Add V7R tech crypto classifier

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+text_classifier.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,4 +1,62 @@
 ---
 license: mit
+tags:
+- text-classification
+- crypto
+- technology
+- twitter
+- x
+- fasttext-distillation
 ---
-this is a classification model that sorts tweets/profiles off the probability that it is tech/crypto related. this was a model created for a job that fell short. this is a tf-idf model, distilled from a transformer model that I also made. maybe ill upload that soon
+this is a classification model that sorts tweets/profiles off the probability that it is tech/crypto related. this was a model created for a job that fell short. this is a tf-idf model, distilled from a transformer model that I also made. maybe ill upload that soon
+
+# Techpto Classifier
+
+This repository contains a lightweight production classifier for detecting whether X/Twitter posts and profiles are crypto-related, tech-related, both, or neither.
+
+## Files
+
+- `text_classifier.json`: Rust-compatible hashed logistic-regression classifier.
+- `model_config.json`: labels, expected inputs, and recommended thresholds.
+- `distill_metrics.json`: proxy evaluation metrics from distillation.
+- `recommended_thresholds_distillation.json`: thresholds tuned against the V7 fastText teacher.
+- `full_run_manifest.json`: counts and thresholds from the large full-corpus run.
+
+## Recommended Thresholds
+
+The high-precision full-corpus run used:
+
+```json
+{
+  "post_crypto": 0.85,
+  "post_tech": 0.90,
+  "profile_crypto": 0.90,
+  "profile_tech": 0.99
+}
+```
+
+The original distillation-tuned thresholds were:
+
+```json
+{
+  "post_crypto": 0.58,
+  "post_tech": 0.44,
+  "profile_crypto": 0.34,
+  "profile_tech": 0.38
+}
+```
+
+## Full-Corpus Run
+
+Using the high-precision thresholds:
+
+- Posts scanned: `928,484,069`
+- Post matches: `7,728,133`
+- Profiles scanned: `2,667,815,773`
+- Profile matches: `7,915,096`
+
+One corrupt post shard was skipped and is listed in `full_run_manifest.json`.
+
+## Notes
+
+This is not a standard Transformers checkpoint. It is a compact hashed-feature linear classifier intended for very high-throughput local scanning. Metrics in `distill_metrics.json` are proxy metrics against teacher/weak labels rather than a final human-labeled benchmark.

distill_metrics.json

@@ -0,0 +1,390 @@
+{
+  "note": "Rust-compatible hashed-logreg student distilled from V7 fastText. Metrics are proxy-only unless compared to human labels.",
+  "args": {
+    "data": "classified_v4_transformer_data_hardneg",
+    "teacher": "classified_v7_fasttext_v5b_120k\\model.bin",
+    "init_model": "classified_accuracy_v2\\text_classifier.json",
+    "output": "classified_v7r_fasttext_student_410k",
+    "train_sample": 0,
+    "eval_sample": 0,
+    "epochs": 3,
+    "learning_rate": 0.04,
+    "seed": 777,
+    "teacher_batch_size": 8192,
+    "crypto_threshold": 0.45,
+    "tech_threshold": 0.36
+  },
+  "teacher": {
+    "model": "classified_v7_fasttext_v5b_120k\\model.bin",
+    "thresholds": {
+      "crypto": 0.45,
+      "tech": 0.36
+    },
+    "train_seconds": 88.16688129999966,
+    "eval_seconds": 10.310907999999472
+  },
+  "student": {
+    "model_version": "hashed-logreg-v7-fasttext-student",
+    "train_rows": 410292,
+    "eval_rows": 50413,
+    "train_seconds": 13.258047600000282,
+    "default_thresholds": {
+      "post_crypto": 0.5,
+      "post_tech": 0.5,
+      "profile_crypto": 0.5,
+      "profile_tech": 0.5
+    },
+    "tuned_thresholds": {
+      "post_crypto": 0.58,
+      "post_tech": 0.44,
+      "profile_crypto": 0.34,
+      "profile_tech": 0.38
+    },
+    "threshold_report": {
+      "post_crypto": {
+        "threshold": 0.58,
+        "precision": 0.9671687073355727,
+        "recall": 0.980232813529541,
+        "fscore": 0.9736569402781566
+      },
+      "post_tech": {
+        "threshold": 0.44,
+        "precision": 0.9377788213195586,
+        "recall": 0.9484682973165519,
+        "fscore": 0.9430932703659975
+      },
+      "profile_crypto": {
+        "threshold": 0.34,
+        "precision": 0.9304776241959765,
+        "recall": 0.711490163248221,
+        "fscore": 0.806380833778094
+      },
+      "profile_tech": {
+        "threshold": 0.38,
+        "precision": 0.940460947503201,
+        "recall": 0.8883289659342875,
+        "fscore": 0.9136519125116617
+      }
+    }
+  },
+  "default_student_vs_teacher": {
+    "overall": {
+      "exact_match_accuracy": 0.9002836569932359,
+      "micro_precision": 0.9620242848993562,
+      "micro_recall": 0.8482431558525544,
+      "micro_f1": 0.9015579654803727,
+      "macro_precision": 0.9595827241171908,
+      "macro_recall": 0.8617410508785691,
+      "macro_f1": 0.9073573249366726,
+      "crypto": {
+        "precision": 0.9679515487981831,
+        "recall": 0.8221519665630693,
+        "f1": 0.8891142468055515,
+        "support": 18662
+      },
+      "tech": {
+        "precision": 0.9512138994361984,
+        "recall": 0.9013301351940689,
+        "f1": 0.9256004030677938,
+        "support": 9172
+      },
+      "pred_counts": {
+        "crypto": 15851,
+        "tech": 8691,
+        "negative": 26168
+      }
+    },
+    "posts_only": {
+      "exact_match_accuracy": 0.9610693606012073,
+      "micro_precision": 0.9562995333678986,
+      "micro_recall": 0.969512652999925,
+      "micro_f1": 0.9628607651577299,
+      "macro_precision": 0.9522663599218083,
+      "macro_recall": 0.9618777142458361,
+      "macro_f1": 0.9570222958759071,
+      "crypto": {
+        "precision": 0.9629789065863108,
+        "recall": 0.9826488029870415,
+        "f1": 0.9727144254810306,
+        "support": 9106
+      },
+      "tech": {
+        "precision": 0.9415538132573058,
+        "recall": 0.9411066255046308,
+        "f1": 0.9413301662707838,
+        "support": 4211
+      },
+      "pred_counts": {
+        "crypto": 9292,
+        "tech": 4209,
+        "negative": 11015
+      }
+    },
+    "profiles_only": {
+      "exact_match_accuracy": 0.8434886040979204,
+      "micro_precision": 0.9690245448781813,
+      "micro_recall": 0.7369980023420817,
+      "micro_f1": 0.8372329603255341,
+      "macro_precision": 0.9676408876174856,
+      "macro_recall": 0.7683900413176734,
+      "macro_f1": 0.852622602377825,
+      "crypto": {
+        "precision": 0.9749961884433602,
+        "recall": 0.6692130598576811,
+        "f1": 0.7936704933291964,
+        "support": 9556
+      },
+      "tech": {
+        "precision": 0.9602855867916109,
+        "recall": 0.8675670227776658,
+        "f1": 0.9115747114264534,
+        "support": 4961
+      },
+      "pred_counts": {
+        "crypto": 6559,
+        "tech": 4482,
+        "negative": 15153
+      }
+    }
+  },
+  "tuned_student_vs_teacher": {
+    "overall": {
+      "exact_match_accuracy": 0.9031003907722215,
+      "micro_precision": 0.9468231231113379,
+      "micro_recall": 0.866781633972839,
+      "micro_f1": 0.905036106161493,
+      "macro_precision": 0.9450696963354239,
+      "macro_recall": 0.8792805932309817,
+      "macro_f1": 0.910466457987787,
+      "crypto": {
+        "precision": 0.9509554910498307,
+        "recall": 0.8426213696281213,
+        "f1": 0.8935166770839252,
+        "support": 18662
+      },
+      "tech": {
+        "precision": 0.9391839016210173,
+        "recall": 0.9159398168338422,
+        "f1": 0.9274162388916487,
+        "support": 9172
+      },
+      "pred_counts": {
+        "crypto": 16536,
+        "tech": 8945,
+        "negative": 25288
+      }
+    },
+    "posts_only": {
+      "exact_match_accuracy": 0.9622602767853476,
+      "micro_precision": 0.9578884934756821,
+      "micro_recall": 0.9701884808890892,
+      "micro_f1": 0.9639992538705465,
+      "macro_precision": 0.9524737643275656,
+      "macro_recall": 0.9643505554230465,
+      "macro_f1": 0.958375105322077,
+      "crypto": {
+        "precision": 0.9671687073355727,
+        "recall": 0.980232813529541,
+        "f1": 0.9736569402781565,
+        "support": 9106
+      },
+      "tech": {
+        "precision": 0.9377788213195586,
+        "recall": 0.9484682973165519,
+        "f1": 0.9430932703659977,
+        "support": 4211
+      },
+      "pred_counts": {
+        "crypto": 9229,
+        "tech": 4259,
+        "negative": 11022
+      }
+    },
+    "profiles_only": {
+      "exact_match_accuracy": 0.8478244186938838,
+      "micro_precision": 0.9343783873926457,
+      "micro_recall": 0.7719225735344768,
+      "micro_f1": 0.8454168238400603,
+      "macro_precision": 0.9354692858495888,
+      "macro_recall": 0.7999095645912542,
+      "macro_f1": 0.8600163731448778,
+      "crypto": {
+        "precision": 0.9304776241959765,
+        "recall": 0.711490163248221,
+        "f1": 0.806380833778094,
+        "support": 9556
+      },
+      "tech": {
+        "precision": 0.940460947503201,
+        "recall": 0.8883289659342875,
+        "f1": 0.9136519125116617,
+        "support": 4961
+      },
+      "pred_counts": {
+        "crypto": 7307,
+        "tech": 4686,
+        "negative": 14266
+      }
+    }
+  },
+  "tuned_student_vs_weak_labels": {
+    "overall": {
+      "exact_match_accuracy": 0.8832840735524567,
+      "micro_precision": 0.9655429535732507,
+      "micro_recall": 0.8161823248407644,
+      "micro_f1": 0.8846022471910112,
+      "macro_precision": 0.965574214755928,
+      "macro_recall": 0.811794173017184,
+      "macro_f1": 0.8819491541016604,
+      "crypto": {
+        "precision": 0.9654692791485244,
+        "recall": 0.8274593137763036,
+        "f1": 0.8911526653642199,
+        "support": 19294
+      },
+      "tech": {
+        "precision": 0.9656791503633315,
+        "recall": 0.7961290322580645,
+        "f1": 0.8727456428391008,
+        "support": 10850
+      },
+      "pred_counts": {
+        "crypto": 16536,
+        "tech": 8945,
+        "negative": 25288
+      }
+    },
+    "posts_only": {
+      "exact_match_accuracy": 0.9659151574883988,
+      "micro_precision": 0.9944395017793595,
+      "micro_recall": 0.9453097469871027,
+      "micro_f1": 0.9692524478809119,
+      "macro_precision": 0.9942297434119423,
+      "macro_recall": 0.936574113553416,
+      "macro_f1": 0.9643727674837768,
+      "crypto": {
+        "precision": 0.994799003142269,
+        "recall": 0.9618648507071765,
+        "f1": 0.9780547565782465,
+        "support": 9545
+      },
+      "tech": {
+        "precision": 0.9936604836816154,
+        "recall": 0.9112833763996555,
+        "f1": 0.9506907783893069,
+        "support": 4644
+      },
+      "pred_counts": {
+        "crypto": 9229,
+        "tech": 4259,
+        "negative": 11022
+      }
+    },
+    "profiles_only": {
+      "exact_match_accuracy": 0.8060778144424833,
+      "micro_precision": 0.9330442758275661,
+      "micro_recall": 0.7013475399561266,
+      "micro_f1": 0.8007728638900816,
+      "macro_precision": 0.9343361720100598,
+      "macro_recall": 0.7029121738755889,
+      "macro_f1": 0.8022656696241247,
+      "crypto": {
+        "precision": 0.928424798138771,
+        "recall": 0.6958662426915581,
+        "f1": 0.7954971857410882,
+        "support": 9749
+      },
+      "tech": {
+        "precision": 0.9402475458813487,
+        "recall": 0.7099581050596198,
+        "f1": 0.8090341535071612,
+        "support": 6206
+      },
+      "pred_counts": {
+        "crypto": 7307,
+        "tech": 4686,
+        "negative": 14266
+      }
+    }
+  },
+  "teacher_vs_weak_labels": {
+    "overall": {
+      "exact_match_accuracy": 0.9257929502310912,
+      "micro_precision": 0.9578932241143925,
+      "micro_recall": 0.8844877919320594,
+      "micro_f1": 0.919728172755183,
+      "macro_precision": 0.9586196156540225,
+      "macro_recall": 0.8686619894248541,
+      "macro_f1": 0.9103972509783973,
+      "crypto": {
+        "precision": 0.9564891222805701,
+        "recall": 0.9251580802321966,
+        "f1": 0.9405627568763831,
+        "support": 19294
+      },
+      "tech": {
+        "precision": 0.9607501090274749,
+        "recall": 0.8121658986175115,
+        "f1": 0.8802317450804116,
+        "support": 10850
+      },
+      "pred_counts": {
+        "crypto": 18662,
+        "tech": 9172,
+        "negative": 22927
+      }
+    },
+    "posts_only": {
+      "exact_match_accuracy": 0.945423185906123,
+      "micro_precision": 0.9804009912142374,
+      "micro_recall": 0.9201494115159631,
+      "micro_f1": 0.9493201483312732,
+      "macro_precision": 0.9776264881654799,
+      "macro_recall": 0.9097467163339348,
+      "macro_f1": 0.9423142636561663,
+      "crypto": {
+        "precision": 0.9851746101471557,
+        "recall": 0.9398638030382399,
+        "f1": 0.9619859524958447,
+        "support": 9545
+      },
+      "tech": {
+        "precision": 0.9700783661838043,
+        "recall": 0.8796296296296297,
+        "f1": 0.9226425748164878,
+        "support": 4644
+      },
+      "pred_counts": {
+        "crypto": 9106,
+        "tech": 4211,
+        "negative": 11202
+      }
+    },
+    "profiles_only": {
+      "exact_match_accuracy": 0.9074514618985496,
+      "micro_precision": 0.9372459874629745,
+      "micro_recall": 0.8527734252585396,
+      "micro_f1": 0.8930165397742189,
+      "macro_precision": 0.9409932741182816,
+      "macro_recall": 0.8362211604736839,
+      "macro_f1": 0.8832334569237945,
+      "crypto": {
+        "precision": 0.9291544579321892,
+        "recall": 0.9107600779567135,
+        "f1": 0.9198653198653198,
+        "support": 9749
+      },
+      "tech": {
+        "precision": 0.9528320903043741,
+        "recall": 0.7616822429906542,
+        "f1": 0.8466015939822692,
+        "support": 6206
+      },
+      "pred_counts": {
+        "crypto": 9556,
+        "tech": 4961,
+        "negative": 11725
+      }
+    }
+  }
+}

full_run_manifest.json

@@ -0,0 +1,24 @@
+{
+    "mode":  "model_parallel_v7r_precision",
+    "classifier_version":  "hashed-logreg-v7-fasttext-student",
+    "created_at":  "2026-06-21T22:24:45",
+    "post_rows_scanned":  928484069,
+    "post_rows_written":  7728133,
+    "post_crypto_matches":  5725112,
+    "post_tech_matches":  2032807,
+    "profile_rows_scanned":  2667815773,
+    "profile_rows_written":  7915096,
+    "profile_crypto_matches":  5221555,
+    "profile_tech_matches":  2756026,
+    "author_stats_written":  1631493,
+    "thresholds":  {
+                       "post_crypto":  0.85,
+                       "post_tech":  0.9,
+                       "profile_crypto":  0.9,
+                       "profile_tech":  0.99
+                   },
+    "notes":  "Parallel run reused current output through part_4, skipped duplicate part_5 rows from current, processed post parts 5-7 in parallel, and scanned profiles with 11 workers. Author stats use matched post counts rather than full denominator pass.",
+    "unreadable_files":  [
+                             "posts/part_3(1).parquet: skipping unreadable parquet file: Parquet error: Invalid Parquet file. Corrupt footer"
+                         ]
+}

model_config.json

@@ -0,0 +1,37 @@
+{
+  "model_type": "hashed-logreg-v7-fasttext-student",
+  "task": "multi-label text classification",
+  "labels": [
+    "crypto",
+    "tech"
+  ],
+  "inputs": {
+    "post": "tweet/post text",
+    "profile": "bio + display_name + username + professional_category + url"
+  },
+  "artifacts": {
+    "classifier": "text_classifier.json",
+    "distillation_metrics": "distill_metrics.json",
+    "distillation_thresholds": "recommended_thresholds_distillation.json",
+    "full_run_manifest": "full_run_manifest.json"
+  },
+  "recommended_thresholds": {
+    "high_precision_full_run": {
+      "post_crypto": 0.85,
+      "post_tech": 0.9,
+      "profile_crypto": 0.9,
+      "profile_tech": 0.99
+    },
+    "distillation_tuned": {
+      "post_crypto": 0.58,
+      "post_tech": 0.44,
+      "profile_crypto": 0.34,
+      "profile_tech": 0.38
+    }
+  },
+  "notes": [
+    "Rust-compatible hashed logistic-regression student distilled from a V7 fastText teacher.",
+    "The production full-corpus run used the high_precision_full_run thresholds.",
+    "Metrics in distill_metrics.json are proxy metrics against teacher/weak labels, not a final human-labeled benchmark."
+  ]
+}

recommended_thresholds_distillation.json

@@ -0,0 +1,6 @@
+{
+  "post_crypto": 0.58,
+  "post_tech": 0.44,
+  "profile_crypto": 0.34,
+  "profile_tech": 0.38
+}

text_classifier.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8e48c5716df193fa8d96c7b85aa956e2bc79a7ece9587fe52f6286bdabf2928
+size 41342153