Publish OpenJevX RLCD checkpoint and ONNX export

README.md

@@ -0,0 +1,39 @@
+---
+license: apache-2.0
+library_name: laya
+pipeline_tag: text-classification
+tags:
+- system-one
+- typed-decisions
+- rlcd
+- jev-compatible
+- onnx
+base_model: convaiinnovations/laya
+---
+
+# OpenJevX
+
+OpenJevX is an open-weight, non-autoregressive System One decision model specialized from
+[Laya](https://huggingface.co/convaiinnovations/laya), which uses
+`answerdotai/ModernBERT-large` plus a dynamic typed-decision head.
+
+It accepts runtime-defined `choice`, `score`, and `noul` questions and returns calibrated
+probabilities in one forward pass. It is compatible with the TypeSafe Jev `/v1/systemone`
+request shape through the OpenJevX server.
+
+## Benchmark
+
+- Typed-decisions accuracy: **0.774**
+- Choice: **0.745**
+- Score: **0.731**
+- Noul: **0.860**
+- GPU: `NVIDIA GeForce RTX 4090`
+- CUDA p50 latency per five-question case: **22.8 ms**
+
+The benchmark uses the untouched 400-case, 2,000-decision test split from
+`LocalLLaMA/typed-decisions`. Training uses only its 1,200-case train split.
+
+## Attribution
+
+OpenJevX is a derivative of Laya by ConvAI Innovations and ModernBERT by Answer.AI and
+LightOn. Laya and ModernBERT are Apache-2.0 licensed.

SHA256SUMS

@@ -0,0 +1,14 @@
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  TRAINING_COMPLETE
+90758b7aa7bd9b9dfb4ea0baec8d885be8a82ec7710f40149d05d6286d2e7dd9  benchmark.json
+5268d24ad3b77c8151de5dcb0762ba4391619aad9ab0bda33e36fb083cfeae6d  checkpoint_latest/encoder/config.json
+d9723d7a7e7fd8173583b191861d98006fb3b18e4c1f67a4bc559f5ae4543da7  checkpoint_latest/model.safetensors
+fd7237529fb5d575a6ea23a6b2964765d94630765971a1e96d577ce597338724  checkpoint_latest/rl_agent_config.json
+6c8aaa9a542084f2457eab775d4eeb51f92a70c0fd9de28d5edb0ddec3c08d30  checkpoint_latest/tokenizer/tokenizer.json
+5926e6ec4294f80294bd98d9176defa9fbae7f140525d06a1da987488e74a973  checkpoint_latest/tokenizer/tokenizer_config.json
+5268d24ad3b77c8151de5dcb0762ba4391619aad9ab0bda33e36fb083cfeae6d  encoder/config.json
+d9723d7a7e7fd8173583b191861d98006fb3b18e4c1f67a4bc559f5ae4543da7  model.safetensors
+02ec28639b86bc9891f396c2a86909234df080336b59ef13addb0cec8eec4b93  openjevx.int8.onnx
+841f2908c9d97736c691668cb025bfc299acaa7b049b94754d9f2c5fbd02b878  openjevx.onnx
+71542d013de1bbba34836f9f3ad854a41f4468d4ec47b7cdfcecb6f9b2cbf2b3  rl_agent_config.json
+6c8aaa9a542084f2457eab775d4eeb51f92a70c0fd9de28d5edb0ddec3c08d30  tokenizer/tokenizer.json
+08d4cf3ac4dca381759441b85b91a6d40e688471dcd33d15d6649eb0a9a854d1  tokenizer/tokenizer_config.json

benchmark.json

@@ -0,0 +1,24 @@
+{
+  "model": "openjevx",
+  "base_model": "convaiinnovations/laya",
+  "accuracy": 0.774,
+  "per_type": {
+    "choice": {
+      "accuracy": 0.745,
+      "n": 600
+    },
+    "score": {
+      "accuracy": 0.73125,
+      "n": 800
+    },
+    "noul": {
+      "accuracy": 0.86,
+      "n": 600
+    }
+  },
+  "latency_ms": {
+    "p50": 22.807621397078037,
+    "p95": 28.716788906604023
+  },
+  "device": "NVIDIA GeForce RTX 4090"
+}

encoder/config.json

@@ -0,0 +1,84 @@
+{
+  "architectures": [
+    "ModernBertForMaskedLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 50281,
+  "classifier_activation": "gelu",
+  "classifier_bias": false,
+  "classifier_dropout": 0.0,
+  "classifier_pooling": "mean",
+  "cls_token_id": 50281,
+  "decoder_bias": true,
+  "deterministic_flash_attn": false,
+  "dtype": "float32",
+  "embedding_dropout": 0.0,
+  "eos_token_id": 50282,
+  "global_attn_every_n_layers": 3,
+  "gradient_checkpointing": false,
+  "hidden_activation": "gelu",
+  "hidden_size": 1024,
+  "initializer_cutoff_factor": 2.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 2624,
+  "layer_norm_eps": 1e-05,
+  "layer_types": [
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "local_attention": 128,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "mlp_dropout": 0.0,
+  "model_type": "modernbert",
+  "norm_bias": false,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "pad_token_id": 50283,
+  "position_embedding_type": "absolute",
+  "repad_logits_with_grad": false,
+  "rope_parameters": {
+    "full_attention": {
+      "rope_theta": 160000.0,
+      "rope_type": "default"
+    },
+    "sliding_attention": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    }
+  },
+  "sep_token_id": 50282,
+  "sparse_pred_ignore_index": -100,
+  "sparse_prediction": false,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.17.0",
+  "vocab_size": 50368
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9723d7a7e7fd8173583b191861d98006fb3b18e4c1f67a4bc559f5ae4543da7
+size 842609220

openjevx.int8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02ec28639b86bc9891f396c2a86909234df080336b59ef13addb0cec8eec4b93
+size 598012474

rl_agent_config.json

@@ -0,0 +1,28 @@
+{
+  "encoder": "answerdotai/ModernBERT-large",
+  "head_layers": 2,
+  "max_len": 1024,
+  "head_max_len": 256,
+  "max_prefixes": 6,
+  "act_costs": {
+    "escalate": 0.5
+  },
+  "cost_wrong_act": 3.0,
+  "amp_dtype": "bf16",
+  "model_name": "openjevx",
+  "temperature": [
+    1.02553391456604,
+    1.034013271331787,
+    1.147902488708496
+  ],
+  "training": {
+    "updates": 7313,
+    "epochs_completed": 1,
+    "hours": 1.96,
+    "world_size": 1,
+    "fine_tuned_from_checkpoint": true
+  },
+  "gradient_checkpointing": true,
+  "max_tokens_per_batch": 4096,
+  "fine_tuned": true
+}

tokenizer/tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "local_files_only": false,
+  "mask_token": "[MASK]",
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 8192,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "unk_token": "[UNK]"
+}