Upload checkpoint at step 1050
Browse files- .gitattributes +5 -0
- 000000001050/metadata.json +115 -0
- 000000001050/state/_CHECKPOINT_METADATA +1 -0
- 000000001050/state/_METADATA +1 -0
- 000000001050/state/commit_success.txt +1 -0
- 000000001050/state/d/dfd00ce6f50f025300ce57dfae542b88 +0 -0
- 000000001050/state/manifest.ocdbt +0 -0
- 000000001050/state/ocdbt.process_0/d/1373496a41e54716b7179af7b027501b +3 -0
- 000000001050/state/ocdbt.process_0/d/22065cdfdcda5ed8c75274ab03011072 +3 -0
- 000000001050/state/ocdbt.process_0/d/28d66e5150511b566b81f4f07a692b2f +3 -0
- 000000001050/state/ocdbt.process_0/d/6ab1b6ad10b2ec718403979e79641f0a +0 -0
- 000000001050/state/ocdbt.process_0/d/9707a089f7485d6502d611ef4239175d +0 -0
- 000000001050/state/ocdbt.process_0/d/af66bdb6aa9b1b7d3055206b0ab0e373 +3 -0
- 000000001050/state/ocdbt.process_0/d/b06aa64a642580ebc830f19f7f5c1175 +3 -0
- 000000001050/state/ocdbt.process_0/d/f19b5fb619a480f0c2c8b86d3894ce79 +0 -0
- 000000001050/state/ocdbt.process_0/manifest.ocdbt +0 -0
.gitattributes
CHANGED
|
@@ -138,3 +138,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 138 |
000000001000/state/ocdbt.process_0/d/5ee3e25a5b190f31988669a0667e379e filter=lfs diff=lfs merge=lfs -text
|
| 139 |
000000001000/state/ocdbt.process_0/d/ac912030c237ffde70ef859a773b7bf4 filter=lfs diff=lfs merge=lfs -text
|
| 140 |
000000001000/state/ocdbt.process_0/d/f3e40944e413e471752465f8e1e812c3 filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138 |
000000001000/state/ocdbt.process_0/d/5ee3e25a5b190f31988669a0667e379e filter=lfs diff=lfs merge=lfs -text
|
| 139 |
000000001000/state/ocdbt.process_0/d/ac912030c237ffde70ef859a773b7bf4 filter=lfs diff=lfs merge=lfs -text
|
| 140 |
000000001000/state/ocdbt.process_0/d/f3e40944e413e471752465f8e1e812c3 filter=lfs diff=lfs merge=lfs -text
|
| 141 |
+
000000001050/state/ocdbt.process_0/d/1373496a41e54716b7179af7b027501b filter=lfs diff=lfs merge=lfs -text
|
| 142 |
+
000000001050/state/ocdbt.process_0/d/22065cdfdcda5ed8c75274ab03011072 filter=lfs diff=lfs merge=lfs -text
|
| 143 |
+
000000001050/state/ocdbt.process_0/d/28d66e5150511b566b81f4f07a692b2f filter=lfs diff=lfs merge=lfs -text
|
| 144 |
+
000000001050/state/ocdbt.process_0/d/af66bdb6aa9b1b7d3055206b0ab0e373 filter=lfs diff=lfs merge=lfs -text
|
| 145 |
+
000000001050/state/ocdbt.process_0/d/b06aa64a642580ebc830f19f7f5c1175 filter=lfs diff=lfs merge=lfs -text
|
000000001050/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": 1050,
|
| 112 |
+
"checkpoint_items": [
|
| 113 |
+
"params"
|
| 114 |
+
]
|
| 115 |
+
}
|
000000001050/state/_CHECKPOINT_METADATA
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"item_handlers": "orbax.checkpoint._src.handlers.standard_checkpoint_handler.StandardCheckpointHandler", "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1789676365773376594, "commit_timestamp_nsecs": 1789676404238322097, "custom_metadata": {}}
|
000000001050/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}
|
000000001050/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/000000001050/state
|
000000001050/state/d/dfd00ce6f50f025300ce57dfae542b88
ADDED
|
Binary file (736 Bytes). View file
|
|
|
000000001050/state/manifest.ocdbt
ADDED
|
Binary file (117 Bytes). View file
|
|
|
000000001050/state/ocdbt.process_0/d/1373496a41e54716b7179af7b027501b
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0bb2d3257548f42971885d9080ae5e3a4f32b9547739339a3a9fa2517572137a
|
| 3 |
+
size 560349184
|
000000001050/state/ocdbt.process_0/d/22065cdfdcda5ed8c75274ab03011072
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c71da552220f0020025254dea989baf775b953b3b3fa3e18527172a0c2bfbd99
|
| 3 |
+
size 998309888
|
000000001050/state/ocdbt.process_0/d/28d66e5150511b566b81f4f07a692b2f
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:78f93f97071d2e0dd9faa1009658292cc93a4de3c07e1c1fe7976d515034751e
|
| 3 |
+
size 62226432
|
000000001050/state/ocdbt.process_0/d/6ab1b6ad10b2ec718403979e79641f0a
ADDED
|
Binary file (201 Bytes). View file
|
|
|
000000001050/state/ocdbt.process_0/d/9707a089f7485d6502d611ef4239175d
ADDED
|
Binary file (430 Bytes). View file
|
|
|
000000001050/state/ocdbt.process_0/d/af66bdb6aa9b1b7d3055206b0ab0e373
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d6912fce990ab1cf5c845b5ba77ef07a2f5bf4cedf92988ab0c4706b9a899dfb
|
| 3 |
+
size 2982506496
|
000000001050/state/ocdbt.process_0/d/b06aa64a642580ebc830f19f7f5c1175
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1302a7b845cea2fc956c71fa2fc399e79e80618a52b7331be7d0e0c6de1f8466
|
| 3 |
+
size 227477
|
000000001050/state/ocdbt.process_0/d/f19b5fb619a480f0c2c8b86d3894ce79
ADDED
|
Binary file (7.11 kB). View file
|
|
|
000000001050/state/ocdbt.process_0/manifest.ocdbt
ADDED
|
Binary file (400 Bytes). View file
|
|
|