feat: add safe distributed checkpoint intake and preflight

This commit is contained in:
Joseph Magly
2026-09-04 19:43:53 -04:00
parent 5cc43c6e52
commit 985c9e9363
108 changed files with 21726 additions and 92 deletions
@@ -0,0 +1,161 @@
{
"case_id": "mixed-model-peft",
"components": [
"full_model",
"peft_adapter"
],
"expected_manifest_digest": "sha256:a37bcf41c60410c78430f767fcd0ab4c413ceca7fb9ef9da9bbb35f542b0043f",
"features": [
"mixed_full_model_peft"
],
"fragments": [
{
"component_id": "peft-adapter",
"dtype": "float32",
"element_extent": [
1,
2
],
"element_offset": [
0,
0
],
"evidence_refs": [
"synthetic:mixed-adapter"
],
"fqn": "adapter.lora_A.weight",
"fragment_digest": "sha256:deea3b24add66f9c401d38a758eb5cb664db0596a3113b5ceaf8c5e774faa321",
"fragment_id": "mixed-adapter",
"global_shape": [
1,
2
],
"local_shape": [
1,
2
],
"logical_tensor_id": "adapter.lora_A.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00001.safetensors",
"payload_key": "mixed-adapter",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00001.safetensors",
"shard_file_id": "rank-00001.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
1
]
]
},
{
"component_id": "full-model",
"dtype": "float32",
"element_extent": [
2,
2
],
"element_offset": [
0,
0
],
"evidence_refs": [
"synthetic:mixed-model"
],
"fqn": "model.weight",
"fragment_digest": "sha256:4c9c4f354e74153db012329d71c8562ec23e498148174b2c49de58f45d47cdbe",
"fragment_id": "mixed-model",
"global_shape": [
2,
2
],
"local_shape": [
2,
2
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "mixed-model",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
0
]
]
}
],
"oracle_file": "oracles.safetensors",
"oracles": [
{
"dtype": "float32",
"logical_tensor_id": "adapter.lora_A.weight",
"payload_key": "tensor_000",
"sha256": "sha256:deea3b24add66f9c401d38a758eb5cb664db0596a3113b5ceaf8c5e774faa321",
"shape": [
1,
2
]
},
{
"dtype": "float32",
"logical_tensor_id": "model.weight",
"payload_key": "tensor_001",
"sha256": "sha256:4c9c4f354e74153db012329d71c8562ec23e498148174b2c49de58f45d47cdbe",
"shape": [
2,
2
]
}
],
"schema_id": "obliteratus.checkpoint-fixture-case",
"schema_version": "1.0.0",
"topology": {
"source": {
"world_size": 2
},
"target": {
"world_size": 1
}
},
"world_size": 2
}
@@ -0,0 +1,310 @@
{
"case_id": "tp2-pp2-to-single",
"components": [
"full_model"
],
"expected_manifest_digest": "sha256:03da2d1dafe858e225174f12394d243dcaf70516c2649bca85c57dce91dff323",
"features": [
"pipeline_parallel",
"topology_a_to_b"
],
"fragments": [
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
2
],
"element_offset": [
0,
0
],
"evidence_refs": [
"synthetic:tp-pp-r0"
],
"fqn": "model.weight",
"fragment_digest": "sha256:36f52612ab7fad5ef2adca0d34f41471c0c38f50cc92d9a47fa662631d3e1268",
"fragment_id": "tp-pp-r0",
"global_shape": [
4,
4
],
"local_shape": [
2,
2
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00000.safetensors",
"payload_key": "tp-pp-r0",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"pp",
0
],
[
"rank",
0
],
[
"tp",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
2
],
"element_offset": [
0,
2
],
"evidence_refs": [
"synthetic:tp-pp-r1"
],
"fqn": "model.weight",
"fragment_digest": "sha256:762921576e058fd1360006105a48a8225d6baa27085f7d30abed455c288f9308",
"fragment_id": "tp-pp-r1",
"global_shape": [
4,
4
],
"local_shape": [
2,
2
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00001.safetensors",
"payload_key": "tp-pp-r1",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00001.safetensors",
"shard_file_id": "rank-00001.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"pp",
0
],
[
"rank",
1
],
[
"tp",
1
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
2
],
"element_offset": [
2,
0
],
"evidence_refs": [
"synthetic:tp-pp-r2"
],
"fqn": "model.weight",
"fragment_digest": "sha256:5cca39bfe5a82135bad95ae98f2df321fe08ea75b68ea8f939ffa90845138146",
"fragment_id": "tp-pp-r2",
"global_shape": [
4,
4
],
"local_shape": [
2,
2
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00002.safetensors",
"payload_key": "tp-pp-r2",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00002.safetensors",
"shard_file_id": "rank-00002.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"pp",
1
],
[
"rank",
2
],
[
"tp",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
2
],
"element_offset": [
2,
2
],
"evidence_refs": [
"synthetic:tp-pp-r3"
],
"fqn": "model.weight",
"fragment_digest": "sha256:dbb7087f7cd13a205e0dcec5093a66dcfac4f6c237afd8dcafd64d166227083a",
"fragment_id": "tp-pp-r3",
"global_shape": [
4,
4
],
"local_shape": [
2,
2
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00003.safetensors",
"payload_key": "tp-pp-r3",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00003.safetensors",
"shard_file_id": "rank-00003.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"pp",
1
],
[
"rank",
3
],
[
"tp",
1
]
]
}
],
"oracle_file": "oracles.safetensors",
"oracles": [
{
"dtype": "float32",
"logical_tensor_id": "model.weight",
"payload_key": "tensor_000",
"sha256": "sha256:58dda328598e2f7fe472621bfc54935aaa354d1a6ebcaf9562cd743fd575eb19",
"shape": [
4,
4
]
}
],
"schema_id": "obliteratus.checkpoint-fixture-case",
"schema_version": "1.0.0",
"topology": {
"source": {
"pp": 2,
"tp": 2
},
"target": {
"world_size": 1
}
},
"world_size": 4
}
@@ -0,0 +1,393 @@
{
"case_id": "world1-complete",
"components": [
"full_model"
],
"expected_manifest_digest": "sha256:a6f4338c200816ce0a71bc9034e7c4792528f5520273c94718d332e995649065",
"features": [
"buffer",
"expert",
"scalar",
"tied_weight"
],
"fragments": [
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:w1-buffer"
],
"fqn": "model.running_mean",
"fragment_digest": "sha256:ab2a21ab5e1262d555eea5678035e8fe3e76542c6b474cc04a16a39d4aa02645",
"fragment_id": "w1-buffer",
"global_shape": [
2
],
"local_shape": [
2
],
"logical_tensor_id": "model.running_mean",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "w1-buffer",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "persistent_buffer",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
4
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:w1-embed"
],
"fqn": "model.embed.weight",
"fragment_digest": "sha256:ad73b9acd6e4a74b2f5bb5386658ce3bb146cd040a1867646ab3b973fb6632b1",
"fragment_id": "w1-embed",
"global_shape": [
4
],
"local_shape": [
4
],
"logical_tensor_id": "model.embed.weight",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "w1-embed",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": "tie-embedding-head",
"tie_group_id": "tie-embedding-head",
"topology_coordinates": [
[
"rank",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
2
],
"element_offset": [
0,
0
],
"evidence_refs": [
"synthetic:w1-expert"
],
"fqn": "model.experts.0.weight",
"fragment_digest": "sha256:4c9c4f354e74153db012329d71c8562ec23e498148174b2c49de58f45d47cdbe",
"fragment_id": "w1-expert",
"global_shape": [
2,
2
],
"local_shape": [
2,
2
],
"logical_tensor_id": "model.experts.0.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "w1-expert",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"ep",
0
],
[
"rank",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
4
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:w1-head"
],
"fqn": "lm_head.weight",
"fragment_digest": "sha256:ad73b9acd6e4a74b2f5bb5386658ce3bb146cd040a1867646ab3b973fb6632b1",
"fragment_id": "w1-head",
"global_shape": [
4
],
"local_shape": [
4
],
"logical_tensor_id": "lm_head.weight",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "w1-head",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": "tie-embedding-head",
"tie_group_id": "tie-embedding-head",
"topology_coordinates": [
[
"rank",
0
]
]
},
{
"component_id": "model",
"dtype": "int64",
"element_extent": [],
"element_offset": [],
"evidence_refs": [
"synthetic:w1-scalar"
],
"fqn": "model.step",
"fragment_digest": "sha256:35be322d094f9d154a8aba4733b8497f180353bd7ae7b0a15f90b586b549f28b",
"fragment_id": "w1-scalar",
"global_shape": [],
"local_shape": [],
"logical_tensor_id": "model.step",
"padding": {
"after": [],
"before": [],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "w1-scalar",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "persistent_buffer",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
3
],
"element_offset": [
0,
0
],
"evidence_refs": [
"synthetic:w1-weight"
],
"fqn": "model.weight",
"fragment_digest": "sha256:e2c0a71510b5394df7773b63fb5f54372b84c3564e67811bde7d665be227976d",
"fragment_id": "w1-weight",
"global_shape": [
2,
3
],
"local_shape": [
2,
3
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "w1-weight",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
0
]
]
}
],
"oracle_file": "oracles.safetensors",
"oracles": [
{
"dtype": "float32",
"logical_tensor_id": "lm_head.weight",
"payload_key": "tensor_000",
"sha256": "sha256:ad73b9acd6e4a74b2f5bb5386658ce3bb146cd040a1867646ab3b973fb6632b1",
"shape": [
4
]
},
{
"dtype": "float32",
"logical_tensor_id": "model.embed.weight",
"payload_key": "tensor_001",
"sha256": "sha256:ad73b9acd6e4a74b2f5bb5386658ce3bb146cd040a1867646ab3b973fb6632b1",
"shape": [
4
]
},
{
"dtype": "float32",
"logical_tensor_id": "model.experts.0.weight",
"payload_key": "tensor_002",
"sha256": "sha256:4c9c4f354e74153db012329d71c8562ec23e498148174b2c49de58f45d47cdbe",
"shape": [
2,
2
]
},
{
"dtype": "float32",
"logical_tensor_id": "model.running_mean",
"payload_key": "tensor_003",
"sha256": "sha256:ab2a21ab5e1262d555eea5678035e8fe3e76542c6b474cc04a16a39d4aa02645",
"shape": [
2
]
},
{
"dtype": "int64",
"logical_tensor_id": "model.step",
"payload_key": "tensor_004",
"sha256": "sha256:35be322d094f9d154a8aba4733b8497f180353bd7ae7b0a15f90b586b549f28b",
"shape": []
},
{
"dtype": "float32",
"logical_tensor_id": "model.weight",
"payload_key": "tensor_005",
"sha256": "sha256:e2c0a71510b5394df7773b63fb5f54372b84c3564e67811bde7d665be227976d",
"shape": [
2,
3
]
}
],
"schema_id": "obliteratus.checkpoint-fixture-case",
"schema_version": "1.0.0",
"topology": {
"source": {
"world_size": 1
},
"target": {
"world_size": 1
}
},
"world_size": 1
}
@@ -0,0 +1,150 @@
{
"case_id": "world2-uneven-1d",
"components": [
"full_model"
],
"expected_manifest_digest": "sha256:d0e2e96c6bdfb6ebd29470ef1f789b1a6a8f1ebc89624538cf56600ad3f4cb88",
"features": [
"padding",
"uneven_1d"
],
"fragments": [
{
"component_id": "model",
"dtype": "int64",
"element_extent": [
3
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:w2-r0"
],
"fqn": "model.weight",
"fragment_digest": "sha256:ab25350e3e65efebe24584461683ecda68725576e825e550038b90e7b1479946",
"fragment_id": "w2-r0",
"global_shape": [
7
],
"local_shape": [
4
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
1
],
"before": [
0
],
"semantic": "producer_declared"
},
"partition_axes": [
0
],
"payload_file": "rank-00000.safetensors",
"payload_key": "w2-r0",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
0
],
[
"tp",
0
]
]
},
{
"component_id": "model",
"dtype": "int64",
"element_extent": [
4
],
"element_offset": [
3
],
"evidence_refs": [
"synthetic:w2-r1"
],
"fqn": "model.weight",
"fragment_digest": "sha256:6df0e128a82e6d8c332ed546fca7d48406970ea60531b4d64ee41bdffc3d5da4",
"fragment_id": "w2-r1",
"global_shape": [
7
],
"local_shape": [
4
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [
0
],
"payload_file": "rank-00001.safetensors",
"payload_key": "w2-r1",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00001.safetensors",
"shard_file_id": "rank-00001.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
1
],
[
"tp",
1
]
]
}
],
"oracle_file": "oracles.safetensors",
"oracles": [
{
"dtype": "int64",
"logical_tensor_id": "model.weight",
"payload_key": "tensor_000",
"sha256": "sha256:81845a01dafa45c9b26e10a7af52a92e8604d5d8ef690f1e3ccdcfe3b5c6ae98",
"shape": [
7
]
}
],
"schema_id": "obliteratus.checkpoint-fixture-case",
"schema_version": "1.0.0",
"topology": {
"source": {
"tp": 2
},
"target": {
"world_size": 1
}
},
"world_size": 2
}
@@ -0,0 +1,255 @@
{
"case_id": "world4-dp-replicas",
"components": [
"full_model"
],
"expected_manifest_digest": "sha256:d45ab4d7db81d4f7bd55e7ccb04bfb7c936b0bd320f1b57e77d7d02d3d488ff3",
"features": [
"dp_replica"
],
"fragments": [
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
3
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:dp-r0"
],
"fqn": "model.weight",
"fragment_digest": "sha256:7f8a2918261b8d80cae770a9aae209b29d8b39654c62aee0013a423796ddb655",
"fragment_id": "dp-r0",
"global_shape": [
3
],
"local_shape": [
3
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00000.safetensors",
"payload_key": "dp-r0",
"replica": {
"group_id": "dp-full",
"member_count": 4,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"dp",
0
],
[
"rank",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
3
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:dp-r1"
],
"fqn": "model.weight",
"fragment_digest": "sha256:7f8a2918261b8d80cae770a9aae209b29d8b39654c62aee0013a423796ddb655",
"fragment_id": "dp-r1",
"global_shape": [
3
],
"local_shape": [
3
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00001.safetensors",
"payload_key": "dp-r1",
"replica": {
"group_id": "dp-full",
"member_count": 4,
"member_index": 1
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00001.safetensors",
"shard_file_id": "rank-00001.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"dp",
1
],
[
"rank",
1
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
3
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:dp-r2"
],
"fqn": "model.weight",
"fragment_digest": "sha256:7f8a2918261b8d80cae770a9aae209b29d8b39654c62aee0013a423796ddb655",
"fragment_id": "dp-r2",
"global_shape": [
3
],
"local_shape": [
3
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00002.safetensors",
"payload_key": "dp-r2",
"replica": {
"group_id": "dp-full",
"member_count": 4,
"member_index": 2
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00002.safetensors",
"shard_file_id": "rank-00002.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"dp",
2
],
[
"rank",
2
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
3
],
"element_offset": [
0
],
"evidence_refs": [
"synthetic:dp-r3"
],
"fqn": "model.weight",
"fragment_digest": "sha256:7f8a2918261b8d80cae770a9aae209b29d8b39654c62aee0013a423796ddb655",
"fragment_id": "dp-r3",
"global_shape": [
3
],
"local_shape": [
3
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0
],
"before": [
0
],
"semantic": "none"
},
"partition_axes": [],
"payload_file": "rank-00003.safetensors",
"payload_key": "dp-r3",
"replica": {
"group_id": "dp-full",
"member_count": 4,
"member_index": 3
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00003.safetensors",
"shard_file_id": "rank-00003.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"dp",
3
],
[
"rank",
3
]
]
}
],
"oracle_file": "oracles.safetensors",
"oracles": [
{
"dtype": "float32",
"logical_tensor_id": "model.weight",
"payload_key": "tensor_000",
"sha256": "sha256:7f8a2918261b8d80cae770a9aae209b29d8b39654c62aee0013a423796ddb655",
"shape": [
3
]
}
],
"schema_id": "obliteratus.checkpoint-fixture-case",
"schema_version": "1.0.0",
"topology": {
"source": {
"dp": 4
},
"target": {
"world_size": 1
}
},
"world_size": 4
}
@@ -0,0 +1,309 @@
{
"case_id": "world4-uneven-2d",
"components": [
"full_model"
],
"expected_manifest_digest": "sha256:356bddfa143562828982c2fd80b83d4a2e0c278814d844568b99aeed1f6cb310",
"features": [
"uneven_2d"
],
"fragments": [
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
3
],
"element_offset": [
0,
0
],
"evidence_refs": [
"synthetic:w4-r0"
],
"fqn": "model.weight",
"fragment_digest": "sha256:e902ad65c5c6702e7ca1da427d13ec926431470bc155ff2983b362cb10d2dff5",
"fragment_id": "w4-r0",
"global_shape": [
5,
7
],
"local_shape": [
2,
3
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00000.safetensors",
"payload_key": "w4-r0",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00000.safetensors",
"shard_file_id": "rank-00000.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
0
],
[
"tp_col",
0
],
[
"tp_row",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
2,
4
],
"element_offset": [
0,
3
],
"evidence_refs": [
"synthetic:w4-r1"
],
"fqn": "model.weight",
"fragment_digest": "sha256:8dc8137a498004d656f20c18b4726bf12aab21d7010b9980480317d6df38e8c8",
"fragment_id": "w4-r1",
"global_shape": [
5,
7
],
"local_shape": [
2,
4
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00001.safetensors",
"payload_key": "w4-r1",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00001.safetensors",
"shard_file_id": "rank-00001.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
1
],
[
"tp_col",
1
],
[
"tp_row",
0
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
3,
3
],
"element_offset": [
2,
0
],
"evidence_refs": [
"synthetic:w4-r2"
],
"fqn": "model.weight",
"fragment_digest": "sha256:cd2d91a4484f0cd6236034895436acbee187e02fd96900593a2512d9882f7e0e",
"fragment_id": "w4-r2",
"global_shape": [
5,
7
],
"local_shape": [
3,
3
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00002.safetensors",
"payload_key": "w4-r2",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00002.safetensors",
"shard_file_id": "rank-00002.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
2
],
[
"tp_col",
0
],
[
"tp_row",
1
]
]
},
{
"component_id": "model",
"dtype": "float32",
"element_extent": [
3,
4
],
"element_offset": [
2,
3
],
"evidence_refs": [
"synthetic:w4-r3"
],
"fqn": "model.weight",
"fragment_digest": "sha256:3c4e748352f778507efe297633562c96b36d3e68030f1c230266d63fd2994179",
"fragment_id": "w4-r3",
"global_shape": [
5,
7
],
"local_shape": [
3,
4
],
"logical_tensor_id": "model.weight",
"padding": {
"after": [
0,
0
],
"before": [
0,
0
],
"semantic": "none"
},
"partition_axes": [
0,
1
],
"payload_file": "rank-00003.safetensors",
"payload_key": "w4-r3",
"replica": {
"group_id": null,
"member_count": 1,
"member_index": 0
},
"role": "parameter",
"shard_digest_ref": "source-digest:rank-00003.safetensors",
"shard_file_id": "rank-00003.safetensors",
"shared_storage_id": null,
"tie_group_id": null,
"topology_coordinates": [
[
"rank",
3
],
[
"tp_col",
1
],
[
"tp_row",
1
]
]
}
],
"oracle_file": "oracles.safetensors",
"oracles": [
{
"dtype": "float32",
"logical_tensor_id": "model.weight",
"payload_key": "tensor_000",
"sha256": "sha256:471d327907fc83cb6703d3424393e5caeefd627fa86d8b1b2f07d3045b6e1433",
"shape": [
5,
7
]
}
],
"schema_id": "obliteratus.checkpoint-fixture-case",
"schema_version": "1.0.0",
"topology": {
"source": {
"tp_columns": 2,
"tp_rows": 2
},
"target": {
"world_size": 1
}
},
"world_size": 4
}
@@ -0,0 +1,233 @@
{
"cases": [
{
"case_id": "mixed-model-peft",
"features": [
"mixed_full_model_peft"
],
"files": [
{
"relative_path": "case.json",
"sha256": "sha256:aab22d557fdc3de06143a8db3222680fb5db42eacc2a782e79882ccf200417c6",
"size_bytes": 3577
},
{
"relative_path": "oracles.safetensors",
"sha256": "sha256:74d4b70269f0c5ad034c5291fb63541629de99a4103c8190215d4c30f3387e1a",
"size_bytes": 168
},
{
"relative_path": "rank-00000.safetensors",
"sha256": "sha256:a7cac4935027f303e4c4b37ba93726f17c49e451c2f49e9ce9b6db2c7c82721f",
"size_bytes": 96
},
{
"relative_path": "rank-00001.safetensors",
"sha256": "sha256:d9231d15d8fbc7b6f80014d4fc8b9da125c0bfcbff219ef8a7941076871260e3",
"size_bytes": 88
}
],
"relative_path": "cases/mixed-model-peft",
"world_size": 2
},
{
"case_id": "tp2-pp2-to-single",
"features": [
"pipeline_parallel",
"topology_a_to_b"
],
"files": [
{
"relative_path": "case.json",
"sha256": "sha256:7a667928a5f142449b2309aa6c22b8cf1996c4fe2d96558fb36f6acc2c5ab629",
"size_bytes": 6235
},
{
"relative_path": "oracles.safetensors",
"sha256": "sha256:03a1b5c42ee9de97d2669f6b7d21dbc30f87edd9bd988e1e154fc03f88b72298",
"size_bytes": 144
},
{
"relative_path": "rank-00000.safetensors",
"sha256": "sha256:7d721dd88baac995c4070fcc639de7fff9d7ff1922b1f7076dcc3470b41c34ab",
"size_bytes": 88
},
{
"relative_path": "rank-00001.safetensors",
"sha256": "sha256:13b231f41a6f11c160dab5cdd58184f9ba0e8b6e97606ea118d2785c6322f7ec",
"size_bytes": 88
},
{
"relative_path": "rank-00002.safetensors",
"sha256": "sha256:f1aa964072188df5bc93fee44a1a139c4f2a82811aefb946a5a7ea6aa38cbe9d",
"size_bytes": 88
},
{
"relative_path": "rank-00003.safetensors",
"sha256": "sha256:620e59497637e8038383b6e926b51e07a63943ef340c2e1b1d48a55a903459c2",
"size_bytes": 88
}
],
"relative_path": "cases/tp2-pp2-to-single",
"world_size": 4
},
{
"case_id": "world1-complete",
"features": [
"buffer",
"expert",
"scalar",
"tied_weight"
],
"files": [
{
"relative_path": "case.json",
"sha256": "sha256:8bd58508b678460697f7a2315ab4f2e48798278a38111479ee518661839a98b2",
"size_bytes": 9218
},
{
"relative_path": "oracles.safetensors",
"sha256": "sha256:eb87c4db023b1272325d27e37e71511240f3cdf80df1712d261fde7d0f7b64b6",
"size_bytes": 488
},
{
"relative_path": "rank-00000.safetensors",
"sha256": "sha256:e42a2a167781454d9f6df4a9f870cd732ba237fd32960b5a66446bda1ed5c458",
"size_bytes": 472
}
],
"relative_path": "cases/world1-complete",
"world_size": 1
},
{
"case_id": "world2-uneven-1d",
"features": [
"padding",
"uneven_1d"
],
"files": [
{
"relative_path": "case.json",
"sha256": "sha256:ad350b19d728c3d45961a6695d581dfd44158f389d7cf06420c444bc6e6eb358",
"size_bytes": 3205
},
{
"relative_path": "oracles.safetensors",
"sha256": "sha256:7e922eba8c68cf28a0444625c79a3c5f88274e47653dae885e8a493ea946b179",
"size_bytes": 128
},
{
"relative_path": "rank-00000.safetensors",
"sha256": "sha256:6fac58511adb9184da344e3f6c21fc5e90c42d2365c951863cf72dcc3a868c6b",
"size_bytes": 104
},
{
"relative_path": "rank-00001.safetensors",
"sha256": "sha256:fa4fb09a4ef30971540267395c156dc60a31390a936cd754280b8a658b56bb09",
"size_bytes": 104
}
],
"relative_path": "cases/world2-uneven-1d",
"world_size": 2
},
{
"case_id": "world4-dp-replicas",
"features": [
"dp_replica"
],
"files": [
{
"relative_path": "case.json",
"sha256": "sha256:61840aaec0045c1d648b5a4bdd93fe66612f66844ff506d1f9d98b88f5e221d5",
"size_bytes": 5576
},
{
"relative_path": "oracles.safetensors",
"sha256": "sha256:42ed9ed388b42f5f0813713f449abec59446c406eca9aef376534ee49394be8e",
"size_bytes": 84
},
{
"relative_path": "rank-00000.safetensors",
"sha256": "sha256:66354f04b342b9c935a54ec5ebf3ddfa6da49fc1c549bf76ca886dc8a9d6d428",
"size_bytes": 84
},
{
"relative_path": "rank-00001.safetensors",
"sha256": "sha256:a58e05fb72c9e48ecf24444e150869225ccdb9ddaec1107015171d1d687e7b18",
"size_bytes": 84
},
{
"relative_path": "rank-00002.safetensors",
"sha256": "sha256:ed9633e59e38118d3e8b7121da9c8b8893e3ab2b00bfc4ccb3947350bdc615ef",
"size_bytes": 84
},
{
"relative_path": "rank-00003.safetensors",
"sha256": "sha256:07ac85546ac39bfea2047461b45ea99f68e76adf65dec5f8c63b544d9c32bd32",
"size_bytes": 84
}
],
"relative_path": "cases/world4-dp-replicas",
"world_size": 4
},
{
"case_id": "world4-uneven-2d",
"features": [
"uneven_2d"
],
"files": [
{
"relative_path": "case.json",
"sha256": "sha256:4ccce62bc46cd9738b0ea21de982438dde8edca10ae2a565d6d8d1094482af06",
"size_bytes": 6212
},
{
"relative_path": "oracles.safetensors",
"sha256": "sha256:7109b6b4db6043689e5483ff56422079f4778abc5f92e375e5ee8e0b564cb06a",
"size_bytes": 220
},
{
"relative_path": "rank-00000.safetensors",
"sha256": "sha256:a062693caed152b759d04229348c5ef9d27f4276ffed7ad5b21cc9e3d381ffa6",
"size_bytes": 96
},
{
"relative_path": "rank-00001.safetensors",
"sha256": "sha256:3932804176b74fe5573e4f4f9496ab95ca911aa0cc2f84d6d2b5553a3bbead84",
"size_bytes": 104
},
{
"relative_path": "rank-00002.safetensors",
"sha256": "sha256:2dc96b01af5c564018de0dd8f0e94783c9dd4321e307901dab486d2456a6e654",
"size_bytes": 108
},
{
"relative_path": "rank-00003.safetensors",
"sha256": "sha256:523140f0770d0b2145d72d03bda03c702455f6b57a7bf6efb35a479eb118b3df",
"size_bytes": 120
}
],
"relative_path": "cases/world4-uneven-2d",
"world_size": 4
}
],
"generator": {
"path": "scripts/generate_checkpoint_fixtures.py",
"version": "1.0.0"
},
"license": "AGPL-3.0-or-later",
"limits": {
"max_case_bytes": 65536,
"max_cases": 16,
"max_files_per_case": 16,
"max_tensors_per_case": 16
},
"provenance": {
"kind": "deterministic_synthetic",
"seed": 0,
"third_party_data": false,
"third_party_weights": false
},
"schema_id": "obliteratus.checkpoint-fixture-corpus",
"schema_version": "1.0.0"
}
@@ -0,0 +1,108 @@
{
"cases": [
{
"case_id": "negative-01",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "coverage_gap",
"mutation": "deterministic:coverage_gap"
},
{
"case_id": "negative-02",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "coverage_overlap",
"mutation": "deterministic:coverage_overlap"
},
{
"case_id": "negative-03",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "dimension_mismatch",
"mutation": "deterministic:dimension_mismatch"
},
{
"case_id": "negative-04",
"expected_code": "DCI_SOURCE_BOUNDARY_VIOLATION",
"failure": "extra_shard",
"mutation": "deterministic:extra_shard"
},
{
"case_id": "negative-05",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "fragment_out_of_bounds",
"mutation": "deterministic:fragment_out_of_bounds"
},
{
"case_id": "negative-06",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "integer_overflow",
"mutation": "deterministic:integer_overflow"
},
{
"case_id": "negative-07",
"expected_code": "DCI_SOURCE_BOUNDARY_VIOLATION",
"failure": "missing_shard",
"mutation": "deterministic:missing_shard"
},
{
"case_id": "negative-08",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "negative_integer",
"mutation": "deterministic:negative_integer"
},
{
"case_id": "negative-09",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "padding_shape_mismatch",
"mutation": "deterministic:padding_shape_mismatch"
},
{
"case_id": "negative-10",
"expected_code": "DCI_SOURCE_BOUNDARY_VIOLATION",
"failure": "path_traversal",
"mutation": "deterministic:path_traversal"
},
{
"case_id": "negative-11",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "payload_dtype_mismatch",
"mutation": "deterministic:payload_dtype_mismatch"
},
{
"case_id": "negative-12",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "payload_shape_mismatch",
"mutation": "deterministic:payload_shape_mismatch"
},
{
"case_id": "negative-13",
"expected_code": "DCI_VALIDATION_FAILED",
"failure": "replica_digest_mismatch",
"mutation": "deterministic:replica_digest_mismatch"
},
{
"case_id": "negative-14",
"expected_code": "DCI_RESOURCE_LIMIT",
"failure": "resource_manifest_bomb",
"mutation": "deterministic:resource_manifest_bomb"
},
{
"case_id": "negative-15",
"expected_code": "DCI_SOURCE_BOUNDARY_VIOLATION",
"failure": "source_special_file",
"mutation": "deterministic:source_special_file"
},
{
"case_id": "negative-16",
"expected_code": "DCI_SOURCE_BOUNDARY_VIOLATION",
"failure": "source_symlink",
"mutation": "deterministic:source_symlink"
},
{
"case_id": "negative-17",
"expected_code": "DCI_SOURCE_BOUNDARY_VIOLATION",
"failure": "truncated_shard",
"mutation": "deterministic:truncated_shard"
}
],
"schema_id": "obliteratus.checkpoint-negative-fixtures",
"schema_version": "1.0.0"
}