Finalize verified models/FCP-SmolLM2-360M-20B/checkpoints/20B
Browse files
models/FCP-SmolLM2-360M-20B/checkpoints/20B/README.md
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# FCP 20B model
|
| 2 |
+
|
| 3 |
+
Artifacts for the SmolLM2-360M 20B comparison. Stored under `models/FCP-SmolLM2-360M-20B/checkpoints/20B/` in `meihant/AnonymousModel-mht` (a model repository).
|
| 4 |
+
|
| 5 |
+
{
|
| 6 |
+
"step": 95158,
|
| 7 |
+
"target_source_tokens": 20000000000,
|
| 8 |
+
"source_tokens": 20000000040,
|
| 9 |
+
"training_tokens": 49672486395,
|
| 10 |
+
"checkpoint_kind": "evaluation",
|
| 11 |
+
"resumable": false,
|
| 12 |
+
"model": "SmolLM2-360M",
|
| 13 |
+
"precision": "BF16 parameters and AdamW moments",
|
| 14 |
+
"context": 4096,
|
| 15 |
+
"global_batch": 128,
|
| 16 |
+
"packing": "BFD",
|
| 17 |
+
"learning_rate": 0.0005,
|
| 18 |
+
"warmup_steps": 2000,
|
| 19 |
+
"scheduler": "constant_with_warmup",
|
| 20 |
+
"final_checkpoint": true
|
| 21 |
+
}
|
| 22 |
+
|
| 23 |
+
See manifest.json for artifact hashes and completeness. The original corpus is the shared nominal-20B FineWeb-Edu source. BFD uses 4096-token sequences and 1000-record candidate buffers.
|
models/FCP-SmolLM2-360M-20B/checkpoints/20B/manifest.json
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"method": "fcp",
|
| 3 |
+
"kind": "model",
|
| 4 |
+
"step": 95158,
|
| 5 |
+
"target_source_tokens": 20000000000,
|
| 6 |
+
"source_tokens": 20000000040,
|
| 7 |
+
"training_tokens": 49672486395,
|
| 8 |
+
"checkpoint_kind": "evaluation",
|
| 9 |
+
"resumable": false,
|
| 10 |
+
"model": "SmolLM2-360M",
|
| 11 |
+
"precision": "BF16 parameters and AdamW moments",
|
| 12 |
+
"context": 4096,
|
| 13 |
+
"global_batch": 128,
|
| 14 |
+
"packing": "BFD",
|
| 15 |
+
"learning_rate": 0.0005,
|
| 16 |
+
"warmup_steps": 2000,
|
| 17 |
+
"scheduler": "constant_with_warmup",
|
| 18 |
+
"final_checkpoint": true,
|
| 19 |
+
"files": [
|
| 20 |
+
{
|
| 21 |
+
"path": "chat_template.jinja",
|
| 22 |
+
"bytes": 391,
|
| 23 |
+
"sha256": "d2771c9694cc22f4d983cbf2344551d4d0933156a927f673584a358e4aee0563"
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"path": "config.json",
|
| 27 |
+
"bytes": 775,
|
| 28 |
+
"sha256": "6fc5a8f7f998517da9963043785832ee8cc4954b9dd3ad9cb757e0c214db49de"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"path": "generation_config.json",
|
| 32 |
+
"bytes": 215,
|
| 33 |
+
"sha256": "fd4d3b53dcff228245b5e91b7f6b7729b9b1f7f6e4cf3cdce57cef26db0772ef"
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"path": "model.safetensors",
|
| 37 |
+
"bytes": 723686432,
|
| 38 |
+
"sha256": "940af68d592c5bf2a920c259c808c17943127800cb4a56ca8326f52a64dfa551"
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"path": "tokenizer.json",
|
| 42 |
+
"bytes": 3523968,
|
| 43 |
+
"sha256": "87fd33a66131c780462a54d7d088ce20fa10216134144acd7e6818bbc36ef8ca"
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"path": "tokenizer_config.json",
|
| 47 |
+
"bytes": 388,
|
| 48 |
+
"sha256": "ac5a687ced18da96ff4bc754136508afe2d363a745a1b9c19eca280c1cdbdb47"
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"path": "training_progress.json",
|
| 52 |
+
"bytes": 3053,
|
| 53 |
+
"sha256": "380252045fcfff6b03a6cad9933109f63f559c6cc89232fb2c0fe0a5db6b410c"
|
| 54 |
+
}
|
| 55 |
+
]
|
| 56 |
+
}
|