exp002 Wave B seed-1 folded in: ordering budget-stable certified 2 seeds (mlp 13.899/13.889 vs relay_pw 13.983/13.932); relay convergence seed-dependent — claim corrected
Browse files
exp002_refine/README.md
CHANGED
|
@@ -18,8 +18,8 @@ pure Adam lr 1e-3 wd 0).
|
|
| 18 |
| **relay_pw (patchwork consumer)** | **13.982** | **14.012** | **winner — certified 2 seeds** |
|
| 19 |
| relay_deep (all budget, last 12 blocks) | 14.576 | — | coverage beats concentration (−0.48) |
|
| 20 |
| mlp_wide (hidden 128 control) | **NaN** | 13.930 | diverged 1/2 seeds; no gain over narrow |
|
| 21 |
-
| relay_pw @ 6000 steps | 13.983 |
|
| 22 |
-
| mlp64 @ 6000 steps (companion) | 13.899 |
|
| 23 |
|
| 24 |
References (exp001, certified): frozen 17.798; relay16 14.093/14.091; mlp64
|
| 25 |
13.927/13.935. `build_results.py` re-asserts every claim from
|
|
@@ -44,14 +44,15 @@ References (exp001, certified): frozen 17.798; relay16 14.093/14.091; mlp64
|
|
| 44 |
diverged to NaN at one of two seeds (50%) and gained nothing over the
|
| 45 |
narrow MLP at the other; every sphere-normalized relay variant — all five,
|
| 46 |
all seeds — trained without aid (0% divergence).
|
| 47 |
-
5. **The perplexity ordering is budget-stable
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
|
|
|
| 55 |
|
| 56 |
Checkpoint note: `adapters/` holds the latest checkpoint per (arm, seed);
|
| 57 |
for relay_pw s0 that is the 6000-step version (the 3000-step run's ppl is
|
|
|
|
| 18 |
| **relay_pw (patchwork consumer)** | **13.982** | **14.012** | **winner — certified 2 seeds** |
|
| 19 |
| relay_deep (all budget, last 12 blocks) | 14.576 | — | coverage beats concentration (−0.48) |
|
| 20 |
| mlp_wide (hidden 128 control) | **NaN** | 13.930 | diverged 1/2 seeds; no gain over narrow |
|
| 21 |
+
| relay_pw @ 6000 steps | 13.983 | 13.932 | s0 flat, s1 −0.08: convergence is seed-dependent |
|
| 22 |
+
| mlp64 @ 6000 steps (companion) | 13.899 | 13.889 | ordering budget-stable, both seeds |
|
| 23 |
|
| 24 |
References (exp001, certified): frozen 17.798; relay16 14.093/14.091; mlp64
|
| 25 |
13.927/13.935. `build_results.py` re-asserts every claim from
|
|
|
|
| 44 |
diverged to NaN at one of two seeds (50%) and gained nothing over the
|
| 45 |
narrow MLP at the other; every sphere-normalized relay variant — all five,
|
| 46 |
all seeds — trained without aid (0% divergence).
|
| 47 |
+
5. **The perplexity ordering is budget-stable — certified 2 seeds at 6000
|
| 48 |
+
steps** (mlp 13.899/13.889 vs relay_pw 13.983/13.932, gap 0.04–0.09).
|
| 49 |
+
Per-seed convergence differs (relay_pw flat at s0, −0.08 at s1), so the
|
| 50 |
+
robust claim is the ordering, not convergence. Free capacity holds a
|
| 51 |
+
small, persistent ppl edge on this well-trained substrate. Combined with
|
| 52 |
+
exp001's gate-vs-ppl dissociation, the relay's value at 0.5B is its
|
| 53 |
+
**stability, its opt-in mechanism, and its discrete addressable surface**
|
| 54 |
+
— the levers the next experiments (sign-code tap, frozen-keyer registry)
|
| 55 |
+
are built on — rather than raw perplexity.
|
| 56 |
|
| 57 |
Checkpoint note: `adapters/` holds the latest checkpoint per (arm, seed);
|
| 58 |
for relay_pw s0 that is the 6000-step version (the 3000-step run's ppl is
|
exp002_refine/adapters/q2_relay_pw_s1.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 22166937
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b15524b4191c13b9940b63e2c4db1d60224fb9709e31961e98f0cd06834d7797
|
| 3 |
size 22166937
|
exp002_refine/build_results.py
CHANGED
|
@@ -12,8 +12,8 @@ rows = [json.loads(l) for l in
|
|
| 12 |
comp = [json.loads(l) for l in
|
| 13 |
open(os.path.join(HERE, "results", "companion_mlp6k.jsonl"),
|
| 14 |
encoding="utf-8")]
|
| 15 |
-
assert all(r["exp"] == "q2" for r in rows) and len(rows) ==
|
| 16 |
-
assert len(comp) ==
|
| 17 |
|
| 18 |
def cell(arm, seed, steps=3000):
|
| 19 |
return next(r for r in rows if r["arm"] == arm and r["seed"] == seed
|
|
@@ -42,10 +42,16 @@ assert sum(1 for p in mw if math.isnan(p)) == 1, mw
|
|
| 42 |
assert any(abs(p - REF_MLP) < 0.01 for p in mw if not math.isnan(p))
|
| 43 |
assert all(not math.isnan(r["ppl"]) for r in rows if r["arm"] != "mlp_wide")
|
| 44 |
|
| 45 |
-
# claim 4 (budget,
|
| 46 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
assert abs(cell("relay_pw", 0, 6000)["ppl"] - cell("relay_pw", 0)["ppl"]) < 0.005
|
| 48 |
-
assert
|
| 49 |
|
| 50 |
out = {"ppl": {f'{r["arm"]}_s{r["seed"]}_t{r["steps"]}': r["ppl"] for r in rows},
|
| 51 |
"companion_mlp6k": comp[0]["ppl"], "n_rows": len(rows) + 1}
|
|
|
|
| 12 |
comp = [json.loads(l) for l in
|
| 13 |
open(os.path.join(HERE, "results", "companion_mlp6k.jsonl"),
|
| 14 |
encoding="utf-8")]
|
| 15 |
+
assert all(r["exp"] == "q2" for r in rows) and len(rows) == 10
|
| 16 |
+
assert len(comp) == 2 and all(c["steps"] == 6000 for c in comp)
|
| 17 |
|
| 18 |
def cell(arm, seed, steps=3000):
|
| 19 |
return next(r for r in rows if r["arm"] == arm and r["seed"] == seed
|
|
|
|
| 42 |
assert any(abs(p - REF_MLP) < 0.01 for p in mw if not math.isnan(p))
|
| 43 |
assert all(not math.isnan(r["ppl"]) for r in rows if r["arm"] != "mlp_wide")
|
| 44 |
|
| 45 |
+
# claim 4 (budget, 2 seeds): the ORDERING is budget-stable — the mlp stays
|
| 46 |
+
# ahead of relay_pw at 6000 steps at both seeds (gap 0.04-0.09); per-seed
|
| 47 |
+
# convergence differs (s0 relay flat 3k->6k, s1 improves 0.08) so the robust
|
| 48 |
+
# claim is the ordering, not convergence
|
| 49 |
+
mlp6k = {c["seed"]: c["ppl"] for c in comp}
|
| 50 |
+
for s in (0, 1):
|
| 51 |
+
r6 = cell("relay_pw", s, 6000)["ppl"]
|
| 52 |
+
assert mlp6k[s] < r6 and (r6 - mlp6k[s]) < 0.1, (s, r6, mlp6k[s])
|
| 53 |
assert abs(cell("relay_pw", 0, 6000)["ppl"] - cell("relay_pw", 0)["ppl"]) < 0.005
|
| 54 |
+
assert cell("relay_pw", 1)["ppl"] - cell("relay_pw", 1, 6000)["ppl"] > 0.05
|
| 55 |
|
| 56 |
out = {"ppl": {f'{r["arm"]}_s{r["seed"]}_t{r["steps"]}': r["ppl"] for r in rows},
|
| 57 |
"companion_mlp6k": comp[0]["ppl"], "n_rows": len(rows) + 1}
|
exp002_refine/results/companion_mlp6k.jsonl
CHANGED
|
@@ -1 +1,2 @@
|
|
| 1 |
{"exp": "q1", "arm": "mlp", "seed": 0, "ppl": 13.899, "steps": 6000, "trainable": 2775576, "vitals": {"gates": [0.0399, 0.0381, 0.0375, 0.0375, 0.0358, 0.0333, "..."], "gate_mean": 0.0292, "drift": []}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and Greece . The Babylonians , who lived in the 2nd millennium BC , were the first to develop a system of mathematics that included the use of fractions . The Egyptians , who lived in the 2nd millennium BC , developed a system of mathematics that included the use of fractions and the use of the decimal system . The Greeks ,", "In a small village by the sea,": " a young man named John is a member of the local gang . He is a member of the gang known as the \" Black Gang \" , and is a member of the gang known as the \" White Gang \" . John is a member of the gang known as the \" Black Gang \" , and is a member of the gang known as the \" White Gang \" . John is a member of the gang", "The most important principle of engineering is": " to ensure that the design is safe and reliable . The design of a building is a complex process that involves many factors , including the materials used , the construction methods , and the design of the building itself . The design of a building is also influenced by the building owner 's needs and preferences . \n The design of a building is a complex process that involves many factors , including the materials used , the"}}
|
|
|
|
|
|
| 1 |
{"exp": "q1", "arm": "mlp", "seed": 0, "ppl": 13.899, "steps": 6000, "trainable": 2775576, "vitals": {"gates": [0.0399, 0.0381, 0.0375, 0.0375, 0.0358, 0.0333, "..."], "gate_mean": 0.0292, "drift": []}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and Greece . The Babylonians , who lived in the 2nd millennium BC , were the first to develop a system of mathematics that included the use of fractions . The Egyptians , who lived in the 2nd millennium BC , developed a system of mathematics that included the use of fractions and the use of the decimal system . The Greeks ,", "In a small village by the sea,": " a young man named John is a member of the local gang . He is a member of the gang known as the \" Black Gang \" , and is a member of the gang known as the \" White Gang \" . John is a member of the gang known as the \" Black Gang \" , and is a member of the gang known as the \" White Gang \" . John is a member of the gang", "The most important principle of engineering is": " to ensure that the design is safe and reliable . The design of a building is a complex process that involves many factors , including the materials used , the construction methods , and the design of the building itself . The design of a building is also influenced by the building owner 's needs and preferences . \n The design of a building is a complex process that involves many factors , including the materials used , the"}}
|
| 2 |
+
{"exp": "q1", "arm": "mlp", "seed": 1, "ppl": 13.889, "steps": 6000, "trainable": 2775576, "vitals": {"gates": [0.0426, 0.0389, 0.0386, 0.0378, 0.0364, 0.0337, "..."], "gate_mean": 0.0299, "drift": []}, "samples": {"The history of mathematics begins": " with the ancient Babylonians , who used a sexagesimal ( base 60 ) number system . The Babylonians used a base 60 number system for their calculations , and the sexagesimal system was used for the calculation of the length of a day , the number of days in a year , and the number of seconds in a minute . The Babylonians also used a sexagesimal", "In a small village by the sea,": " a young man named John is a victim of a violent crime . He is a victim of a violent crime . He is a victim of a violent crime . He is a victim of a violent crime . He is a victim of a violent crime . He is a victim of a violent crime . He is a victim of a violent crime . He is a victim of a violent crime . He is a victim", "The most important principle of engineering is": " that the design of a structure should be based on the principles of physics . \n The most important principle of engineering is that the design of a structure should be based on the principles of physics . \n The most important principle of engineering is that the design of a structure should be based on the principles of physics . \n The most important principle of engineering is that the design of a structure should be based on the"}}
|
exp002_refine/results/ledger.jsonl
CHANGED
|
@@ -7,3 +7,4 @@
|
|
| 7 |
{"exp": "q2", "arm": "mlp_wide", "seed": 1, "ppl": 13.93, "steps": 3000, "trainable": 5529624, "vitals": {"gate_mean": 0.0366, "gates_head_tail": [0.0441, 0.0409, 0.0398, 0.0261, 0.0286, 0.0246], "drift_head_tail": null}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and Greece . The Babylonians , who lived in the 19th century BC , were the first to use a base @-@ 10 positional numeral system , and they were the first to use zero as a place holder . The Egyptians used a base @-@ 20 positional numeral system , and the Greeks used a", "In a small village by the sea,": " a young man named John is a skilled sailor and a skilled fisherman . He is also a skilled hunter and a skilled tracker . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled", "The most important principle of engineering is": " that of safety . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The"}}
|
| 8 |
{"exp": "q2", "arm": "relay32", "seed": 1, "ppl": 14.069, "steps": 3000, "trainable": 5511192, "vitals": {"gate_mean": 0.0613, "gates_head_tail": [0.0535, 0.0497, 0.0495, 0.1354, 0.1159, 0.0853], "drift_head_tail": [0.156, 0.147, 0.133, 0.261, 0.358, 0.185]}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and Greece . The Babylonians , for example , developed a system of mathematics that was based on the concept of zero , which was used to represent the absence of a quantity . The Egyptians used a sexagesimal ( base 60 ) numeral system , which was based on the fact that 60 is the number of days in a", "In a small village by the sea,": " there are 100 houses . Each house has a dog , and each dog has a collar . The number of dogs in the village is 10 more than the number of dogs in the village with a collar . How many dogs are there in the village with a collar ? \n To solve this problem , we can set up a system of equations based on the information given . Let 's", "The most important principle of engineering is": " to make the best use of the materials and energy available . The most important principle of engineering is to make the best use of the materials and energy available . \n The most important principle of engineering is to make the best use of the materials and energy available . The most important principle of engineering is to make the best use of the materials and energy available . \n The most important principle of engineering is to make"}}
|
| 9 |
{"exp": "q2", "arm": "relay_pw", "seed": 0, "ppl": 13.983, "steps": 6000, "trainable": 5517864, "vitals": {"gate_mean": 0.0387, "gates_head_tail": [0.0368, 0.0225, 0.0239, 0.0836, 0.0704, 0.0632], "drift_head_tail": [0.167, 0.131, 0.118, 0.132, 0.146, 0.154]}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and China . The Babylonians developed a base 60 positional notation system , and the Chinese developed a base 10 positional notation system . The ancient Greeks developed a base 10 positional notation system , and the Babylonians developed a base 60 positional notation system . The Babylonians used a base 60 positional notation system", "In a small village by the sea,": " a young man named Tom is a skilled sailor . He is a good swimmer and a good swimmer at sea . He is also a good swimmer on land . He is a good swimmer in the sea and on land . He is a good swimmer in the sea and on land . He is a good swimmer in the sea and on land . He is a good swimmer in", "The most important principle of engineering is": " to make the best use of the materials available . The materials used in the construction of a building are not only the materials used in the construction of the building itself , but also the materials used in the construction of the building 's supporting structures . The materials used in the construction of the supporting structures are not only the materials used in the construction of the supporting structures themselves , but also the materials used in"}}
|
|
|
|
|
|
| 7 |
{"exp": "q2", "arm": "mlp_wide", "seed": 1, "ppl": 13.93, "steps": 3000, "trainable": 5529624, "vitals": {"gate_mean": 0.0366, "gates_head_tail": [0.0441, 0.0409, 0.0398, 0.0261, 0.0286, 0.0246], "drift_head_tail": null}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and Greece . The Babylonians , who lived in the 19th century BC , were the first to use a base @-@ 10 positional numeral system , and they were the first to use zero as a place holder . The Egyptians used a base @-@ 20 positional numeral system , and the Greeks used a", "In a small village by the sea,": " a young man named John is a skilled sailor and a skilled fisherman . He is also a skilled hunter and a skilled tracker . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled thief . He is also a skilled", "The most important principle of engineering is": " that of safety . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The engineer is responsible for the safety of the people who use his services . The"}}
|
| 8 |
{"exp": "q2", "arm": "relay32", "seed": 1, "ppl": 14.069, "steps": 3000, "trainable": 5511192, "vitals": {"gate_mean": 0.0613, "gates_head_tail": [0.0535, 0.0497, 0.0495, 0.1354, 0.1159, 0.0853], "drift_head_tail": [0.156, 0.147, 0.133, 0.261, 0.358, 0.185]}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and Greece . The Babylonians , for example , developed a system of mathematics that was based on the concept of zero , which was used to represent the absence of a quantity . The Egyptians used a sexagesimal ( base 60 ) numeral system , which was based on the fact that 60 is the number of days in a", "In a small village by the sea,": " there are 100 houses . Each house has a dog , and each dog has a collar . The number of dogs in the village is 10 more than the number of dogs in the village with a collar . How many dogs are there in the village with a collar ? \n To solve this problem , we can set up a system of equations based on the information given . Let 's", "The most important principle of engineering is": " to make the best use of the materials and energy available . The most important principle of engineering is to make the best use of the materials and energy available . \n The most important principle of engineering is to make the best use of the materials and energy available . The most important principle of engineering is to make the best use of the materials and energy available . \n The most important principle of engineering is to make"}}
|
| 9 |
{"exp": "q2", "arm": "relay_pw", "seed": 0, "ppl": 13.983, "steps": 6000, "trainable": 5517864, "vitals": {"gate_mean": 0.0387, "gates_head_tail": [0.0368, 0.0225, 0.0239, 0.0836, 0.0704, 0.0632], "drift_head_tail": [0.167, 0.131, 0.118, 0.132, 0.146, 0.154]}, "samples": {"The history of mathematics begins": " with the ancient civilizations of Mesopotamia , Egypt , and China . The Babylonians developed a base 60 positional notation system , and the Chinese developed a base 10 positional notation system . The ancient Greeks developed a base 10 positional notation system , and the Babylonians developed a base 60 positional notation system . The Babylonians used a base 60 positional notation system", "In a small village by the sea,": " a young man named Tom is a skilled sailor . He is a good swimmer and a good swimmer at sea . He is also a good swimmer on land . He is a good swimmer in the sea and on land . He is a good swimmer in the sea and on land . He is a good swimmer in the sea and on land . He is a good swimmer in", "The most important principle of engineering is": " to make the best use of the materials available . The materials used in the construction of a building are not only the materials used in the construction of the building itself , but also the materials used in the construction of the building 's supporting structures . The materials used in the construction of the supporting structures are not only the materials used in the construction of the supporting structures themselves , but also the materials used in"}}
|
| 10 |
+
{"exp": "q2", "arm": "relay_pw", "seed": 1, "ppl": 13.932, "steps": 6000, "trainable": 5517864, "vitals": {"gate_mean": 0.0388, "gates_head_tail": [0.0353, 0.0231, 0.0234, 0.0795, 0.0684, 0.06], "drift_head_tail": [0.107, 0.116, 0.107, 0.133, 0.149, 0.138]}, "samples": {"The history of mathematics begins": " with the ancient Babylonians , who developed a system of mathematics that was based on the use of a base @-@ ten numeral system . The Babylonians used a base @-@ ten numeral system , which is still used today . The Babylonians also developed a system of mathematics that was based on the use of a base @-@ ten positional numeral system . The Babylonians used a base @", "In a small village by the sea,": " a young man named John is a fisherman . He is a good fisherman , but he is not very good at fishing . He is a good swimmer , but he is not very good at swimming . He is a good swimmer , but he is not very good at swimming . He is a good swimmer , but he is not very good at swimming . He is a good swimmer", "The most important principle of engineering is": " that of safety . The engineer is responsible for the safety of the public and the safety of the engineer himself . The engineer is responsible for the safety of the public and the safety of the engineer himself . The engineer is responsible for the safety of the public and the safety of the engineer himself . \n The engineer is responsible for the safety of the public and the safety of the engineer himself . The engineer is"}}
|
exp002_refine/results/results.json
CHANGED
|
@@ -8,8 +8,9 @@
|
|
| 8 |
"relay_pw_s1_t3000": 14.012,
|
| 9 |
"mlp_wide_s1_t3000": 13.93,
|
| 10 |
"relay32_s1_t3000": 14.069,
|
| 11 |
-
"relay_pw_s0_t6000": 13.983
|
|
|
|
| 12 |
},
|
| 13 |
"companion_mlp6k": 13.899,
|
| 14 |
-
"n_rows":
|
| 15 |
}
|
|
|
|
| 8 |
"relay_pw_s1_t3000": 14.012,
|
| 9 |
"mlp_wide_s1_t3000": 13.93,
|
| 10 |
"relay32_s1_t3000": 14.069,
|
| 11 |
+
"relay_pw_s0_t6000": 13.983,
|
| 12 |
+
"relay_pw_s1_t6000": 13.932
|
| 13 |
},
|
| 14 |
"companion_mlp6k": 13.899,
|
| 15 |
+
"n_rows": 11
|
| 16 |
}
|