Finalize verified models/KLLM-SmolLM2-360M-20B/checkpoints/16B
Browse files
models/KLLM-SmolLM2-360M-20B/checkpoints/16B/README.md
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# KLLM 16B model
|
| 2 |
+
|
| 3 |
+
Artifacts for the SmolLM2-360M 20B comparison. Stored under `models/KLLM-SmolLM2-360M-20B/checkpoints/16B/` in `meihant/AnonymousModel-mht` (a model repository).
|
| 4 |
+
|
| 5 |
+
{
|
| 6 |
+
"step": 31905,
|
| 7 |
+
"target_source_tokens": 16000000000,
|
| 8 |
+
"source_tokens": 16000000433,
|
| 9 |
+
"training_tokens": 16639228313,
|
| 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": false
|
| 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/KLLM-SmolLM2-360M-20B/checkpoints/16B/manifest.json
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"method": "kllm",
|
| 3 |
+
"kind": "model",
|
| 4 |
+
"step": 31905,
|
| 5 |
+
"target_source_tokens": 16000000000,
|
| 6 |
+
"source_tokens": 16000000433,
|
| 7 |
+
"training_tokens": 16639228313,
|
| 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": false,
|
| 19 |
+
"files": [
|
| 20 |
+
{
|
| 21 |
+
"path": "config.json",
|
| 22 |
+
"bytes": 778,
|
| 23 |
+
"sha256": "d0321fb8700678e23f4cd206f7958d7488ac982f5097bb7dedce6f24197fb3fa"
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"path": "generation_config.json",
|
| 27 |
+
"bytes": 194,
|
| 28 |
+
"sha256": "b8733fba00d21edda23e6f7cca5cfe0312e5236dfa2b253e9f389f4ad0a48f58"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"path": "model.safetensors",
|
| 32 |
+
"bytes": 723893792,
|
| 33 |
+
"sha256": "d33368d904d795fde20e22f1188fcf54d98383ef0f9a245ff424a095d860e939"
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"path": "tokenizer.json",
|
| 37 |
+
"bytes": 3544733,
|
| 38 |
+
"sha256": "f75fde26d0654775af4deb7f0cd0a4d94e7e92ce4b72ad4442aa9d352bc445f5"
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"path": "tokenizer_config.json",
|
| 42 |
+
"bytes": 466,
|
| 43 |
+
"sha256": "fd96339eaaad0449c7df04df9e38291e380418f7caa5157825d27da7a92b1751"
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"path": "training_progress.json",
|
| 47 |
+
"bytes": 663,
|
| 48 |
+
"sha256": "319170d330fa88e16f81475899e1b167cfa6d64536cda1b6e2f52a5b5abcfe98"
|
| 49 |
+
}
|
| 50 |
+
]
|
| 51 |
+
}
|