Preserve complete v5.5 scaling checkpoint base__64m__30tpp__arithmetic_simple__pinned-engine
Browse files- refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/complete.json +28 -0
- refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/rl.pt +3 -0
- refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/sft.pt +3 -0
- refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/terminal-sft.json +7 -0
- refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/validated.json +39 -0
refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/complete.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"task": "arithmetic_simple",
|
| 3 |
+
"model": "64m",
|
| 4 |
+
"sft_tokens": 32768,
|
| 5 |
+
"sft_budget": 32768,
|
| 6 |
+
"rl_rollouts": 1024,
|
| 7 |
+
"rl_budget": 1024,
|
| 8 |
+
"terminal_sft_hash": "f23f452df84a94f5812744327f87171e7918aeddbb5ffb7a7d31df67448a2562",
|
| 9 |
+
"final_rl_hash": "eb88ee1914a13e24135c6c3eb90a58474fc0cc5311be3912339bea5a4032223b",
|
| 10 |
+
"rl_epochs": 1,
|
| 11 |
+
"rl_optimizer_updates": 8,
|
| 12 |
+
"sft_seconds": 3.0106322383508086,
|
| 13 |
+
"rl_seconds": 7.499335938133299,
|
| 14 |
+
"no_checkpoint_selection": true,
|
| 15 |
+
"evaluation_generations": 224,
|
| 16 |
+
"calibration_generations": 0,
|
| 17 |
+
"hint_distribution": "Uniform integer 1..16 per training query",
|
| 18 |
+
"samples_per_query": 32,
|
| 19 |
+
"development_episodes": 16,
|
| 20 |
+
"development_unique_groups": 13,
|
| 21 |
+
"evaluation_hint_max": 16,
|
| 22 |
+
"new_sft_tokens": 32768,
|
| 23 |
+
"sft_parent": null,
|
| 24 |
+
"rollout_temperature": 1.0,
|
| 25 |
+
"code_training_reward": "binary",
|
| 26 |
+
"code_origin_balanced": false,
|
| 27 |
+
"math_training_band": "all"
|
| 28 |
+
}
|
refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/rl.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a60264b0bdf1ad38286885097f33af84054bf6bf795e3a392ff0bda122c78e98
|
| 3 |
+
size 254038237
|
refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/sft.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:257f7e8af52e2bbeaaff352fa3e6028e414c0473aa18111daa8b34835126c550
|
| 3 |
+
size 254038368
|
refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/terminal-sft.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"tokens": 32768,
|
| 3 |
+
"budget": 32768,
|
| 4 |
+
"weights_sha256": "f23f452df84a94f5812744327f87171e7918aeddbb5ffb7a7d31df67448a2562",
|
| 5 |
+
"seconds": 3.0106322383508086,
|
| 6 |
+
"role": "Only terminal full-budget SFT weights initialize RL"
|
| 7 |
+
}
|
refined-tasks-v5.5-scaling/smokes/base__64m__30tpp__arithmetic_simple__pinned-engine/validated.json
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"case": "base__64m__30tpp__arithmetic_simple__pinned-engine",
|
| 3 |
+
"mix": "base",
|
| 4 |
+
"model": "64m",
|
| 5 |
+
"tpp": 30,
|
| 6 |
+
"source": "/nlp/scr/dhei/tiny-rl/experiments/refined-tasks-v5.5-scaling/source/8a8524d8e178b23e",
|
| 7 |
+
"pretrained_completion_verified": true,
|
| 8 |
+
"model_sha256": "ca3ac0c59db71de1bd56c29d6f3b51a419085084e0ab5d421da7f9debf47e1b9",
|
| 9 |
+
"complete": {
|
| 10 |
+
"task": "arithmetic_simple",
|
| 11 |
+
"model": "64m",
|
| 12 |
+
"sft_tokens": 32768,
|
| 13 |
+
"sft_budget": 32768,
|
| 14 |
+
"rl_rollouts": 1024,
|
| 15 |
+
"rl_budget": 1024,
|
| 16 |
+
"terminal_sft_hash": "f23f452df84a94f5812744327f87171e7918aeddbb5ffb7a7d31df67448a2562",
|
| 17 |
+
"final_rl_hash": "eb88ee1914a13e24135c6c3eb90a58474fc0cc5311be3912339bea5a4032223b",
|
| 18 |
+
"rl_epochs": 1,
|
| 19 |
+
"rl_optimizer_updates": 8,
|
| 20 |
+
"sft_seconds": 3.0106322383508086,
|
| 21 |
+
"rl_seconds": 7.499335938133299,
|
| 22 |
+
"no_checkpoint_selection": true,
|
| 23 |
+
"evaluation_generations": 224,
|
| 24 |
+
"calibration_generations": 0,
|
| 25 |
+
"hint_distribution": "Uniform integer 1..16 per training query",
|
| 26 |
+
"samples_per_query": 32,
|
| 27 |
+
"development_episodes": 16,
|
| 28 |
+
"development_unique_groups": 13,
|
| 29 |
+
"evaluation_hint_max": 16,
|
| 30 |
+
"new_sft_tokens": 32768,
|
| 31 |
+
"sft_parent": null,
|
| 32 |
+
"rollout_temperature": 1.0,
|
| 33 |
+
"code_training_reward": "binary",
|
| 34 |
+
"code_origin_balanced": false,
|
| 35 |
+
"math_training_band": "all"
|
| 36 |
+
},
|
| 37 |
+
"smoke": true,
|
| 38 |
+
"raw_initialization_verified": false
|
| 39 |
+
}
|