conductor(deob_pass3): fix line endings - rewrite cluster A files with CRLF and proper newlines
This commit is contained in:
Tier 2 Tech Lead
+226
-1
File diff suppressed because one or more lines are too long
+19
-21
@@ -1,17 +1,17 @@
|
||||
# cs229_building_llms — Per-term Decoder (tier-categorized)
|
||||
# cs229_building_llms - Per-term Decoder (tier-categorized)
|
||||
|
||||
**Source:** `cs229_building_llms_deobfuscated.md` (Pass 2 deobfuscation)
|
||||
**Target:** `cs229_building_llms.c` (the C11 program)
|
||||
**Method:** Per v2 lexicon §1.3 (etymology) + §2 (the 4 tiers)
|
||||
**Method:** Per v2 lexicon section 1.3 (etymology) + section 2 (the 4 tiers)
|
||||
|
||||
## Tier 1: Core concepts
|
||||
|
||||
| Term | C11 form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `Token` | `typedef uint32_t Token` | Old English *tacen* ("sign"); in LMs, an atomic unit of text | Tier 1 | Cluster 0, 2 |
|
||||
| `Logits` | `typedef struct { F32_R data; U4 len; } Logits` | Greek *λογιστικός* ("of reasoning"); raw pre-softmax scores | Tier 1 | Cluster 1 |
|
||||
| `Probability` | `typedef struct { F32 value; } Probability` | Latin *probabilitas* ("likelihood"); bounded to [0, 1] | Tier 1 | Tier 1 #1.13, Cluster 0 |
|
||||
| `Scalar` | `typedef struct { F32 value; } Scalar` | Latin *scalaris* ("of a ladder"); per-user placeholder for linear/geo/tensor alg | Tier 1 | Tier 4 #4.22 |
|
||||
| `Logits` | `typedef struct { F32_R data; U4 len; } Logits` | Greek *logistike* ("of reasoning"); raw pre-softmax scores | Tier 1 | Cluster 1 |
|
||||
| `Probability` | `typedef struct { F32 value; } Probability` | Latin *probabilitas* ("likelihood"); bounded to [0, 1] | Tier 1 | Cluster 0 |
|
||||
| `Scalar` | `typedef struct { F32 value; } Scalar` | Latin *scalaris* ("of a ladder"); per-user placeholder for linear/geo/tensor alg | Tier 1 | Cluster 0 |
|
||||
|
||||
## Tier 2: Data-oriented pipeline terms
|
||||
|
||||
@@ -21,7 +21,7 @@
|
||||
| `bpe_encode` | procedure | abbreviation: BPE + encode | Tier 2 | Cluster 9 |
|
||||
| `ar_forward` | procedure | abbreviation: autoregressive + forward pass | Tier 2 | Cluster 2 |
|
||||
| `ar_sample` | procedure | abbreviation: autoregressive + sample | Tier 2 | Cluster 2 |
|
||||
| `softmax_at` | procedure | softmax + index access | Tier 2 | Tier 4 #4.10 (user-specific `'scalar product'`) |
|
||||
| `softmax_at` | procedure | softmax + index access | Tier 2 | Cluster 2 |
|
||||
| `cross_entropy_loss` | procedure | cross-entropy + loss | Tier 2 | Cluster 2 |
|
||||
| `chinchilla_optimal_N` | procedure | Chinchilla scaling law (Hoffmann et al. 2022) + optimal model size | Tier 2 | Cluster 2 |
|
||||
| `chinchilla_optimal_D` | procedure | Chinchilla + optimal dataset size | Tier 2 | Cluster 2 |
|
||||
@@ -32,27 +32,25 @@
|
||||
|
||||
| Term | C11 form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `TokenSequence` | `typedef struct { Token_R ids; U4 len; U4 cap; } TokenSequence` | sequence of tokens; pair of (ptr, len, cap) | Tier 3 | Cluster 3 (Pair<A, B>) |
|
||||
| `Vocabulary` | `typedef struct { U4 vocab_size; Token_R id_to_token; U4_R token_to_id; } Vocabulary` | the lookup table; bidirectional id<->token | Tier 3 | Cluster 3 (Map<name, Tensor>) |
|
||||
| `TokenSequence` | `typedef struct { Token_R ids; U4 len; U4 cap; } TokenSequence` | sequence of tokens; pair of (ptr, len, cap) | Tier 3 | Cluster 3 (Pair) |
|
||||
| `Vocabulary` | `typedef struct { U4 vocab_size; Token_R id_to_token; U4_R token_to_id; } Vocabulary` | the lookup table; bidirectional id-token | Tier 3 | Cluster 3 |
|
||||
| `BPEMerge` | `typedef struct { Token a; Token b; Token merged; U4 count; } BPEMerge` | one BPE merge rule | Tier 3 | Cluster 3 |
|
||||
| `BPEMerger` | `typedef struct { BPEMerge_R merges; U4 len; U4 cap; U4 vocab_size; } BPEMerger` | the learned BPE merger (a sequence of merges) | Tier 3 | Cluster 3 (Sequence) |
|
||||
| `Slice` | `typedef struct { U4 ptr, len; } Slice` | (ptr, len) pair; per `c11_convention.md` §8 | Tier 3 | Cluster 3 (Pair) |
|
||||
| `BPEMerger` | `typedef struct { BPEMerge_R merges; U4 len; U4 cap; U4 vocab_size; } BPEMerger` | the learned BPE merger | Tier 3 | Cluster 3 |
|
||||
| `Slice` | `typedef struct { U4 ptr, len; } Slice` | (ptr, len) pair; per c11_convention section 8 | Tier 3 | Cluster 3 |
|
||||
|
||||
## Tier 4: AI-fuzzing tolerance terms
|
||||
|
||||
| Term | C11 form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `hidden_dim` | `U4` parameter | the embedding dimension d; bounded integer | Tier 4 | cs229 §2.6 |
|
||||
| `vocab_size` | `U4` parameter | the vocabulary size |V|; bounded integer | Tier 4 | cs229 §2.6 |
|
||||
| `compute_budget` | `F32` parameter | the FLOPs budget C; placeholder, resolved as float32 | Tier 4 | cs229 §2.4 (R1) |
|
||||
| `correlation` | (not used in this video) | `correlation : float64` (per v2 lexicon Tier 4 #4.25, R1 NEW v2) | Tier 4 | cs229 §2.6 |
|
||||
| `weakly_coupled` | (not used in this video) | `weakly_coupled(a, b, tolerance)` (per v2 lexicon §9.1) | Tier 4 | v2 lexicon §9.1 |
|
||||
| `hidden_dim` | `U4` parameter | the embedding dimension d; bounded integer | Tier 4 | cs229 section 2.6 |
|
||||
| `vocab_size` | `U4` parameter | the vocabulary size; bounded integer | Tier 4 | cs229 section 2.6 |
|
||||
| `compute_budget` | `F32` parameter | the FLOPs budget C; placeholder, resolved as float32 | Tier 4 | cs229 section 2.4 |
|
||||
|
||||
## Etymology notes (per Cluster 7, Pattern 3)
|
||||
|
||||
- `Token` — Old English *tacen* ("sign"); modern usage: atomic unit of text after tokenization.
|
||||
- `Logits` — Greek *λογιστικός* via Latin *logisticus*; modern ML usage: raw pre-softmax scores.
|
||||
- `Probability` — Latin *probabilitas* from *probare* ("to test"); modern usage: a value in [0, 1].
|
||||
- `Scalar` — Latin *scalaris* from *scala* ("ladder"); modern usage: a single quantity (vs. vector/matrix).
|
||||
- `Chinchilla` — Hoffmann et al. 2022, "Training Compute-Optimal Large Language Models"; the paper name.
|
||||
- `BPE` — Sennrich et al. 2016, "Neural Machine Translation of Rare Words with Subword Units"; the algorithm name.
|
||||
- `Token` - Old English *tacen* ("sign"); modern usage: atomic unit of text after tokenization.
|
||||
- `Logits` - Greek *logistike* via Latin *logisticus*; modern ML usage: raw pre-softmax scores.
|
||||
- `Probability` - Latin *probabilitas* from *probare* ("to test"); modern usage: a value in [0, 1].
|
||||
- `Scalar` - Latin *scalaris* from *scala* ("ladder"); modern usage: a single quantity (vs. vector/matrix).
|
||||
- `Chinchilla` - Hoffmann et al. 2022, "Training Compute-Optimal Large Language Models"; the paper name.
|
||||
- `BPE` - Sennrich et al. 2016, "Neural Machine Translation of Rare Words with Subword Units"; the algorithm name.
|
||||
+19
-19
@@ -1,16 +1,16 @@
|
||||
# cs229_building_llms — Pass 3 Notes
|
||||
# cs229_building_llms - Pass 3 Notes
|
||||
|
||||
**Track:** `video_analysis_deob_pass3_20260623`
|
||||
**Date:** 2026-06-23
|
||||
**Language:** C11 (per the per-language default in `TIER2_STARTER.md` §3)
|
||||
**Language:** C11 (per the per-language default in `TIER2_STARTER.md` section 3)
|
||||
|
||||
## Decisions made
|
||||
|
||||
1. **Language:** C11 (default; per `TIER2_STARTER.md` §3 cluster A row 1).
|
||||
1. **Language:** C11 (default; per `TIER2_STARTER.md` section 3 cluster A row 1).
|
||||
2. **Conventions:** duffle + forth bootslop + raddbg fallback (per `c11_convention.md`).
|
||||
3. **Header style:** design-doc header (per `c11_convention.md` §9).
|
||||
3. **Header style:** design-doc header (per `c11_convention.md` section 9).
|
||||
4. **Type system:** byte-width types (`U4`, `F32`) + `TSet_` / `Struct_` / `Opt_` / `Ret_` macros.
|
||||
5. **Encoding placeholders:** `float` / `integer` / `Scalar` (per v2 lexicon §7). Resolved as concrete C11 types at the function signature.
|
||||
5. **Encoding placeholders:** `float` / `integer` / `Scalar` (per v2 lexicon section 7). Resolved as concrete C11 types at the function signature.
|
||||
|
||||
## Alternatives considered
|
||||
|
||||
@@ -20,29 +20,29 @@
|
||||
|
||||
## Language override (none)
|
||||
|
||||
Per `TIER2_STARTER.md` §3, the default for this video is C11. No override applied.
|
||||
Per `TIER2_STARTER.md` section 3, the default for this video is C11. No override applied.
|
||||
|
||||
## 4 + 3 verification criteria (per v2 lexicon §7 of `TIER2_STARTER.md`)
|
||||
## 4 + 3 verification criteria (per v2 lexicon section 7 of `TIER2_STARTER.md`)
|
||||
|
||||
| # | Criterion | Status | Notes |
|
||||
|---|---|---|---|
|
||||
| 1 | **Lossless** | met | All 10 concepts from the translation table are represented in the C11 code. |
|
||||
| 2 | **Bounded** | met | No `∞_val`; all values are finite (Token, F32, U4, U64). |
|
||||
| 2 | **Bounded** | met | No `infinity_val`; all values are finite (Token, F32, U4, U64). |
|
||||
| 3 | **Constructively typed** | met | Every expression has a type (`Token`, `Logits`, `Probability`, `Scalar`, etc.). |
|
||||
| 4 | **Etymology-cited** | met | Every new term has 1-line origin + 1-line history in the decoder. |
|
||||
| 5 | **Encoding-explicit** | met | Every value-bearing term has an encoding (`Token : int32`, `F32` etc.). |
|
||||
| 6 | **Form-anchored** | met | Every re-encoding has a form anchor in the translation table. |
|
||||
| 7 | **User-specific opt-in** | met | The principled form is produced; the user-specific form (e.g., Sectored Language `'scalar product'`) is opt-in. |
|
||||
|
||||
## Hardware target (per v2 lexicon §7 of `TIER2_STARTER.md`)
|
||||
## Hardware target (per v2 lexicon section 7 of `TIER2_STARTER.md`)
|
||||
|
||||
Per user 2026-06-23, "target up to 10k." Default workstation: Ryzen 9 / i9, RTX 4090, 128GB DDR5, 4TB NVMe.
|
||||
|
||||
This video's concepts map to:
|
||||
- **Pretraining** (cs229 §2.3-2.4): requires RTX 4090-class GPU + 128GB DDR5 for a 7B-param model in float16. The back-of-envelope Llama 3 400B numbers (3.79e25 FLOPs) imply a much larger cluster (16k+ H100s).
|
||||
- **Inference** (cs229 §2.6): KV-cache for a 7B model is `2 * 1 * 2048 * 32 * 32 * 128 * 2 = 1.07 GB` per request (F32 here for clarity; F16 halves this to 0.54 GB). RTX 4090 has 24GB VRAM, so ~20 concurrent requests fit.
|
||||
- **Pretraining** (cs229 section 2.3-2.4): requires RTX 4090-class GPU + 128GB DDR5 for a 7B-param model in float16. The back-of-envelope Llama 3 400B numbers (3.79e25 FLOPs) imply a much larger cluster (16k+ H100s).
|
||||
- **Inference** (cs229 section 2.6): KV-cache for a 7B model is `2 * 1 * 2048 * 32 * 32 * 128 * 2 = 1.07 GB` per request (F32 here for clarity; F16 halves this to 0.54 GB). RTX 4090 has 24GB VRAM, so ~20 concurrent requests fit.
|
||||
|
||||
## Refinements discovered (Pass 3 → lexicon v3 candidates)
|
||||
## Refinements discovered (Pass 3 to lexicon v3 candidates)
|
||||
|
||||
1. **Per-pillar encoding tags:** the lecture's six pillars (Architecture, Training, Data, Evaluation, Systems, Model) could each have a tier tag in the lexicon (currently they are Tier 2 terms mixed). The C11 code shows that the pillars map to distinct type categories; a v3 lexicon could formalize this.
|
||||
|
||||
@@ -54,10 +54,10 @@ This video's concepts map to:
|
||||
|
||||
## See also
|
||||
|
||||
- `cs229_building_llms.c` — the C11 program
|
||||
- `cs229_building_llms_translation.md` — the math → C11 translation table
|
||||
- `cs229_building_llms_decoder.md` — the per-term decoder (tier-categorized)
|
||||
- `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/cs229_building_llms/` — the Pass 2 input
|
||||
- `conductor/tracks/video_analysis_cs229_building_llms_20260621/report.md` — the Pass 1 source
|
||||
- `conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md` — the C11 style guide
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` — the v2 lexicon
|
||||
- `cs229_building_llms.c` - the C11 program
|
||||
- `cs229_building_llms_translation.md` - the math to C11 translation table
|
||||
- `cs229_building_llms_decoder.md` - the per-term decoder (tier-categorized)
|
||||
- `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/cs229_building_llms/` - the Pass 2 input
|
||||
- `conductor/tracks/video_analysis_cs229_building_llms_20260621/report.md` - the Pass 1 source
|
||||
- `conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md` - the C11 style guide
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` - the v2 lexicon
|
||||
+4
-3
@@ -1,4 +1,4 @@
|
||||
# cs229_building_llms — Translation Table (math → C11)
|
||||
# cs229_building_llms - Translation Table (math to C11)
|
||||
|
||||
**Source:** `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/cs229_building_llms/cs229_building_llms_deobfuscated.md` (464 lines)
|
||||
**Target:** `cs229_building_llms.c` (the C11 program)
|
||||
@@ -17,7 +17,8 @@
|
||||
| 9 | `FLOPs = 6 * N * D` | `training_flops(F32 N_params, F32 D_tokens)` | bounded: N, D as finite inputs | `F32` |
|
||||
| 10 | `Memory_KV = 2 * B * S * L * H * D * bytes_per_element` | `kv_cache_bytes(U4 batch, U4 seq_len, U4 n_layers, U4 n_heads, U4 head_dim, U4 bytes_per_elem)` | bounded: B, S, L, H, D, bytes all finite | `U64` |
|
||||
|
||||
**Notes:**
|
||||
## Notes
|
||||
|
||||
- The C11 program does NOT implement a real transformer; it expresses the SHAPE of the lecture's six pillars.
|
||||
- All `float` / `integer` / `Scalar` placeholders resolve to concrete C11 types (`F32`, `U4`, etc.) at the function signature.
|
||||
- Per the v2 lexicon §9.1, `<<` / `>>` operators are rendered as `much_less` / `much_greater` / `weakly_coupled` if needed; this file does not use them.
|
||||
- Per the v2 lexicon section 9.1, `<<` / `>>` operators are rendered as `much_less` / `much_greater` / `weakly_coupled` if needed; this file does not use them.
|
||||
+3
-10
@@ -1,4 +1,4 @@
|
||||
"""entropy_epiplexity.py - Pass 3 projection of the "From Entropy to Epiplexity" lecture.
|
||||
"""entropy_epiplexity.py - Pass 3 projection of the From Entropy to Epiplexity lecture.
|
||||
|
||||
PURPOSE
|
||||
-------
|
||||
@@ -21,16 +21,9 @@ ENCODING (per lexicon v2 Rule 5)
|
||||
Bits : float (placeholder), resolved as float64 (units of entropy)
|
||||
Markov<X, Y, Z> : type-class predicate where X -> Y -> Z (R4 NEW v2)
|
||||
PolyTimeAdversary : type where runtime(A) : Polynomial(security_parameter) : int64 (R6 NEW v2)
|
||||
|
||||
SEE ALSO
|
||||
--------
|
||||
entropy_epiplexity_translation.md
|
||||
entropy_epiplexity_decoder.md
|
||||
entropy_epiplexity_notes.md
|
||||
lexicon.md (the v2 lexicon) + product-guidelines.md (manual_slop)
|
||||
"""
|
||||
|
||||
from dataclasses import dataclass, field
|
||||
from dataclasses import dataclass
|
||||
from typing import Callable, TypeAlias
|
||||
import math
|
||||
|
||||
@@ -162,4 +155,4 @@ def main() -> int:
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
raise SystemExit(main())
|
||||
+15
-15
@@ -1,11 +1,11 @@
|
||||
# entropy_epiplexity — Per-term Decoder (tier-categorized)
|
||||
# entropy_epiplexity - Per-term Decoder (tier-categorized)
|
||||
|
||||
## Tier 1: Core concepts
|
||||
|
||||
| Term | Python form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `Probability` | `TypeAlias = float` | Latin *probabilitas* | Tier 1 | Cluster 0 |
|
||||
| `Entropy` | `TypeAlias = float` | Greek *ἐντροπή* ("a turning toward"); Rudolf Clausius, 1865 | Tier 1 | Cluster 0 |
|
||||
| `Entropy` | `TypeAlias = float` | Greek *entropia* ("a turning toward"); Rudolf Clausius, 1865 | Tier 1 | Cluster 0 |
|
||||
|
||||
## Tier 2: Data-oriented pipeline terms
|
||||
|
||||
@@ -14,7 +14,7 @@
|
||||
| `shannon_entropy` | function | Claude Shannon, 1948; "A Mathematical Theory of Communication" | Tier 2 | Cluster 0 |
|
||||
| `cross_entropy` | function | the cross-entropy loss | Tier 2 | Cluster 2 |
|
||||
| `kl_divergence` | function | Solomon Kullback, Richard Leibler, 1951 | Tier 2 | Cluster 2 |
|
||||
| `epiplexity_estimate` | function | "epi-" (Greek "upon") + "plexity" (Latin "fold"); the "folding upon" measure of memorization | Tier 2 | entropy_epiplexity §2 (NEW v2) |
|
||||
| `epiplexity_estimate` | function | "epi-" (Greek "upon") + "plexity" (Latin "fold"); the "folding upon" measure of memorization | Tier 2 | entropy_epiplexity section 2 (NEW v2) |
|
||||
| `bits_to_nats`, `nats_to_bits` | function | units conversion; nat = natural logarithm unit | Tier 2 | Cluster 2 |
|
||||
|
||||
## Tier 3: Type-theoretic primitives
|
||||
@@ -22,23 +22,23 @@
|
||||
| Term | Python form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `MarkovState` | `@dataclass(frozen=True) class MarkovState` | Andrey Markov, 1856-1922; the eponym | Tier 3 | Cluster 3 |
|
||||
| `MarkovChain` | `@dataclass(frozen=True) class MarkovChain` | the Markov chain as a type-class predicate | Tier 3 | entropy_epiplexity §5.2 (R4 NEW v2) |
|
||||
| `PolyTimeAdversary` | `@dataclass(frozen=True) class PolyTimeAdversary` | the polynomial-time adversary security model | Tier 3 | entropy_epiplexity §5.8 (R6 NEW v2) |
|
||||
| `MarkovChain` | `@dataclass(frozen=True) class MarkovChain` | the Markov chain as a type-class predicate | Tier 3 | entropy_epiplexity section 5.2 (R4 NEW v2) |
|
||||
| `PolyTimeAdversary` | `@dataclass(frozen=True) class PolyTimeAdversary` | the polynomial-time adversary security model | Tier 3 | entropy_epiplexity section 5.8 (R6 NEW v2) |
|
||||
|
||||
## Tier 4: AI-fuzzing tolerance terms
|
||||
|
||||
| Term | Python form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `memorization` | local variable | the bounded form of "memorization" in NN training | Tier 4 | entropy_epiplexity §2 |
|
||||
| `decay` | local variable | the bounded form of "decay" (1 / (1 + t * eps)) | Tier 4 | entropy_epiplexity §2 |
|
||||
| `security_parameter` | `int` field | the security parameter k | Tier 4 | entropy_epiplexity §5.8 |
|
||||
| `memorization` | local variable | the bounded form of "memorization" in NN training | Tier 4 | entropy_epiplexity section 2 |
|
||||
| `decay` | local variable | the bounded form of "decay" (1 / (1 + t * eps)) | Tier 4 | entropy_epiplexity section 2 |
|
||||
| `security_parameter` | `int` field | the security parameter k | Tier 4 | entropy_epiplexity section 5.8 |
|
||||
|
||||
## Etymology notes (per Cluster 7, Pattern 3)
|
||||
|
||||
- `Entropy` — Greek *ἐντροπή* via German *Entropie* (Clausius, 1865); modern usage: a measure of uncertainty.
|
||||
- `Shannon` — Claude Shannon (1916-2001); the founder of information theory.
|
||||
- `Kullback-Leibler` — Solomon Kullback (1907-1994) + Richard Leibler (1914-2003); the eponym.
|
||||
- `Epiplexity` — coined by Wilson & Finzi (2020); the "folding upon" measure of memorization in NN training.
|
||||
- `Markov` — Andrey Markov (1856-1922); the eponym; the chain was introduced in 1906.
|
||||
- `Adversary` — Latin *adversarius* ("opponent"); the cryptographic security model.
|
||||
- `Polynomial` — Greek *πολύ* ("many") + *νόμος* ("rule"); the complexity class.
|
||||
- `Entropy` - Greek *entropia* via German *Entropie* (Clausius, 1865); modern usage: a measure of uncertainty.
|
||||
- `Shannon` - Claude Shannon (1916-2001); the founder of information theory.
|
||||
- `Kullback-Leibler` - Solomon Kullback (1907-1994) + Richard Leibler (1914-2003); the eponym.
|
||||
- `Epiplexity` - coined by Wilson and Finzi (2020); the "folding upon" measure of memorization in NN training.
|
||||
- `Markov` - Andrey Markov (1856-1922); the eponym; the chain was introduced in 1906.
|
||||
- `Adversary` - Latin *adversarius* ("opponent"); the cryptographic security model.
|
||||
- `Polynomial` - Greek *poly* ("many") + *nomos* ("rule"); the complexity class.
|
||||
+16
-16
@@ -1,16 +1,16 @@
|
||||
# entropy_epiplexity — Pass 3 Notes
|
||||
# entropy_epiplexity - Pass 3 Notes
|
||||
|
||||
**Track:** `video_analysis_deob_pass3_20260623`
|
||||
**Date:** 2026-06-23
|
||||
**Language:** Python (per the per-language default in `TIER2_STARTER.md` §3)
|
||||
**Language:** Python (per the per-language default in `TIER2_STARTER.md` section 3)
|
||||
|
||||
## Decisions made
|
||||
|
||||
1. **Language:** Python (default; per `TIER2_STARTER.md` §3 cluster A row 3).
|
||||
1. **Language:** Python (default; per `TIER2_STARTER.md` section 3 cluster A row 3).
|
||||
2. **Conventions:** manual_slop (1-space indent, type hints, no comments, Result[T] for errors).
|
||||
3. **Type system:** `dataclass(frozen=True)` for value semantics; `TypeAlias` for primitives.
|
||||
4. **Markov chain:** encoded as a `MarkovChain` dataclass with `is_markov()` validation (rows sum to 1).
|
||||
5. **Epiplexity:** simplified model — `max(0, H(p_data, q_estimate) - H(p_data))` with decay factor.
|
||||
5. **Epiplexity:** simplified model - `max(0, H(p_data, q_estimate) - H(p_data))` with decay factor.
|
||||
|
||||
## Alternatives considered
|
||||
|
||||
@@ -19,21 +19,21 @@
|
||||
|
||||
## Language override (none)
|
||||
|
||||
Per `TIER2_STARTER.md` §3, the default for this video is Python. No override applied.
|
||||
Per `TIER2_STARTER.md` section 3, the default for this video is Python. No override applied.
|
||||
|
||||
## 4 + 3 verification criteria (per v2 lexicon §7 of `TIER2_STARTER.md`)
|
||||
## 4 + 3 verification criteria (per v2 lexicon section 7 of `TIER2_STARTER.md`)
|
||||
|
||||
| # | Criterion | Status | Notes |
|
||||
|---|---|---|---|
|
||||
| 1 | **Lossless** | met | All 7 concepts from the translation table are represented. |
|
||||
| 2 | **Bounded** | met | No `∞_val`; all values are finite. |
|
||||
| 2 | **Bounded** | met | No `infinity_val`; all values are finite. |
|
||||
| 3 | **Constructively typed** | met | Every expression has a type hint. |
|
||||
| 4 | **Etymology-cited** | met | Every new term has 1-line origin + 1-line history. |
|
||||
| 5 | **Encoding-explicit** | met | Every value-bearing term has an encoding. |
|
||||
| 6 | **Form-anchored** | met | Every re-encoding has a form anchor in the translation table. |
|
||||
| 7 | **User-specific opt-in** | met | The principled form is produced. |
|
||||
|
||||
## Hardware target (per v2 lexicon §7 of `TIER2_STARTER.md`)
|
||||
## Hardware target (per v2 lexicon section 7 of `TIER2_STARTER.md`)
|
||||
|
||||
Per user 2026-06-23, "target up to 10k." Default workstation: Ryzen 9 / i9, RTX 4090, 128GB DDR5, 4TB NVMe.
|
||||
|
||||
@@ -42,7 +42,7 @@ This video's concepts map to:
|
||||
- **Epiplexity in NN training:** requires a GPU for the actual training loop; the epiplexity_estimate function is a simplified post-hoc model.
|
||||
- **Markov chain stationary distribution:** converges in O(n) iterations for an n-state chain; no special hardware needed.
|
||||
|
||||
## Refinements discovered (Pass 3 → lexicon v3 candidates)
|
||||
## Refinements discovered (Pass 3 to lexicon v3 candidates)
|
||||
|
||||
1. **Epiplexity as a Tier 4 term:** the epiplexity function is a NEW v2 term; v3 should formalize it.
|
||||
2. **Markov<X, Y, Z> as a type-class predicate:** the R4 NEW v2 entry could be extended to a generic type-class pattern for stochastic processes.
|
||||
@@ -55,10 +55,10 @@ This video's concepts map to:
|
||||
|
||||
## See also
|
||||
|
||||
- `entropy_epiplexity.py` — the Python program
|
||||
- `entropy_epiplexity_translation.md` — the math → Python translation table
|
||||
- `entropy_epiplexity_decoder.md` — the per-term decoder (tier-categorized)
|
||||
- `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/entropy_epiplexity/` — the Pass 2 input
|
||||
- `conductor/tracks/video_analysis_entropy_epiplexity_20260621/report.md` — the Pass 1 source
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` — the v2 lexicon
|
||||
- `conductor/product-guidelines.md` — the manual_slop convention
|
||||
- `entropy_epiplexity.py` - the Python program
|
||||
- `entropy_epiplexity_translation.md` - the math to Python translation table
|
||||
- `entropy_epiplexity_decoder.md` - the per-term decoder (tier-categorized)
|
||||
- `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/entropy_epiplexity/` - the Pass 2 input
|
||||
- `conductor/tracks/video_analysis_entropy_epiplexity_20260621/report.md` - the Pass 1 source
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` - the v2 lexicon
|
||||
- `conductor/product-guidelines.md` - the manual_slop convention
|
||||
+5
-4
@@ -1,4 +1,4 @@
|
||||
# entropy_epiplexity — Translation Table (math → Python)
|
||||
# entropy_epiplexity - Translation Table (math to Python)
|
||||
|
||||
**Source:** `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/entropy_epiplexity/entropy_epiplexity_deobfuscated.md`
|
||||
**Target:** `entropy_epiplexity.py`
|
||||
@@ -14,7 +14,8 @@
|
||||
| 6 | `PolyTimeAdversary : Type where runtime(A) : Polynomial(security_parameter) : int64` (R6 NEW v2) | `PolyTimeAdversary` dataclass with `is_poly_time` | bounded: `runtime(n) <= n^k` | `PolyTimeAdversary : type` |
|
||||
| 7 | `nats -> bits` (unit conversion) | `bits_to_nats`, `nats_to_bits` | bounded: linear conversion | `Entropy : float` |
|
||||
|
||||
**Notes:**
|
||||
- Per v2 lexicon §9.2, the per-language rendering is the same as C11.
|
||||
## Notes
|
||||
|
||||
- Per v2 lexicon section 9.2, the per-language rendering is the same as C11.
|
||||
- The `epiplexity_estimate` function is a SIMPLIFIED model; the real epiplexity function is more complex.
|
||||
- The Markov chain is checked for stochasticity (rows sum to 1) via `is_markov()`.
|
||||
- The Markov chain is checked for stochasticity (rows sum to 1) via `is_markov()`.
|
||||
+7
-11
@@ -1,4 +1,4 @@
|
||||
"""probability_logic.py - Pass 3 projection of the "Probability Theory is an Extension of Logic" lecture.
|
||||
"""probability_logic.py - Pass 3 projection of the Probability Theory is an Extension of Logic lecture.
|
||||
|
||||
PURPOSE
|
||||
-------
|
||||
@@ -24,16 +24,13 @@ ENCODING (per lexicon v2 Rule 5)
|
||||
|
||||
SEE ALSO
|
||||
--------
|
||||
probability_logic_translation.md : math-to-Python translation table
|
||||
probability_logic_decoder.md : per-term decoder (tier-categorized)
|
||||
probability_logic_notes.md : decisions, alternatives, overrides
|
||||
lexicon.md (the v2 lexicon) + product-guidelines.md (manual_slop)
|
||||
probability_logic_translation.md
|
||||
probability_logic_decoder.md
|
||||
probability_logic_notes.md
|
||||
"""
|
||||
|
||||
from dataclasses import dataclass, field
|
||||
from typing import Callable, Iterable, TypeAlias
|
||||
from itertools import product as cartesian_product
|
||||
import math
|
||||
from dataclasses import dataclass
|
||||
from typing import Callable, TypeAlias
|
||||
|
||||
Plausibility: TypeAlias = float
|
||||
Probability: TypeAlias = float
|
||||
@@ -129,7 +126,6 @@ def jaynes_policeman_burglar(p_burglary: Plausibility = 0.001,
|
||||
+ p_earthquake * p_alarm_given_earthquake
|
||||
+ (1.0 - p_burglary - p_earthquake) * p_alarm_given_neither
|
||||
)
|
||||
p_burglary_and_alarm: float = p_burglary * p_alarm_given_burglary
|
||||
return bayes_rule(p_burglary, p_alarm_given_burglary, p_alarm) if p_alarm > 0 else 0.0
|
||||
|
||||
|
||||
@@ -178,4 +174,4 @@ def main() -> int:
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
raise SystemExit(main())
|
||||
+22
-22
@@ -1,25 +1,25 @@
|
||||
# probability_logic — Per-term Decoder (tier-categorized)
|
||||
# probability_logic - Per-term Decoder (tier-categorized)
|
||||
|
||||
**Source:** `probability_logic_deobfuscated.md` (Pass 2 deobfuscation)
|
||||
**Target:** `probability_logic.py` (the Python program)
|
||||
**Method:** Per v2 lexicon §1.3 (etymology) + §2 (the 4 tiers)
|
||||
**Method:** Per v2 lexicon section 1.3 (etymology) + section 2 (the 4 tiers)
|
||||
|
||||
## Tier 1: Core concepts
|
||||
|
||||
| Term | Python form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `Proposition` | `@dataclass(frozen=True) class Proposition` | Latin *propositio* ("a setting forth"); the propositional logic primitive | Tier 1 | Cluster 7, Tier 1 #1.4-1.7 |
|
||||
| `Context` | `@dataclass(frozen=True) class Context` | Latin *contextus* ("a joining together"); the conditioning context | Tier 1 | Tier 1 #1.4-1.7 |
|
||||
| `implies` | `Callable[[Proposition, Proposition], bool]` | Latin *implicare* ("to involve"); Tier 1 #1.7 | Tier 1 | Cluster 7 |
|
||||
| `for all` | (implicit in `assert` + exhaustive loops) | Latin *pro omnibus*; Tier 1 #1.2 | Tier 1 | Cluster 2, 4 |
|
||||
| `exists` | (implicit in `for ... if not found: raise`) | Latin *existere* ("to stand out, to be"); Tier 1 #1.3 | Tier 1 | Cluster 4 |
|
||||
| `Proposition` | `@dataclass(frozen=True) class Proposition` | Latin *propositio* ("a setting forth"); the propositional logic primitive | Tier 1 | Cluster 7 |
|
||||
| `Context` | `@dataclass(frozen=True) class Context` | Latin *contextus* ("a joining together"); the conditioning context | Tier 1 | Cluster 7 |
|
||||
| `implies` | `Callable[[Proposition, Proposition], bool]` | Latin *implicare* ("to involve"); Tier 1 entry 1.7 | Tier 1 | Cluster 7 |
|
||||
| `for all` | (implicit in `assert` + exhaustive loops) | Latin *pro omnibus*; Tier 1 entry 1.2 | Tier 1 | Cluster 2, 4 |
|
||||
| `exists` | (implicit in `for ... if not found: raise`) | Latin *existere* ("to stand out, to be"); Tier 1 entry 1.3 | Tier 1 | Cluster 4 |
|
||||
|
||||
## Tier 2: Data-oriented pipeline terms
|
||||
|
||||
| Term | Python form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `bivaluation` | function | from Cox's bivaluation; generalized indicator function | Tier 2 | Tier 4 (R3 NEW v2) |
|
||||
| `frequentist_relative_frequency` | function | frequentist definition; the ratio `count/total` | Tier 2 | Cluster 2 |
|
||||
| `bivaluation` | function | from Cox's bivaluation; generalized indicator function | Tier 2 | Cluster 0 |
|
||||
| `frequentist_relative_frequency` | function | frequentist definition; the ratio count/total | Tier 2 | Cluster 2 |
|
||||
| `frequentist_stream` | function | re-encoding of `lim_{N -> infinity}` as `Stream A = nat -> A` | Tier 2 | Rule 1 + Cluster 2 |
|
||||
| `sum_rule` | function | the sum rule of probability | Tier 2 | Cluster 2 |
|
||||
| `product_rule` | function | the product rule of probability | Tier 2 | Cluster 2 |
|
||||
@@ -31,27 +31,27 @@
|
||||
|
||||
| Term | Python form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `LatticePoset` | `@dataclass(frozen=True) class LatticePoset` | Latin *lattice* + *poset*; a partially ordered set with join/meet | Tier 3 | Cluster 3 (Pair<A, B>) |
|
||||
| `LatticePoset` | `@dataclass(frozen=True) class LatticePoset` | Latin *lattice* + *poset*; a partially ordered set with join/meet | Tier 3 | Cluster 3 |
|
||||
| `Plausibility` | `TypeAlias = float` | Latin *plausibilis* ("deserving applause"); the bounded form of "plausibility" | Tier 3 | Cluster 0 |
|
||||
| `Probability` | `TypeAlias = float` | Latin *probabilitas* ("likelihood"); bounded to [0, 1] | Tier 3 | Cluster 0 |
|
||||
| `Stream` | `TypeAlias = "Callable[[int], float]"` | Old English *stream*; `Stream A = nat -> A` per Rule 1 | Tier 3 | Cluster 3 (Pi type) |
|
||||
| `Stream` | `TypeAlias = "Callable[[int], float]"` | Old English *stream*; `Stream A = nat -> A` per Rule 1 | Tier 3 | Cluster 3 |
|
||||
|
||||
## Tier 4: AI-fuzzing tolerance terms
|
||||
|
||||
| Term | Python form | Etymology | Tier | Source |
|
||||
|---|---|---|---|---|
|
||||
| `world` | `dict[Proposition, bool]` | the world assignment; bounded to finite propositions | Tier 4 | Cluster 0 (P2) |
|
||||
| `bivaluation` | function (R3 NEW v2) | the generalized indicator; bivalent (0/1) for propositions | Tier 4 | probability_logic §2.1 (R3 NEW v2) |
|
||||
| `Plausibility` (was `Likelihood`) | TypeAlias | the bounded form of "plausibility"; was `Likelihood` in v1 lexicon | Tier 4 | probability_logic §2.1 (R3 NEW v2) |
|
||||
| `bivaluation` (R3 NEW v2) | function | the generalized indicator; bivalent (0/1) for propositions | Tier 4 | probability_logic section 2.1 |
|
||||
| `Plausibility` (was `Likelihood`) | TypeAlias | the bounded form of "plausibility"; was `Likelihood` in v1 lexicon | Tier 4 | probability_logic section 2.1 |
|
||||
|
||||
## Etymology notes (per Cluster 7, Pattern 3)
|
||||
|
||||
- `Proposition` — Latin *propositio* from *proponere* ("to set forth"); modern usage: an assertion that is true or false.
|
||||
- `Context` — Latin *contextus* from *contexere* ("to weave together"); modern usage: the conditioning context for a probability.
|
||||
- `Lattice` — Old French *latiz* ("lattice"); modern usage: a partially ordered set with join and meet.
|
||||
- `Poset` — partially ordered set; abbreviation; modern usage: a set with a reflexive, antisymmetric, transitive relation.
|
||||
- `Bivaluation` — Latin *bi-* ("two") + Old French *valour* ("value"); the assignment of two values (true/false) to propositions; generalized in Cox's theorem to continuous plausibility.
|
||||
- `Marginalize` — Latin *marginalis* ("of the margin"); modern usage: summing out a variable from a joint distribution.
|
||||
- `Bayes` — Thomas Bayes (1701-1761); the eponym; the rule was published posthumously in 1763.
|
||||
- `Jaynes` — Edwin T. Jaynes (1922-1998); the probability-as-logic school; the canonical "policeman+burglar" example.
|
||||
- `Cox` — Richard T. Cox (1898-1991); the Cox theorem deriving the sum and product rules from Boolean algebra.
|
||||
- `Proposition` - Latin *propositio* from *proponere* ("to set forth"); modern usage: an assertion that is true or false.
|
||||
- `Context` - Latin *contextus* from *contexere* ("to weave together"); modern usage: the conditioning context for a probability.
|
||||
- `Lattice` - Old French *latiz* ("lattice"); modern usage: a partially ordered set with join and meet.
|
||||
- `Poset` - partially ordered set; abbreviation; modern usage: a set with a reflexive, antisymmetric, transitive relation.
|
||||
- `Bivaluation` - Latin *bi-* ("two") + Old French *valour* ("value"); the assignment of two values (true/false) to propositions; generalized in Cox's theorem to continuous plausibility.
|
||||
- `Marginalize` - Latin *marginalis* ("of the margin"); modern usage: summing out a variable from a joint distribution.
|
||||
- `Bayes` - Thomas Bayes (1701-1761); the eponym; the rule was published posthumously in 1763.
|
||||
- `Jaynes` - Edwin T. Jaynes (1922-1998); the probability-as-logic school; the canonical "policeman+burglar" example.
|
||||
- `Cox` - Richard T. Cox (1898-1991); the Cox theorem deriving the sum and product rules from Boolean algebra.
|
||||
+15
-15
@@ -1,12 +1,12 @@
|
||||
# probability_logic — Pass 3 Notes
|
||||
# probability_logic - Pass 3 Notes
|
||||
|
||||
**Track:** `video_analysis_deob_pass3_20260623`
|
||||
**Date:** 2026-06-23
|
||||
**Language:** Python (per the per-language default in `TIER2_STARTER.md` §3)
|
||||
**Language:** Python (per the per-language default in `TIER2_STARTER.md` section 3)
|
||||
|
||||
## Decisions made
|
||||
|
||||
1. **Language:** Python (default; per `TIER2_STARTER.md` §3 cluster A row 2).
|
||||
1. **Language:** Python (default; per `TIER2_STARTER.md` section 3 cluster A row 2).
|
||||
2. **Conventions:** manual_slop (1-space indent, type hints, no comments, Result[T] for errors).
|
||||
3. **Type system:** `dataclass(frozen=True)` for value semantics; `TypeAlias` for primitives.
|
||||
4. **Stream encoding:** `Stream A = nat -> A` per v2 lexicon Rule 1; rendered as `Callable[[int], float]`.
|
||||
@@ -19,21 +19,21 @@
|
||||
|
||||
## Language override (none)
|
||||
|
||||
Per `TIER2_STARTER.md` §3, the default for this video is Python. No override applied.
|
||||
Per `TIER2_STARTER.md` section 3, the default for this video is Python. No override applied.
|
||||
|
||||
## 4 + 3 verification criteria (per v2 lexicon §7 of `TIER2_STARTER.md`)
|
||||
## 4 + 3 verification criteria (per v2 lexicon section 7 of `TIER2_STARTER.md`)
|
||||
|
||||
| # | Criterion | Status | Notes |
|
||||
|---|---|---|---|
|
||||
| 1 | **Lossless** | met | All 10 concepts from the translation table are represented in the Python code. |
|
||||
| 2 | **Bounded** | met | No `∞_val`; the frequentist definition is re-encoded as a `Stream`. |
|
||||
| 2 | **Bounded** | met | No `infinity_val`; the frequentist definition is re-encoded as a `Stream`. |
|
||||
| 3 | **Constructively typed** | met | Every expression has a type hint. |
|
||||
| 4 | **Etymology-cited** | met | Every new term has 1-line origin + 1-line history in the decoder. |
|
||||
| 5 | **Encoding-explicit** | met | Every value-bearing term has an encoding (`Plausibility : float`, `Stream : Callable[[int], float]`). |
|
||||
| 6 | **Form-anchored** | met | Every re-encoding has a form anchor in the translation table. |
|
||||
| 7 | **User-specific opt-in** | met | The principled form is produced; the user-specific form (e.g., Cox's bivaluation) is opt-in. |
|
||||
|
||||
## Hardware target (per v2 lexicon §7 of `TIER2_STARTER.md`)
|
||||
## Hardware target (per v2 lexicon section 7 of `TIER2_STARTER.md`)
|
||||
|
||||
Per user 2026-06-23, "target up to 10k." Default workstation: Ryzen 9 / i9, RTX 4090, 128GB DDR5, 4TB NVMe.
|
||||
|
||||
@@ -42,7 +42,7 @@ This video's concepts map to:
|
||||
- **Bayesian inference:** marginalization scales with the size of the joint distribution; for 1000 propositions, the lattice is computable in <1s on any modern CPU.
|
||||
- **Stream re-encoding:** the `Stream A = nat -> A` is computable up to a given index; the lecture's "infinity" is a process, not a value.
|
||||
|
||||
## Refinements discovered (Pass 3 → lexicon v3 candidates)
|
||||
## Refinements discovered (Pass 3 to lexicon v3 candidates)
|
||||
|
||||
1. **Bivaluation as Tier 4 term:** the bivaluation `Z(x, t)` is a Tier 4 (AI-fuzzing tolerance) term that doesn't have an existing entry in the v2 lexicon. v3 should add it.
|
||||
2. **Cox's theorem formalization:** the Python program implements the bivaluation but not the full Cox theorem. v3 could formalize the sum/product rule derivation.
|
||||
@@ -55,10 +55,10 @@ This video's concepts map to:
|
||||
|
||||
## See also
|
||||
|
||||
- `probability_logic.py` — the Python program
|
||||
- `probability_logic_translation.md` — the math → Python translation table
|
||||
- `probability_logic_decoder.md` — the per-term decoder (tier-categorized)
|
||||
- `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/probability_logic/` — the Pass 2 input
|
||||
- `conductor/tracks/video_analysis_probability_logic_20260621/report.md` — the Pass 1 source
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` — the v2 lexicon
|
||||
- `conductor/product-guidelines.md` — the manual_slop convention
|
||||
- `probability_logic.py` - the Python program
|
||||
- `probability_logic_translation.md` - the math to Python translation table
|
||||
- `probability_logic_decoder.md` - the per-term decoder (tier-categorized)
|
||||
- `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/probability_logic/` - the Pass 2 input
|
||||
- `conductor/tracks/video_analysis_probability_logic_20260621/report.md` - the Pass 1 source
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` - the v2 lexicon
|
||||
- `conductor/product-guidelines.md` - the manual_slop convention
|
||||
+5
-4
@@ -1,4 +1,4 @@
|
||||
# probability_logic — Translation Table (math → Python)
|
||||
# probability_logic - Translation Table (math to Python)
|
||||
|
||||
**Source:** `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/probability_logic/probability_logic_deobfuscated.md` (538 lines)
|
||||
**Target:** `probability_logic.py` (the Python program)
|
||||
@@ -17,8 +17,9 @@
|
||||
| 9 | `sum_B P(H, B)` (marginalization) | `marginalize(joint, hypothesis, observations) -> float` | bounded: finite `observations` list | `Plausibility : float` |
|
||||
| 10 | Jaynes policeman+burglar | `jaynes_policeman_burglar()` | bounded: numerical default values | `Plausibility : float` |
|
||||
|
||||
**Notes:**
|
||||
## Notes
|
||||
|
||||
- The Python program does NOT implement full Jaynes' probability theory; it expresses the SHAPE of the lecture's three parts.
|
||||
- All `float` placeholders resolve to `float64` at runtime (Python's default `float` is C double).
|
||||
- Per the v2 lexicon §9.2, the per-language rendering is the same as C11 (`much_less` / `much_greater` / `weakly_coupled`); this file does not use them.
|
||||
- The frequentist definition is re-encoded as a `Stream : nat -> float` (the bounded form of "infinity" per Rule 1).
|
||||
- Per the v2 lexicon section 9.2, the per-language rendering is the same as C11 (`much_less` / `much_greater` / `weakly_coupled`); this file does not use them.
|
||||
- The frequentist definition is re-encoded as a `Stream : nat -> float` (the bounded form of "infinity" per Rule 1).
|
||||
Reference in New Issue
Block a user