upload checkpoint at step 100
Browse files- .gitattributes +6 -0
- 000000000100/metadata.json +115 -0
- 000000000100/state/_CHECKPOINT_METADATA +1 -0
- 000000000100/state/_METADATA +1 -0
- 000000000100/state/commit_success.txt +1 -0
- 000000000100/state/d/2943697fbd0dc541d63c2c0f68568cfb +0 -0
- 000000000100/state/manifest.ocdbt +0 -0
- 000000000100/state/ocdbt.process_0/d/12750572a8fdbec61e241db449effc31 +3 -0
- 000000000100/state/ocdbt.process_0/d/4fb74e4ee73e1b43e6d7020da1e5e08f +0 -0
- 000000000100/state/ocdbt.process_0/d/596e53ef489a6541847f7586a4965542 +3 -0
- 000000000100/state/ocdbt.process_0/d/9d01fef9c4afa5e195f7f7e6822089c4 +3 -0
- 000000000100/state/ocdbt.process_0/d/a8a67f4316592c204a7627c6b3e17933 +0 -0
- 000000000100/state/ocdbt.process_0/d/b9630c6aa6008fc03acf39dd1727fc36 +3 -0
- 000000000100/state/ocdbt.process_0/d/cba2e2d19a2c57f2c19d9ceb161b240d +3 -0
- 000000000100/state/ocdbt.process_0/d/f68d28b69b79396539c27cd8472a8b15 +3 -0
- 000000000100/state/ocdbt.process_0/manifest.ocdbt +0 -0
.gitattributes
CHANGED
|
@@ -38,3 +38,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 38 |
000000000050/state/ocdbt.process_0/d/b57612afc4ecdc1350c1a99c63512635 filter=lfs diff=lfs merge=lfs -text
|
| 39 |
000000000050/state/ocdbt.process_0/d/c1c7382ebe8ae68b7361aa0afaaaf183 filter=lfs diff=lfs merge=lfs -text
|
| 40 |
000000000050/state/ocdbt.process_0/d/f9a78892bcd74c77bab1acfb0896ae13 filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
000000000050/state/ocdbt.process_0/d/b57612afc4ecdc1350c1a99c63512635 filter=lfs diff=lfs merge=lfs -text
|
| 39 |
000000000050/state/ocdbt.process_0/d/c1c7382ebe8ae68b7361aa0afaaaf183 filter=lfs diff=lfs merge=lfs -text
|
| 40 |
000000000050/state/ocdbt.process_0/d/f9a78892bcd74c77bab1acfb0896ae13 filter=lfs diff=lfs merge=lfs -text
|
| 41 |
+
000000000100/state/ocdbt.process_0/d/12750572a8fdbec61e241db449effc31 filter=lfs diff=lfs merge=lfs -text
|
| 42 |
+
000000000100/state/ocdbt.process_0/d/596e53ef489a6541847f7586a4965542 filter=lfs diff=lfs merge=lfs -text
|
| 43 |
+
000000000100/state/ocdbt.process_0/d/9d01fef9c4afa5e195f7f7e6822089c4 filter=lfs diff=lfs merge=lfs -text
|
| 44 |
+
000000000100/state/ocdbt.process_0/d/b9630c6aa6008fc03acf39dd1727fc36 filter=lfs diff=lfs merge=lfs -text
|
| 45 |
+
000000000100/state/ocdbt.process_0/d/cba2e2d19a2c57f2c19d9ceb161b240d filter=lfs diff=lfs merge=lfs -text
|
| 46 |
+
000000000100/state/ocdbt.process_0/d/f68d28b69b79396539c27cd8472a8b15 filter=lfs diff=lfs merge=lfs -text
|
000000000100/metadata.json
ADDED
|
@@ -0,0 +1,115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"git_commit": "a34fdf8bdceb4b644a9106819b9d6f7d6d0020e0",
|
| 3 |
+
"wandb_run_id": "a01e0icg",
|
| 4 |
+
"config": {
|
| 5 |
+
"seed": 42,
|
| 6 |
+
"notes": null,
|
| 7 |
+
"trainer": {
|
| 8 |
+
"run_dir": "local_data/runs/baseline",
|
| 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",
|
| 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": 100,
|
| 112 |
+
"checkpoint_items": [
|
| 113 |
+
"params"
|
| 114 |
+
]
|
| 115 |
+
}
|
000000000100/state/_CHECKPOINT_METADATA
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"item_handlers": "orbax.checkpoint._src.handlers.standard_checkpoint_handler.StandardCheckpointHandler", "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1789626603722906698, "commit_timestamp_nsecs": 1789626646017503766, "custom_metadata": {}}
|
000000000100/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}
|
000000000100/state/commit_success.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
Checkpoint commit was successful to /home/aklein4/outer-loop-jax/local_data/runs/baseline/checkpoints/000000000100/state
|
000000000100/state/d/2943697fbd0dc541d63c2c0f68568cfb
ADDED
|
Binary file (730 Bytes). View file
|
|
|
000000000100/state/manifest.ocdbt
ADDED
|
Binary file (117 Bytes). View file
|
|
|
000000000100/state/ocdbt.process_0/d/12750572a8fdbec61e241db449effc31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc9b678949780323ce8f613a47d8fe136385e24c365b0da71e3943c18c6439b9
|
| 3 |
+
size 998309888
|
000000000100/state/ocdbt.process_0/d/4fb74e4ee73e1b43e6d7020da1e5e08f
ADDED
|
Binary file (430 Bytes). View file
|
|
|
000000000100/state/ocdbt.process_0/d/596e53ef489a6541847f7586a4965542
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ba9774cb3fdfbbb7a0b072d1dc3c55ada940f75a30e65110888758ba13e7158d
|
| 3 |
+
size 117568
|
000000000100/state/ocdbt.process_0/d/9d01fef9c4afa5e195f7f7e6822089c4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a161b653c445b4f8df558129612da5920cbeb5267c24e1b25a0e2e35e5036bda
|
| 3 |
+
size 560349184
|
000000000100/state/ocdbt.process_0/d/a8a67f4316592c204a7627c6b3e17933
ADDED
|
Binary file (201 Bytes). View file
|
|
|
000000000100/state/ocdbt.process_0/d/b9630c6aa6008fc03acf39dd1727fc36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd2446ca6054e90f1ecd4c0e0990604ac03675499ae7b32a426094ca535ad851
|
| 3 |
+
size 62226432
|
000000000100/state/ocdbt.process_0/d/cba2e2d19a2c57f2c19d9ceb161b240d
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:17e183eec8ff83a688f8b4e4b70d604902a67e7ce554a0e35380b50a06bb498e
|
| 3 |
+
size 111337
|
000000000100/state/ocdbt.process_0/d/f68d28b69b79396539c27cd8472a8b15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4cea9e485692e213d85299e90deee2c2f925aa0dcf2d9e3cd35f13340386cb2a
|
| 3 |
+
size 2982490112
|
000000000100/state/ocdbt.process_0/manifest.ocdbt
ADDED
|
Binary file (400 Bytes). View file
|
|
|