diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/DONE b/e33/S1-E33-M10-rt-n16k-a100-syn/DONE new file mode 100644 index 0000000000000000000000000000000000000000..85322eceec28ef2135ba16791281f76eca001ab7 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/DONE @@ -0,0 +1 @@ +2026-10-24 12:31:59 \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/README.md b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/README.md new file mode 100644 index 0000000000000000000000000000000000000000..57167079ef3f181aacdbd678a733d71697d18f63 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/README.md @@ -0,0 +1,206 @@ +--- +base_model: /workspace/jev_kt/models/hf/JEV-9B +library_name: peft +tags: +- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **Developed by:** [More Information Needed] +- **Funded by [optional]:** [More Information Needed] +- **Shared by [optional]:** [More Information Needed] +- **Model type:** [More Information Needed] +- **Language(s) (NLP):** [More Information Needed] +- **License:** [More Information Needed] +- **Finetuned from model [optional]:** [More Information Needed] + +### Model Sources [optional] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. + +## How to Get Started with the Model + +Use the code below to get started with the model. + +[More Information Needed] + +## Training Details + +### Training Data + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). + +- **Hardware Type:** [More Information Needed] +- **Hours used:** [More Information Needed] +- **Cloud Provider:** [More Information Needed] +- **Compute Region:** [More Information Needed] +- **Carbon Emitted:** [More Information Needed] + +## Technical Specifications [optional] + +### Model Architecture and Objective + +[More Information Needed] + +### Compute Infrastructure + +[More Information Needed] + +#### Hardware + +[More Information Needed] + +#### Software + +[More Information Needed] + +## Citation [optional] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[More Information Needed] + +## More Information [optional] + +[More Information Needed] + +## Model Card Authors [optional] + +[More Information Needed] + +## Model Card Contact + +[More Information Needed] +### Framework versions + +- PEFT 0.21.2 \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..83ff7283c2262b6957d99dc5d0456957fd2adb52 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json @@ -0,0 +1,57 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 32, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 16, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "q_proj", + "out_proj", + "down_proj", + "v_proj", + "o_proj", + "k_proj", + "up_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9de9ec5219ba63287ed17876fc6a5284b6fd8655 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3a8ca54346d8be0f2b16b0d745cf749492f4d53dd9e2d475c96efee98b01839 +size 160486456 diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d093d03334d1bcd8511a3867adb084f382fd3c5 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json @@ -0,0 +1,58 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": null, + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64.0, + "lora_bias": false, + "lora_dropout": 0.0, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "lm_head", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5b3ee6244056e2c92d55a31722bf44dd3f2c1c6c --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:923bb59225a7e6a362da7f25ac4849ad5380fd7e8caba63ef739f292bec44675 +size 176641832 diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json new file mode 100644 index 0000000000000000000000000000000000000000..cae139a3288faf6eb7e182adc33890763c598b4c --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json @@ -0,0 +1,99 @@ +{ + "bias": [ + 0.00045958356349729, + -0.0004598347295541316, + -0.0022042542695999146, + -0.0013978549977764487, + -0.00023735652212053537, + 0.0020917123183608055, + 0.0016224675346165895, + -0.002466864651069045, + -0.0005318059120327234, + -8.45411341288127e-05, + -7.737032137811184e-05, + 0.0007597437361255288, + -0.00017794813902582973, + -0.0037094249855726957, + -0.004070935305207968, + -0.005084027070552111, + -0.005918621551245451, + -0.004334166180342436, + -0.004935281816869974, + -0.005004563368856907, + -0.00402710260823369, + -0.003410839010030031, + -0.005615165922790766, + -0.0034875874407589436 + ], + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature" +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/calibration.json b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/calibration.json new file mode 100644 index 0000000000000000000000000000000000000000..ab3d843cda87700e508d8e33127155e35248dc50 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/calibration.json @@ -0,0 +1,29 @@ +{ + "version": 1, + "per_kind": { + "noul": 0.7724464809625557, + "choice": 1.0270244461420386, + "score": 0.9929109453941127 + }, + "fit": { + "noul": { + "n": 1087, + "T": 0.7724464809625557, + "ce_before": 0.5699275732040405, + "ce_after": 0.5652568340301514 + }, + "choice": { + "n": 60, + "T": 1.0270244461420386, + "ce_before": 0.6716148257255554, + "ce_after": 0.6714505553245544 + }, + "score": { + "n": 53, + "T": 0.9929109453941127, + "ce_before": 1.0511720180511475, + "ce_after": 1.0511550903320312 + } + }, + "source": "models/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000" +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/head.safetensors b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/head.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..aa9249a8669fd8d5a0744eb94dcfc96b4641b7e6 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/head.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9330c8f5460cbe3894d3cf5eccd9ceca662a9daaeb7a787d139aec96c6c1990b +size 393464 diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/judge_config.json b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/judge_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3fbdcc8e6d9b57adddb4f47c129e308ec07b3d7f --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000/judge_config.json @@ -0,0 +1,164 @@ +{ + "base_model_path": "/root/models/Qwen3.5-9B", + "hidden_size": 4096, + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "kinds": [ + "noul", + "choice", + "score" + ], + "model_name": "jev-judge-qwen35-9b", + "model_version": "0.8.0", + "lm_head_included": true, + "weights_mode": "unmerged", + "adapter_subfolder": "adapter", + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ] + }, + "trained_stage": "e16", + "source_checkpoint": "models/e33/S1-E33-M10-rt-n16k-a100-syn/ckpt-16000", + "softcap": null, + "e16": { + "data": "data/e33/E33-M10-rt-n16k-a100", + "data_manifest": { + "arm": "E33-M10-rt-n16k-a100", + "name": null, + "train_sha256": "e0cc7b84a6deb613bff644587c8fc7dda043c2c3266043dc7df4c0c79180428f", + "n_train": 16000 + }, + "init": "released", + "objective": "s-log", + "sigma": 0.5, + "K": 8, + "reward": "brier", + "clip": 0.2, + "passes": 1, + "lr": 0.0001, + "lr_head": 0.0002, + "rows_per_step": 32, + "train_seed": 0, + "replay": "data/e16/replay/train.jsonl", + "replay_frac": 0.05, + "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", + "max_len": 3072, + "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", + "total_rows": 16000, + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05 + }, + "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", + "event": "ckpt", + "tag": "ckpt-16000", + "rows_seen": 16000, + "replay_seen": 842, + "step": 500, + "tokens": 9820167, + "tok_per_s": 854.7992102709214, + "rows_per_s": 1.4660166465301607, + "val": { + "n": 1200, + "dropped_long": 0, + "noul_ce": 0.5699275732040405, + "noul_T": 0.7724464809625557, + "choice_ce": 0.6716148257255554, + "choice_T": 1.0270244461420386, + "score_ce": 1.0511720180511475, + "score_T": 0.9929109453941127, + "noul_auc": 0.7252821333333332, + "noul_acc": 0.699, + "noul_f1": 0.7762081784386617, + "noul_nll": 0.5881762043386698, + "noul_ece": 0.059616081506013885, + "noul_brier": 0.201311856030733, + "noul_sharp": 0.1597384661436081, + "entropy": 0.6134923100471497 + }, + "dropped_long": 0, + "peak_gib": 18.236003875732422 + } +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-a100-syn/train_log.jsonl b/e33/S1-E33-M10-rt-n16k-a100-syn/train_log.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..b0a642861e60d99810ed28b94bc25283b1b90e2d --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-a100-syn/train_log.jsonl @@ -0,0 +1,57 @@ +{"time": "09:20:27", "event": "start", "data": "data/e33/E33-M10-rt-n16k-a100", "data_manifest": {"arm": "E33-M10-rt-n16k-a100", "name": null, "train_sha256": "e0cc7b84a6deb613bff644587c8fc7dda043c2c3266043dc7df4c0c79180428f", "n_train": 16000}, "init": "released", "objective": "s-log", "sigma": 0.5, "K": 8, "reward": "brier", "clip": 0.2, "passes": 1, "lr": 0.0001, "lr_head": 0.0002, "rows_per_step": 32, "train_seed": 0, "replay": "data/e16/replay/train.jsonl", "replay_frac": 0.05, "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", "max_len": 3072, "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", "total_rows": 16000, "lora": {"r": 16, "lora_alpha": 32, "lora_dropout": 0.05}, "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", "device": "cuda", "load_s": 77.06579637527466} +{"time": "09:24:08", "event": "step", "step": 10, "rows_seen": 320, "lr": 7.333333333333333e-05, "tok_per_s": 891.0474855226341, "loss": 0.6033337190747261, "entropy": 0.6253835633397102, "sharp": 0.15371828898787498} +{"time": "09:27:49", "event": "step", "step": 20, "rows_seen": 640, "lr": 9.997508952966197e-05, "tok_per_s": 898.7678657186418, "loss": 0.7838748916983604, "entropy": 0.6202669739723206, "sharp": 0.17797065526247025} +{"time": "09:31:29", "event": "step", "step": 30, "rows_seen": 960, "lr": 9.977596250541417e-05, "tok_per_s": 900.5382753446562, "loss": 0.6850737277418375, "entropy": 0.6641247570514679, "sharp": 0.10811960324645042} +{"time": "09:35:01", "event": "step", "step": 40, "rows_seen": 1280, "lr": 9.937854366485982e-05, "tok_per_s": 900.3117851157423, "loss": 0.6847868859767914, "entropy": 0.6819761171936989, "sharp": 0.05960580799728632} +{"time": "09:38:33", "event": "step", "step": 50, "rows_seen": 1600, "lr": 9.8784499920729e-05, "tok_per_s": 900.4366446851287, "loss": 0.5863783843815327, "entropy": 0.575633630156517, "sharp": 0.19002662599086761} +{"time": "09:42:12", "event": "step", "step": 60, "rows_seen": 1920, "lr": 9.79963228989294e-05, "tok_per_s": 903.3374492922703, "loss": 0.5031997989863157, "entropy": 0.6125824078917503, "sharp": 0.1816745474934578} +{"time": "09:45:44", "event": "step", "step": 70, "rows_seen": 2240, "lr": 9.701731848780134e-05, "tok_per_s": 903.7665703758253, "loss": 0.606873907148838, "entropy": 0.6636044085025787, "sharp": 0.10620175953954458} +{"time": "09:49:21", "event": "step", "step": 80, "rows_seen": 2560, "lr": 9.585159297207315e-05, "tok_per_s": 903.3837596631068, "loss": 0.67482590675354, "entropy": 0.6118676699697971, "sharp": 0.15935452934354544} +{"time": "09:52:53", "event": "step", "step": 90, "rows_seen": 2880, "lr": 9.450403580967612e-05, "tok_per_s": 903.3855343113702, "loss": 0.6431846879422665, "entropy": 0.6385258324444294, "sharp": 0.1421925025060773} +{"time": "09:56:36", "event": "step", "step": 100, "rows_seen": 3200, "lr": 9.298029912365855e-05, "tok_per_s": 903.9834459122642, "loss": 0.6846701428294182, "entropy": 0.6390752457082272, "sharp": 0.14440846256911755} +{"time": "09:58:00", "event": "val", "step": 100, "rows_seen": 3200, "n": 500, "dropped_long": 0, "noul_ce": 0.6084082722663879, "noul_T": 0.6766352911501362, "noul_auc": 0.7320551551872544, "noul_acc": 0.702, "noul_f1": 0.764612954186414, "noul_nll": 0.6007609467208386, "noul_ece": 0.10155431270599366, "noul_brier": 0.20541022797599395, "noul_sharp": 0.13225561380386353, "entropy": 0.6460034251213074} +{"time": "10:01:37", "event": "step", "step": 110, "rows_seen": 3520, "lr": 9.128677399521752e-05, "tok_per_s": 873.0366049535218, "loss": 0.6482396014034748, "entropy": 0.6171336583793163, "sharp": 0.17463437281548977} +{"time": "10:05:05", "event": "step", "step": 120, "rows_seen": 3840, "lr": 8.943056365728335e-05, "tok_per_s": 875.1641802576061, "loss": 0.5996145531535149, "entropy": 0.6567100770771503, "sharp": 0.11905446369200945} +{"time": "10:08:38", "event": "step", "step": 130, "rows_seen": 4160, "lr": 8.741945370109202e-05, "tok_per_s": 876.9032903879504, "loss": 0.6384086310863495, "entropy": 0.6788572743535042, "sharp": 0.06982394959777594} +{"time": "10:12:14", "event": "step", "step": 140, "rows_seen": 4480, "lr": 8.526187942070997e-05, "tok_per_s": 878.9804242819296, "loss": 0.5768000446259975, "entropy": 0.6693370379507542, "sharp": 0.09189106337726116} +{"time": "10:15:49", "event": "step", "step": 150, "rows_seen": 4800, "lr": 8.296689043247915e-05, "tok_per_s": 880.4348184722512, "loss": 0.6017400855198503, "entropy": 0.5698162391781807, "sharp": 0.21178777143359184} +{"time": "10:19:27", "event": "step", "step": 160, "rows_seen": 5120, "lr": 8.054411271778132e-05, "tok_per_s": 881.5823790589051, "loss": 0.6542864292860031, "entropy": 0.5234885523095727, "sharp": 0.2540337862446904} +{"time": "10:23:00", "event": "step", "step": 170, "rows_seen": 5440, "lr": 7.800370824832674e-05, "tok_per_s": 882.7998236958621, "loss": 0.5910687409341335, "entropy": 0.6635227352380753, "sharp": 0.10423886124044657} +{"time": "10:26:37", "event": "step", "step": 180, "rows_seen": 5760, "lr": 7.535633236331324e-05, "tok_per_s": 883.9654856469632, "loss": 0.5948001351207495, "entropy": 0.6536477170884609, "sharp": 0.11783802974969149} +{"time": "10:30:13", "event": "step", "step": 190, "rows_seen": 6080, "lr": 7.26130890772308e-05, "tok_per_s": 884.6143325550358, "loss": 0.6801256858743727, "entropy": 0.6189041864126921, "sharp": 0.16625779680907726} +{"time": "10:33:53", "event": "step", "step": 200, "rows_seen": 6400, "lr": 6.978548450576568e-05, "tok_per_s": 885.7659601500735, "loss": 0.6023488454520702, "entropy": 0.6618353426456451, "sharp": 0.10348301753401756} +{"time": "10:35:16", "event": "val", "step": 200, "rows_seen": 6400, "n": 500, "dropped_long": 0, "noul_ce": 0.613629937171936, "noul_T": 0.8413533877688976, "noul_auc": 0.7025704207523923, "noul_acc": 0.674, "noul_f1": 0.7634252539912917, "noul_nll": 0.6114704060256482, "noul_ece": 0.06213674157857896, "noul_brier": 0.21065328342820616, "noul_sharp": 0.14712855219841003, "entropy": 0.6329501271247864} +{"time": "10:38:58", "event": "step", "step": 210, "rows_seen": 6720, "lr": 6.688537860515266e-05, "tok_per_s": 871.2871728674021, "loss": 0.63688157312572, "entropy": 0.6147606391459703, "sharp": 0.17704907432198524} +{"time": "10:42:32", "event": "step", "step": 220, "rows_seen": 7040, "lr": 6.392493542739706e-05, "tok_per_s": 872.5228054587427, "loss": 0.6519083455204964, "entropy": 0.6725699678063393, "sharp": 0.08100294321775436} +{"time": "10:46:13", "event": "step", "step": 230, "rows_seen": 7360, "lr": 6.091657210001309e-05, "tok_per_s": 873.2950804448907, "loss": 0.6356181167066097, "entropy": 0.6462880596518517, "sharp": 0.12306220084428787} +{"time": "10:49:53", "event": "step", "step": 240, "rows_seen": 7680, "lr": 5.7872906744276015e-05, "tok_per_s": 874.3815137635282, "loss": 0.6321443058550358, "entropy": 0.652955025434494, "sharp": 0.11930426955223083} +{"time": "10:53:33", "event": "step", "step": 250, "rows_seen": 8000, "lr": 5.48067055504363e-05, "tok_per_s": 875.7611935040712, "loss": 0.6394157260656357, "entropy": 0.6025685258209705, "sharp": 0.1883924026042223} +{"time": "10:57:14", "event": "step", "step": 260, "rows_seen": 8320, "lr": 5.173082923188119e-05, "tok_per_s": 876.6742822955374, "loss": 0.6191202476620674, "entropy": 0.6671462059020996, "sharp": 0.09179499931633472} +{"time": "11:00:47", "event": "step", "step": 270, "rows_seen": 8640, "lr": 4.865817908283331e-05, "tok_per_s": 877.4646161094252, "loss": 0.6553642265498638, "entropy": 0.6592379845678806, "sharp": 0.1110451053828001} +{"time": "11:04:26", "event": "step", "step": 280, "rows_seen": 8960, "lr": 4.560164286583922e-05, "tok_per_s": 877.971221797522, "loss": 0.6024574413895607, "entropy": 0.6117802187800407, "sharp": 0.17410529032349586} +{"time": "11:07:55", "event": "step", "step": 290, "rows_seen": 9280, "lr": 4.257404075601443e-05, "tok_per_s": 878.6396164404216, "loss": 0.6674257572740316, "entropy": 0.6629147864878178, "sharp": 0.10723049286752939} +{"time": "11:11:39", "event": "step", "step": 300, "rows_seen": 9600, "lr": 3.958807156877371e-05, "tok_per_s": 879.5878966609004, "loss": 0.6327837016433477, "entropy": 0.6372476555407047, "sharp": 0.14154105819761753} +{"time": "11:13:02", "event": "val", "step": 300, "rows_seen": 9600, "n": 500, "dropped_long": 0, "noul_ce": 0.5927555561065674, "noul_T": 0.9140837438194751, "noul_auc": 0.7317513628461967, "noul_acc": 0.666, "noul_f1": 0.7727891156462585, "noul_nll": 0.5982471793740988, "noul_ece": 0.07837080037593842, "noul_brier": 0.20586142497615276, "noul_sharp": 0.17416174709796906, "entropy": 0.6058613657951355} +{"time": "11:16:37", "event": "step", "step": 310, "rows_seen": 9920, "lr": 3.665625949658603e-05, "tok_per_s": 869.6137309110122, "loss": 0.6072483845055103, "entropy": 0.5658436417579651, "sharp": 0.20860496163368225} +{"time": "11:20:14", "event": "step", "step": 320, "rows_seen": 10240, "lr": 3.379090157815887e-05, "tok_per_s": 870.7963165762047, "loss": 0.6785305291414261, "entropy": 0.6227179393172264, "sharp": 0.1529350159689784} +{"time": "11:23:52", "event": "step", "step": 330, "rows_seen": 10560, "lr": 3.10040161203844e-05, "tok_per_s": 871.4461092863057, "loss": 0.6080680526793003, "entropy": 0.640902929008007, "sharp": 0.1289919689297676} +{"time": "11:27:28", "event": "step", "step": 340, "rows_seen": 10880, "lr": 2.8307292289383776e-05, "tok_per_s": 872.4537510502763, "loss": 0.6423982172273099, "entropy": 0.6232390608638525, "sharp": 0.1561298780143261} +{"time": "11:31:08", "event": "step", "step": 350, "rows_seen": 11200, "lr": 2.571204108208194e-05, "tok_per_s": 873.4323797092313, "loss": 0.65891507640481, "entropy": 0.6288746707141399, "sharp": 0.14588055666536093} +{"time": "11:34:44", "event": "step", "step": 360, "rows_seen": 11520, "lr": 2.322914788395595e-05, "tok_per_s": 874.1338999630548, "loss": 0.6176988501101732, "entropy": 0.6252961456775665, "sharp": 0.13359309546649456} +{"time": "11:38:28", "event": "step", "step": 370, "rows_seen": 11840, "lr": 2.0869026811945157e-05, "tok_per_s": 874.6831089882023, "loss": 0.744150634855032, "entropy": 0.6299264468252659, "sharp": 0.1376659180969} +{"time": "11:41:58", "event": "step", "step": 380, "rows_seen": 12160, "lr": 1.8641577034025627e-05, "tok_per_s": 875.1158409723324, "loss": 0.5802575796842575, "entropy": 0.6352507062256336, "sharp": 0.14148460514843464} +{"time": "11:45:39", "event": "step", "step": 390, "rows_seen": 12480, "lr": 1.6556141248659472e-05, "tok_per_s": 875.5994456569073, "loss": 0.7059405953623354, "entropy": 0.6364332772791386, "sharp": 0.12399807758629322} +{"time": "11:49:25", "event": "step", "step": 400, "rows_seen": 12800, "lr": 1.4621466498270773e-05, "tok_per_s": 876.3615689884818, "loss": 0.6372781246900558, "entropy": 0.6148921549320221, "sharp": 0.15839910320937634} +{"time": "11:50:48", "event": "val", "step": 400, "rows_seen": 12800, "n": 500, "dropped_long": 0, "noul_ce": 0.5831224322319031, "noul_T": 0.973220551345713, "noul_auc": 0.7322745607669069, "noul_acc": 0.71, "noul_f1": 0.7792998477929984, "noul_nll": 0.5908164388295263, "noul_ece": 0.06788681399822236, "noul_brier": 0.20089749663804382, "noul_sharp": 0.18074163794517517, "entropy": 0.5872051119804382} +{"time": "11:54:30", "event": "step", "step": 410, "rows_seen": 13120, "lr": 1.284566748110963e-05, "tok_per_s": 869.1485172518481, "loss": 0.5657886452972889, "entropy": 0.6486924085766077, "sharp": 0.11286498792469501} +{"time": "11:58:14", "event": "step", "step": 420, "rows_seen": 13440, "lr": 1.1236192515387821e-05, "tok_per_s": 870.0233991919794, "loss": 0.5447256825864315, "entropy": 0.6085926666855812, "sharp": 0.16909427661448717} +{"time": "12:02:00", "event": "step", "step": 430, "rows_seen": 13760, "lr": 9.799792298443728e-06, "tok_per_s": 870.6947731484165, "loss": 0.6258899383246899, "entropy": 0.6019481364637613, "sharp": 0.16560775693506002} +{"time": "12:05:35", "event": "step", "step": 440, "rows_seen": 14080, "lr": 8.542491591971554e-06, "tok_per_s": 871.4700596618806, "loss": 0.6144727133214474, "entropy": 0.6315025687217712, "sharp": 0.1404762426391244} +{"time": "12:09:21", "event": "step", "step": 450, "rows_seen": 14400, "lr": 7.469563952077155e-06, "tok_per_s": 872.1645453687298, "loss": 0.5760183185338974, "entropy": 0.656073272228241, "sharp": 0.10922098997980356} +{"time": "12:12:58", "event": "step", "step": 460, "rows_seen": 14720, "lr": 6.585509610150972e-06, "tok_per_s": 872.7241558853473, "loss": 0.6531402003020048, "entropy": 0.6523221842944622, "sharp": 0.12087107170373201} +{"time": "12:16:37", "event": "step", "step": 470, "rows_seen": 15040, "lr": 5.894036597333646e-06, "tok_per_s": 873.3651392286066, "loss": 0.6043087020516396, "entropy": 0.6481751687824726, "sharp": 0.1281145429238677} +{"time": "12:20:16", "event": "step", "step": 480, "rows_seen": 15360, "lr": 5.398045191744809e-06, "tok_per_s": 873.6749996879207, "loss": 0.6318537872284651, "entropy": 0.6360546965152025, "sharp": 0.14198052696883678} +{"time": "12:23:56", "event": "step", "step": 490, "rows_seen": 15680, "lr": 5.099615753708697e-06, "tok_per_s": 874.2588882022271, "loss": 0.6480370759963989, "entropy": 0.6473933756351471, "sharp": 0.11683804169297218} +{"time": "12:27:35", "event": "step", "step": 500, "rows_seen": 16000, "lr": 5e-06, "tok_per_s": 874.6643402276964, "loss": 0.6334035154432058, "entropy": 0.6499393992125988, "sharp": 0.12684703432023525} +{"time": "12:28:58", "event": "val", "step": 500, "rows_seen": 16000, "n": 500, "dropped_long": 0, "noul_ce": 0.5850250124931335, "noul_T": 0.7595249686054862, "noul_auc": 0.7318188722553206, "noul_acc": 0.702, "noul_f1": 0.7682737169517885, "noul_nll": 0.590603579364717, "noul_ece": 0.06252238416671752, "noul_brier": 0.20186487573155232, "noul_sharp": 0.1538901925086975, "entropy": 0.6182107925415039} +{"time": "12:31:59", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "replay_seen": 842, "step": 500, "tokens": 9820167, "tok_per_s": 854.7992102709214, "rows_per_s": 1.4660166465301607, "val": {"n": 1200, "dropped_long": 0, "noul_ce": 0.5699275732040405, "noul_T": 0.7724464809625557, "choice_ce": 0.6716148257255554, "choice_T": 1.0270244461420386, "score_ce": 1.0511720180511475, "score_T": 0.9929109453941127, "noul_auc": 0.7252821333333332, "noul_acc": 0.699, "noul_f1": 0.7762081784386617, "noul_nll": 0.5881762043386698, "noul_ece": 0.059616081506013885, "noul_brier": 0.201311856030733, "noul_sharp": 0.1597384661436081, "entropy": 0.6134923100471497}, "dropped_long": 0, "peak_gib": 18.236003875732422} diff --git a/e33/S1-E33-M10-rt-n16k-syn/DONE b/e33/S1-E33-M10-rt-n16k-syn/DONE new file mode 100644 index 0000000000000000000000000000000000000000..9d5eefc9c93b5a8475edd28c9c1717cdddec737d --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/DONE @@ -0,0 +1 @@ +2026-10-23 04:32:35 \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/README.md b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/README.md new file mode 100644 index 0000000000000000000000000000000000000000..57167079ef3f181aacdbd678a733d71697d18f63 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/README.md @@ -0,0 +1,206 @@ +--- +base_model: /workspace/jev_kt/models/hf/JEV-9B +library_name: peft +tags: +- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **Developed by:** [More Information Needed] +- **Funded by [optional]:** [More Information Needed] +- **Shared by [optional]:** [More Information Needed] +- **Model type:** [More Information Needed] +- **Language(s) (NLP):** [More Information Needed] +- **License:** [More Information Needed] +- **Finetuned from model [optional]:** [More Information Needed] + +### Model Sources [optional] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. + +## How to Get Started with the Model + +Use the code below to get started with the model. + +[More Information Needed] + +## Training Details + +### Training Data + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). + +- **Hardware Type:** [More Information Needed] +- **Hours used:** [More Information Needed] +- **Cloud Provider:** [More Information Needed] +- **Compute Region:** [More Information Needed] +- **Carbon Emitted:** [More Information Needed] + +## Technical Specifications [optional] + +### Model Architecture and Objective + +[More Information Needed] + +### Compute Infrastructure + +[More Information Needed] + +#### Hardware + +[More Information Needed] + +#### Software + +[More Information Needed] + +## Citation [optional] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[More Information Needed] + +## More Information [optional] + +[More Information Needed] + +## Model Card Authors [optional] + +[More Information Needed] + +## Model Card Contact + +[More Information Needed] +### Framework versions + +- PEFT 0.21.2 \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/adapter_config.json b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..0ca2f71094bd9a58bcebcb024fb9d40ee1ec4de4 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/adapter_config.json @@ -0,0 +1,57 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 32, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 16, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "out_proj", + "v_proj", + "k_proj", + "q_proj", + "up_proj", + "in_proj_qkv", + "gate_proj", + "in_proj_z", + "down_proj", + "o_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..06b96bab16d2dafcd6d2d83e9a8691599aa758b9 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acd7398bcc7b780b9076cd2cd710473219c65da73c87a462f1bf7db725e9bcd1 +size 160486456 diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d093d03334d1bcd8511a3867adb084f382fd3c5 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json @@ -0,0 +1,58 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": null, + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64.0, + "lora_bias": false, + "lora_dropout": 0.0, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "lm_head", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..71fadd3595506966399e96a6a1fccba8663d6f26 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1296a823eceecc39bff4c46d0e1947b9302822acce5e88ddf53fcd4389962d5c +size 176641832 diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json new file mode 100644 index 0000000000000000000000000000000000000000..f36e3d5c0553778de5cf6e59b703956dfbbf979c --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json @@ -0,0 +1,99 @@ +{ + "bias": [ + 0.0003609308914747089, + -0.0003611863648984581, + -0.002497137524187565, + -0.0015482216840609908, + -0.0005122810835018754, + 0.002160529140383005, + 0.0014962729765102267, + -0.00209335470572114, + -0.0005815276526845992, + -7.023397483862936e-05, + 2.4424463845207356e-05, + 0.0003453893295954913, + -0.00025366476620547473, + -0.003527699038386345, + -0.004392900038510561, + -0.005109368823468685, + -0.00589381018653512, + -0.004366089124232531, + -0.004935386590659618, + -0.00498735997825861, + -0.003816308919340372, + -0.0033886691089719534, + -0.005703995935618877, + -0.0034793075174093246 + ], + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature" +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/calibration.json b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/calibration.json new file mode 100644 index 0000000000000000000000000000000000000000..d4035c9865e306941f3b54919704f9209de6ec77 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/calibration.json @@ -0,0 +1,29 @@ +{ + "version": 1, + "per_kind": { + "noul": 0.7800532877395329, + "choice": 1.0466806238139477, + "score": 0.9804041633946682 + }, + "fit": { + "noul": { + "n": 1087, + "T": 0.7800532877395329, + "ce_before": 0.5706181526184082, + "ce_after": 0.5660994648933411 + }, + "choice": { + "n": 60, + "T": 1.0466806238139477, + "ce_before": 0.672325849533081, + "ce_after": 0.6718399524688721 + }, + "score": { + "n": 53, + "T": 0.9804041633946682, + "ce_before": 1.0526014566421509, + "ce_after": 1.0524842739105225 + } + }, + "source": "models/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000" +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/head.safetensors b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/head.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b7171638fa12af60e2ffe1ca394ca4f6faa25d70 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/head.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1ea418ef3e6efca8f1a8bd124c42306c9abf5f03aae0d9dec4a89cf24f60687 +size 393464 diff --git a/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/judge_config.json b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/judge_config.json new file mode 100644 index 0000000000000000000000000000000000000000..8f46aaa06d0993ce2e555ae2b0d36982dc4b5fa5 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000/judge_config.json @@ -0,0 +1,164 @@ +{ + "base_model_path": "/root/models/Qwen3.5-9B", + "hidden_size": 4096, + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "kinds": [ + "noul", + "choice", + "score" + ], + "model_name": "jev-judge-qwen35-9b", + "model_version": "0.8.0", + "lm_head_included": true, + "weights_mode": "unmerged", + "adapter_subfolder": "adapter", + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ] + }, + "trained_stage": "e16", + "source_checkpoint": "models/e33/S1-E33-M10-rt-n16k-syn/ckpt-16000", + "softcap": null, + "e16": { + "data": "data/e33/E33-M10-rt-n16k", + "data_manifest": { + "arm": "E33-M10-rt-n16k", + "name": null, + "train_sha256": "39574ddaa158a99572227f29fcc2be761352b45f2f70805ff1c7883379ef2228", + "n_train": 16000 + }, + "init": "released", + "objective": "s-log", + "sigma": 0.5, + "K": 8, + "reward": "brier", + "clip": 0.2, + "passes": 1, + "lr": 0.0001, + "lr_head": 0.0002, + "rows_per_step": 32, + "train_seed": 0, + "replay": "data/e16/replay/train.jsonl", + "replay_frac": 0.05, + "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", + "max_len": 3072, + "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", + "total_rows": 16000, + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05 + }, + "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", + "event": "ckpt", + "tag": "ckpt-16000", + "rows_seen": 16000, + "replay_seen": 842, + "step": 500, + "tokens": 9820167, + "tok_per_s": 853.3575821037343, + "rows_per_s": 1.4635441932658964, + "val": { + "n": 1200, + "dropped_long": 0, + "noul_ce": 0.5706181526184082, + "noul_T": 0.7800532877395329, + "choice_ce": 0.672325849533081, + "choice_T": 1.0466806238139477, + "score_ce": 1.0526014566421509, + "score_T": 0.9804041633946682, + "noul_auc": 0.7283541333333334, + "noul_acc": 0.698, + "noul_f1": 0.7719033232628398, + "noul_nll": 0.5866083004325628, + "noul_ece": 0.05212497982382775, + "noul_brier": 0.20057906316333285, + "noul_sharp": 0.16233906149864197, + "entropy": 0.6147943139076233 + }, + "dropped_long": 0, + "peak_gib": 18.236003875732422 + } +} \ No newline at end of file diff --git a/e33/S1-E33-M10-rt-n16k-syn/train_log.jsonl b/e33/S1-E33-M10-rt-n16k-syn/train_log.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..0cf78b1a90bed652281f6209afc7e40dda2ca950 --- /dev/null +++ b/e33/S1-E33-M10-rt-n16k-syn/train_log.jsonl @@ -0,0 +1,57 @@ +{"time": "01:20:46", "event": "start", "data": "data/e33/E33-M10-rt-n16k", "data_manifest": {"arm": "E33-M10-rt-n16k", "name": null, "train_sha256": "39574ddaa158a99572227f29fcc2be761352b45f2f70805ff1c7883379ef2228", "n_train": 16000}, "init": "released", "objective": "s-log", "sigma": 0.5, "K": 8, "reward": "brier", "clip": 0.2, "passes": 1, "lr": 0.0001, "lr_head": 0.0002, "rows_per_step": 32, "train_seed": 0, "replay": "data/e16/replay/train.jsonl", "replay_frac": 0.05, "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", "max_len": 3072, "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", "total_rows": 16000, "lora": {"r": 16, "lora_alpha": 32, "lora_dropout": 0.05}, "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", "device": "cuda", "load_s": 68.17974424362183} +{"time": "01:24:26", "event": "step", "step": 10, "rows_seen": 320, "lr": 7.333333333333333e-05, "tok_per_s": 894.3302368416892, "loss": 0.6407141238451004, "entropy": 0.6530866995453835, "sharp": 0.11467453930526972} +{"time": "01:28:07", "event": "step", "step": 20, "rows_seen": 640, "lr": 9.997508952966197e-05, "tok_per_s": 898.5712368978669, "loss": 0.7670875117182732, "entropy": 0.5850043445825577, "sharp": 0.19216507486999035} +{"time": "01:31:47", "event": "step", "step": 30, "rows_seen": 960, "lr": 9.977596250541417e-05, "tok_per_s": 901.0519607560497, "loss": 0.7097546458244324, "entropy": 0.5784373320639133, "sharp": 0.2102995691820979} +{"time": "01:35:19", "event": "step", "step": 40, "rows_seen": 1280, "lr": 9.937854366485982e-05, "tok_per_s": 900.2845654414082, "loss": 0.7051415480673313, "entropy": 0.6509120985865593, "sharp": 0.11466245353221893} +{"time": "01:38:52", "event": "step", "step": 50, "rows_seen": 1600, "lr": 9.8784499920729e-05, "tok_per_s": 899.3461954764921, "loss": 0.6417021974921227, "entropy": 0.6230399906635284, "sharp": 0.16287371329963207} +{"time": "01:42:32", "event": "step", "step": 60, "rows_seen": 1920, "lr": 9.79963228989294e-05, "tok_per_s": 902.1724212711766, "loss": 0.5933823697268963, "entropy": 0.6530337370932102, "sharp": 0.1106806006282568} +{"time": "01:46:04", "event": "step", "step": 70, "rows_seen": 2240, "lr": 9.701731848780134e-05, "tok_per_s": 902.7435101228933, "loss": 0.6372644007205963, "entropy": 0.6816029846668243, "sharp": 0.060833889059722424} +{"time": "01:49:43", "event": "step", "step": 80, "rows_seen": 2560, "lr": 9.585159297207315e-05, "tok_per_s": 901.7272742006631, "loss": 0.6699827425181866, "entropy": 0.6287437509745359, "sharp": 0.1442428957670927} +{"time": "01:53:15", "event": "step", "step": 90, "rows_seen": 2880, "lr": 9.450403580967612e-05, "tok_per_s": 901.3173038745224, "loss": 0.6240650974214077, "entropy": 0.6412920206785202, "sharp": 0.1320671895518899} +{"time": "01:57:00", "event": "step", "step": 100, "rows_seen": 3200, "lr": 9.298029912365855e-05, "tok_per_s": 901.9174655351277, "loss": 0.6517882402986288, "entropy": 0.6239370983093977, "sharp": 0.15501127019524574} +{"time": "01:58:23", "event": "val", "step": 100, "rows_seen": 3200, "n": 500, "dropped_long": 0, "noul_ce": 0.5938156843185425, "noul_T": 0.7507550395879644, "noul_auc": 0.7372365023375133, "noul_acc": 0.714, "noul_f1": 0.7906295754026355, "noul_nll": 0.5917655223309993, "noul_ece": 0.06725453257560729, "noul_brier": 0.20199009536325635, "noul_sharp": 0.14850011467933655, "entropy": 0.6278250813484192} +{"time": "02:02:00", "event": "step", "step": 110, "rows_seen": 3520, "lr": 9.128677399521752e-05, "tok_per_s": 871.1336387513961, "loss": 0.6488936170935631, "entropy": 0.6459421142935753, "sharp": 0.1318453261628747} +{"time": "02:05:30", "event": "step", "step": 120, "rows_seen": 3840, "lr": 8.943056365728335e-05, "tok_per_s": 873.0455970204106, "loss": 0.6333391219377518, "entropy": 0.6602405309677124, "sharp": 0.0976364342495799} +{"time": "02:09:03", "event": "step", "step": 130, "rows_seen": 4160, "lr": 8.741945370109202e-05, "tok_per_s": 874.8895332796241, "loss": 0.646759383380413, "entropy": 0.5998613610863686, "sharp": 0.1769361738115549} +{"time": "02:12:40", "event": "step", "step": 140, "rows_seen": 4480, "lr": 8.526187942070997e-05, "tok_per_s": 876.7330429188311, "loss": 0.6052324529737234, "entropy": 0.6510478630661964, "sharp": 0.10768228489905596} +{"time": "02:16:15", "event": "step", "step": 150, "rows_seen": 4800, "lr": 8.296689043247915e-05, "tok_per_s": 878.2464503690718, "loss": 0.6153158433735371, "entropy": 0.631971400231123, "sharp": 0.16027738526463509} +{"time": "02:19:55", "event": "step", "step": 160, "rows_seen": 5120, "lr": 8.054411271778132e-05, "tok_per_s": 879.2137809500738, "loss": 0.6790948957204819, "entropy": 0.5736175198107958, "sharp": 0.20291051920503378} +{"time": "02:23:29", "event": "step", "step": 170, "rows_seen": 5440, "lr": 7.800370824832674e-05, "tok_per_s": 880.3711935175188, "loss": 0.6004719063639641, "entropy": 0.6468684375286102, "sharp": 0.12351896334439516} +{"time": "02:27:06", "event": "step", "step": 180, "rows_seen": 5760, "lr": 7.535633236331324e-05, "tok_per_s": 881.5599301165362, "loss": 0.6054106764495373, "entropy": 0.6608152016997337, "sharp": 0.09726034570485353} +{"time": "02:30:42", "event": "step", "step": 190, "rows_seen": 6080, "lr": 7.26130890772308e-05, "tok_per_s": 882.3999166719803, "loss": 0.6535965269431472, "entropy": 0.6553077027201653, "sharp": 0.11877382826060057} +{"time": "02:34:22", "event": "step", "step": 200, "rows_seen": 6400, "lr": 6.978548450576568e-05, "tok_per_s": 883.6472831287725, "loss": 0.61347995698452, "entropy": 0.6301144063472748, "sharp": 0.14622782915830612} +{"time": "02:35:45", "event": "val", "step": 200, "rows_seen": 6400, "n": 500, "dropped_long": 0, "noul_ce": 0.5859304070472717, "noul_T": 0.8799948384607998, "noul_auc": 0.7436330188520024, "noul_acc": 0.716, "noul_f1": 0.7874251497005988, "noul_nll": 0.5828873897492886, "noul_ece": 0.06448433697223664, "noul_brier": 0.1981293654491963, "noul_sharp": 0.17154835164546967, "entropy": 0.6047180891036987} +{"time": "02:39:28", "event": "step", "step": 210, "rows_seen": 6720, "lr": 6.688537860515266e-05, "tok_per_s": 869.194350995947, "loss": 0.6664544409140944, "entropy": 0.6718409713357687, "sharp": 0.08500322513282299} +{"time": "02:43:03", "event": "step", "step": 220, "rows_seen": 7040, "lr": 6.392493542739706e-05, "tok_per_s": 870.3098890608844, "loss": 0.6444327458739281, "entropy": 0.6366416960954666, "sharp": 0.13889538496732712} +{"time": "02:46:45", "event": "step", "step": 230, "rows_seen": 7360, "lr": 6.091657210001309e-05, "tok_per_s": 870.9679598953579, "loss": 0.6266806535422802, "entropy": 0.6547847539186478, "sharp": 0.11867003049701452} +{"time": "02:50:25", "event": "step", "step": 240, "rows_seen": 7680, "lr": 5.7872906744276015e-05, "tok_per_s": 872.162954408354, "loss": 0.6558815725147724, "entropy": 0.6340029835700989, "sharp": 0.1507103629410267} +{"time": "02:54:06", "event": "step", "step": 250, "rows_seen": 8000, "lr": 5.48067055504363e-05, "tok_per_s": 873.4662758411903, "loss": 0.642772326245904, "entropy": 0.6260318588465452, "sharp": 0.16632072255015373} +{"time": "02:57:47", "event": "step", "step": 260, "rows_seen": 8320, "lr": 5.173082923188119e-05, "tok_per_s": 874.4971117879062, "loss": 0.6335619688034058, "entropy": 0.6518338844180107, "sharp": 0.12628033384680748} +{"time": "03:01:20", "event": "step", "step": 270, "rows_seen": 8640, "lr": 4.865817908283331e-05, "tok_per_s": 875.2975710934581, "loss": 0.6495114751160145, "entropy": 0.6683188453316689, "sharp": 0.09399524331092834} +{"time": "03:04:59", "event": "step", "step": 280, "rows_seen": 8960, "lr": 4.560164286583922e-05, "tok_per_s": 875.8806219727835, "loss": 0.6487135104835033, "entropy": 0.6416224911808968, "sharp": 0.1189706027507782} +{"time": "03:08:29", "event": "step", "step": 290, "rows_seen": 9280, "lr": 4.257404075601443e-05, "tok_per_s": 876.5671341788939, "loss": 0.6583934891968966, "entropy": 0.6552927903831005, "sharp": 0.11414727661758661} +{"time": "03:12:12", "event": "step", "step": 300, "rows_seen": 9600, "lr": 3.958807156877371e-05, "tok_per_s": 877.6184190910565, "loss": 0.661140576004982, "entropy": 0.6403373926877975, "sharp": 0.13451462611556053} +{"time": "03:13:36", "event": "val", "step": 300, "rows_seen": 9600, "n": 500, "dropped_long": 0, "noul_ce": 0.5856049656867981, "noul_T": 0.8816176327090242, "noul_auc": 0.7410507839530135, "noul_acc": 0.684, "noul_f1": 0.7799442896935933, "noul_nll": 0.5889199020117521, "noul_ece": 0.07037900412082672, "noul_brier": 0.20162840001933632, "noul_sharp": 0.1706424504518509, "entropy": 0.6040006875991821} +{"time": "03:17:11", "event": "step", "step": 310, "rows_seen": 9920, "lr": 3.665625949658603e-05, "tok_per_s": 867.636449941811, "loss": 0.6064267233014107, "entropy": 0.6226922497153282, "sharp": 0.1504072267562151} +{"time": "03:20:48", "event": "step", "step": 320, "rows_seen": 10240, "lr": 3.379090157815887e-05, "tok_per_s": 868.825550660966, "loss": 0.6666115932166576, "entropy": 0.6551536694169044, "sharp": 0.10939735546708107} +{"time": "03:24:27", "event": "step", "step": 330, "rows_seen": 10560, "lr": 3.10040161203844e-05, "tok_per_s": 869.444203220318, "loss": 0.6336435861885548, "entropy": 0.6463650986552238, "sharp": 0.12483089882880449} +{"time": "03:28:03", "event": "step", "step": 340, "rows_seen": 10880, "lr": 2.8307292289383776e-05, "tok_per_s": 870.5032601087813, "loss": 0.6576888412237167, "entropy": 0.6181784169748425, "sharp": 0.1659212587401271} +{"time": "03:31:44", "event": "step", "step": 350, "rows_seen": 11200, "lr": 2.571204108208194e-05, "tok_per_s": 871.4820855569737, "loss": 0.6604277305305004, "entropy": 0.6469684541225433, "sharp": 0.11548988427966833} +{"time": "03:35:20", "event": "step", "step": 360, "rows_seen": 11520, "lr": 2.322914788395595e-05, "tok_per_s": 872.2455562457085, "loss": 0.6381114348769188, "entropy": 0.6450099423527718, "sharp": 0.11503359023481607} +{"time": "03:39:04", "event": "step", "step": 370, "rows_seen": 11840, "lr": 2.0869026811945157e-05, "tok_per_s": 872.8070292594366, "loss": 0.6908784322440624, "entropy": 0.633655920624733, "sharp": 0.13484611455351114} +{"time": "03:42:33", "event": "step", "step": 380, "rows_seen": 12160, "lr": 1.8641577034025627e-05, "tok_per_s": 873.3877540759133, "loss": 0.6250544004142284, "entropy": 0.6358201876282692, "sharp": 0.1346720615401864} +{"time": "03:46:14", "event": "step", "step": 390, "rows_seen": 12480, "lr": 1.6556141248659472e-05, "tok_per_s": 873.9114803828065, "loss": 0.6878696195781231, "entropy": 0.6477202232927084, "sharp": 0.10971695929765701} +{"time": "03:50:00", "event": "step", "step": 400, "rows_seen": 12800, "lr": 1.4621466498270773e-05, "tok_per_s": 874.7192303517428, "loss": 0.6515332609415054, "entropy": 0.6342602297663689, "sharp": 0.13561679888516665} +{"time": "03:51:23", "event": "val", "step": 400, "rows_seen": 12800, "n": 500, "dropped_long": 0, "noul_ce": 0.5833331346511841, "noul_T": 0.8217515025225614, "noul_auc": 0.740055020168436, "noul_acc": 0.714, "noul_f1": 0.7816793893129771, "noul_nll": 0.584851800262928, "noul_ece": 0.05926298546791077, "noul_brier": 0.1992804793556761, "noul_sharp": 0.16393373906612396, "entropy": 0.6101114749908447} +{"time": "03:55:05", "event": "step", "step": 410, "rows_seen": 13120, "lr": 1.284566748110963e-05, "tok_per_s": 867.5312933374554, "loss": 0.627738818526268, "entropy": 0.6642075832933187, "sharp": 0.09371273778378963} +{"time": "03:58:49", "event": "step", "step": 420, "rows_seen": 13440, "lr": 1.1236192515387821e-05, "tok_per_s": 868.4505163717379, "loss": 0.5989300683140755, "entropy": 0.6216349601745605, "sharp": 0.15461522340774536} +{"time": "04:02:35", "event": "step", "step": 430, "rows_seen": 13760, "lr": 9.799792298443728e-06, "tok_per_s": 869.1675793578265, "loss": 0.6381424367427826, "entropy": 0.6140604615211487, "sharp": 0.15680895280092955} +{"time": "04:06:12", "event": "step", "step": 440, "rows_seen": 14080, "lr": 8.542491591971554e-06, "tok_per_s": 869.8846409524637, "loss": 0.6422123610973358, "entropy": 0.6449132598936558, "sharp": 0.12053444609045982} +{"time": "04:09:58", "event": "step", "step": 450, "rows_seen": 14400, "lr": 7.469563952077155e-06, "tok_per_s": 870.5906451027971, "loss": 0.6214579604566097, "entropy": 0.6666248366236687, "sharp": 0.09311869088560343} +{"time": "04:13:35", "event": "step", "step": 460, "rows_seen": 14720, "lr": 6.585509610150972e-06, "tok_per_s": 871.2019719938318, "loss": 0.6460805404931307, "entropy": 0.6567777544260025, "sharp": 0.1107736174017191} +{"time": "04:17:13", "event": "step", "step": 470, "rows_seen": 15040, "lr": 5.894036597333646e-06, "tok_per_s": 871.8909668828694, "loss": 0.6471943072974682, "entropy": 0.6520841270685196, "sharp": 0.12355391774326563} +{"time": "04:20:53", "event": "step", "step": 480, "rows_seen": 15360, "lr": 5.398045191744809e-06, "tok_per_s": 872.1641297327224, "loss": 0.6499506458640099, "entropy": 0.6403382867574692, "sharp": 0.13805944379419088} +{"time": "04:24:34", "event": "step", "step": 490, "rows_seen": 15680, "lr": 5.099615753708697e-06, "tok_per_s": 872.7285434720358, "loss": 0.6431107819080353, "entropy": 0.6519273668527603, "sharp": 0.11090059950947762} +{"time": "04:28:13", "event": "step", "step": 500, "rows_seen": 16000, "lr": 5e-06, "tok_per_s": 873.1511200997285, "loss": 0.6419705078005791, "entropy": 0.6479619462043047, "sharp": 0.12581786420196295} +{"time": "04:29:36", "event": "val", "step": 500, "rows_seen": 16000, "n": 500, "dropped_long": 0, "noul_ce": 0.5854170918464661, "noul_T": 0.7670265124397351, "noul_auc": 0.7415402271691616, "noul_acc": 0.712, "noul_f1": 0.7714285714285715, "noul_nll": 0.5860456532686948, "noul_ece": 0.062291146516799935, "noul_brier": 0.19968873407987509, "noul_sharp": 0.15745879709720612, "entropy": 0.6191316246986389} +{"time": "04:32:35", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "replay_seen": 842, "step": 500, "tokens": 9820167, "tok_per_s": 853.3575821037343, "rows_per_s": 1.4635441932658964, "val": {"n": 1200, "dropped_long": 0, "noul_ce": 0.5706181526184082, "noul_T": 0.7800532877395329, "choice_ce": 0.672325849533081, "choice_T": 1.0466806238139477, "score_ce": 1.0526014566421509, "score_T": 0.9804041633946682, "noul_auc": 0.7283541333333334, "noul_acc": 0.698, "noul_f1": 0.7719033232628398, "noul_nll": 0.5866083004325628, "noul_ece": 0.05212497982382775, "noul_brier": 0.20057906316333285, "noul_sharp": 0.16233906149864197, "entropy": 0.6147943139076233}, "dropped_long": 0, "peak_gib": 18.236003875732422} diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/DONE b/e33/S1-E33-M20P40-n16k-a100-syn/DONE new file mode 100644 index 0000000000000000000000000000000000000000..0f88b50b36f9158fc9bc56036383b9a2e98ab239 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/DONE @@ -0,0 +1 @@ +2026-10-23 17:35:55 \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/README.md b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/README.md new file mode 100644 index 0000000000000000000000000000000000000000..57167079ef3f181aacdbd678a733d71697d18f63 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/README.md @@ -0,0 +1,206 @@ +--- +base_model: /workspace/jev_kt/models/hf/JEV-9B +library_name: peft +tags: +- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **Developed by:** [More Information Needed] +- **Funded by [optional]:** [More Information Needed] +- **Shared by [optional]:** [More Information Needed] +- **Model type:** [More Information Needed] +- **Language(s) (NLP):** [More Information Needed] +- **License:** [More Information Needed] +- **Finetuned from model [optional]:** [More Information Needed] + +### Model Sources [optional] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. + +## How to Get Started with the Model + +Use the code below to get started with the model. + +[More Information Needed] + +## Training Details + +### Training Data + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). + +- **Hardware Type:** [More Information Needed] +- **Hours used:** [More Information Needed] +- **Cloud Provider:** [More Information Needed] +- **Compute Region:** [More Information Needed] +- **Carbon Emitted:** [More Information Needed] + +## Technical Specifications [optional] + +### Model Architecture and Objective + +[More Information Needed] + +### Compute Infrastructure + +[More Information Needed] + +#### Hardware + +[More Information Needed] + +#### Software + +[More Information Needed] + +## Citation [optional] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[More Information Needed] + +## More Information [optional] + +[More Information Needed] + +## Model Card Authors [optional] + +[More Information Needed] + +## Model Card Contact + +[More Information Needed] +### Framework versions + +- PEFT 0.21.2 \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..35977f13782af3f3f9849b035f73d656897582a4 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json @@ -0,0 +1,57 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 32, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 16, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "out_proj", + "in_proj_z", + "down_proj", + "q_proj", + "v_proj", + "gate_proj", + "in_proj_qkv", + "up_proj", + "k_proj", + "o_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c8fa2d98c83b6974db0b4041f5479ecf37f9fdd3 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b06f968a6db3848efc7e61de7ca5803a555d400c261bfb55f4e605bd2cc15e8 +size 160486456 diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d093d03334d1bcd8511a3867adb084f382fd3c5 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json @@ -0,0 +1,58 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": null, + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64.0, + "lora_bias": false, + "lora_dropout": 0.0, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "lm_head", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1ab8fac215db9dc437e80aef316a8c9019574121 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:343b9dba12d72646e7c3cad503764c38b9b9bda2d18b929dffd0341d2715f4dc +size 176641832 diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json new file mode 100644 index 0000000000000000000000000000000000000000..0a9f56c952d2e43b520bbfbe12d2571ee313368f --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json @@ -0,0 +1,99 @@ +{ + "bias": [ + 0.00010125698463525623, + -0.00010150979505851865, + -0.002252081409096718, + -0.0015649446286261082, + -0.0010697965044528246, + 0.0015692494343966246, + 0.0011667461367323995, + -0.0025911687407642603, + -1.5943171092658304e-05, + -0.0004627322778105736, + 0.0002803638344630599, + 0.0005846124258823693, + -0.0008904358255676925, + -0.0034926794469356537, + -0.0042582787573337555, + -0.005071217194199562, + -0.005935357417911291, + -0.004361479543149471, + -0.004931907169520855, + -0.004948125220835209, + -0.00379412854090333, + -0.0033585415221750736, + -0.005670496728271246, + -0.0034690620377659798 + ], + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature" +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/calibration.json b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/calibration.json new file mode 100644 index 0000000000000000000000000000000000000000..07bd4319acb463f088807aeb588a92a5a5ade2a8 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/calibration.json @@ -0,0 +1,29 @@ +{ + "version": 1, + "per_kind": { + "noul": 0.8002108628973789, + "choice": 1.0728412258162334, + "score": 0.9491235278105069 + }, + "fit": { + "noul": { + "n": 1087, + "T": 0.8002108628973789, + "ce_before": 0.5691902041435242, + "ce_after": 0.5654259920120239 + }, + "choice": { + "n": 60, + "T": 1.0728412258162334, + "ce_before": 0.6705418825149536, + "ce_after": 0.6693586111068726 + }, + "score": { + "n": 53, + "T": 0.9491235278105069, + "ce_before": 1.0455740690231323, + "ce_after": 1.0447719097137451 + } + }, + "source": "models/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000" +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/head.safetensors b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/head.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..80e4da26708a5582588d8239a87836c8cbeb5a4b --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/head.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e7dd06ef2275896db2a9ad8e41f09f0ef7babb8b31997efa4d1e3e013489b25 +size 393464 diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/judge_config.json b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/judge_config.json new file mode 100644 index 0000000000000000000000000000000000000000..9be296bb70fdf3b7cc0b5206f66d625e96a75ba6 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000/judge_config.json @@ -0,0 +1,164 @@ +{ + "base_model_path": "/root/models/Qwen3.5-9B", + "hidden_size": 4096, + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "kinds": [ + "noul", + "choice", + "score" + ], + "model_name": "jev-judge-qwen35-9b", + "model_version": "0.8.0", + "lm_head_included": true, + "weights_mode": "unmerged", + "adapter_subfolder": "adapter", + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ] + }, + "trained_stage": "e16", + "source_checkpoint": "models/e33/S1-E33-M20P40-n16k-a100-syn/ckpt-16000", + "softcap": null, + "e16": { + "data": "data/e33/E33-M20P40-n16k-a100", + "data_manifest": { + "arm": "E33-M20P40-n16k-a100", + "name": null, + "train_sha256": "4c2cbc7e61a30f31b9bd5413d5605555b3ac9485e7e3cb9e113abd1aa1394b12", + "n_train": 16000 + }, + "init": "released", + "objective": "s-log", + "sigma": 0.5, + "K": 8, + "reward": "brier", + "clip": 0.2, + "passes": 1, + "lr": 0.0001, + "lr_head": 0.0002, + "rows_per_step": 32, + "train_seed": 0, + "replay": "data/e16/replay/train.jsonl", + "replay_frac": 0.05, + "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", + "max_len": 3072, + "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", + "total_rows": 16000, + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05 + }, + "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", + "event": "ckpt", + "tag": "ckpt-16000", + "rows_seen": 16000, + "replay_seen": 842, + "step": 500, + "tokens": 9814579, + "tok_per_s": 806.9924313028692, + "rows_per_s": 1.384814012135178, + "val": { + "n": 1200, + "dropped_long": 0, + "noul_ce": 0.5691902041435242, + "noul_T": 0.8002108628973789, + "choice_ce": 0.6705418825149536, + "choice_T": 1.0728412258162334, + "score_ce": 1.0455740690231323, + "score_T": 0.9491235278105069, + "noul_auc": 0.7222741333333333, + "noul_acc": 0.698, + "noul_f1": 0.7739520958083832, + "noul_nll": 0.5880829917564988, + "noul_ece": 0.04662516665458681, + "noul_brier": 0.20118704083871727, + "noul_sharp": 0.1676119565963745, + "entropy": 0.6107596158981323 + }, + "dropped_long": 0, + "peak_gib": 18.236003875732422 + } +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-a100-syn/train_log.jsonl b/e33/S1-E33-M20P40-n16k-a100-syn/train_log.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..5bf71daf9daa93aea3e8eccd95496ad59485f5a2 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-a100-syn/train_log.jsonl @@ -0,0 +1,57 @@ +{"time": "14:13:08", "event": "start", "data": "data/e33/E33-M20P40-n16k-a100", "data_manifest": {"arm": "E33-M20P40-n16k-a100", "name": null, "train_sha256": "4c2cbc7e61a30f31b9bd5413d5605555b3ac9485e7e3cb9e113abd1aa1394b12", "n_train": 16000}, "init": "released", "objective": "s-log", "sigma": 0.5, "K": 8, "reward": "brier", "clip": 0.2, "passes": 1, "lr": 0.0001, "lr_head": 0.0002, "rows_per_step": 32, "train_seed": 0, "replay": "data/e16/replay/train.jsonl", "replay_frac": 0.05, "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", "max_len": 3072, "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", "total_rows": 16000, "lora": {"r": 16, "lora_alpha": 32, "lora_dropout": 0.05}, "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", "device": "cuda", "load_s": 120.29126191139221} +{"time": "14:16:54", "event": "step", "step": 10, "rows_seen": 320, "lr": 7.333333333333333e-05, "tok_per_s": 839.8715478391055, "loss": 0.6187404282391071, "entropy": 0.6414551772177219, "sharp": 0.13162385113537312} +{"time": "14:20:44", "event": "step", "step": 20, "rows_seen": 640, "lr": 9.997508952966197e-05, "tok_per_s": 847.5550024837847, "loss": 0.710713678970933, "entropy": 0.6639010868966579, "sharp": 0.094074378721416} +{"time": "14:24:42", "event": "step", "step": 30, "rows_seen": 960, "lr": 9.977596250541417e-05, "tok_per_s": 846.8465414505208, "loss": 0.7149962484836578, "entropy": 0.6894855201244354, "sharp": 0.038734118454158306} +{"time": "14:28:40", "event": "step", "step": 40, "rows_seen": 1280, "lr": 9.937854366485982e-05, "tok_per_s": 847.0194710535048, "loss": 0.7016126066446304, "entropy": 0.6360101252794266, "sharp": 0.15717380121350288} +{"time": "14:32:21", "event": "step", "step": 50, "rows_seen": 1600, "lr": 9.8784499920729e-05, "tok_per_s": 846.5476430796652, "loss": 0.7025396898388863, "entropy": 0.6633415594696999, "sharp": 0.09588960744440556} +{"time": "14:36:18", "event": "step", "step": 60, "rows_seen": 1920, "lr": 9.79963228989294e-05, "tok_per_s": 847.856212094996, "loss": 0.6576686762273312, "entropy": 0.5767230838537216, "sharp": 0.2077706828713417} +{"time": "14:40:04", "event": "step", "step": 70, "rows_seen": 2240, "lr": 9.701731848780134e-05, "tok_per_s": 849.3255868206792, "loss": 0.7632252350449562, "entropy": 0.6031612306833267, "sharp": 0.1766234626993537} +{"time": "14:43:56", "event": "step", "step": 80, "rows_seen": 2560, "lr": 9.585159297207315e-05, "tok_per_s": 848.8644693577511, "loss": 0.562406275421381, "entropy": 0.5192938018590212, "sharp": 0.2664580028504133} +{"time": "14:47:44", "event": "step", "step": 90, "rows_seen": 2880, "lr": 9.450403580967612e-05, "tok_per_s": 847.7852483871987, "loss": 0.677379759028554, "entropy": 0.6627754978835583, "sharp": 0.10070306714624166} +{"time": "14:51:39", "event": "step", "step": 100, "rows_seen": 3200, "lr": 9.298029912365855e-05, "tok_per_s": 847.9684956914746, "loss": 0.6650274395942688, "entropy": 0.6410431265830994, "sharp": 0.14800265431404114} +{"time": "14:53:08", "event": "val", "step": 100, "rows_seen": 3200, "n": 500, "dropped_long": 0, "noul_ce": 0.6212870478630066, "noul_T": 0.7023261478585969, "noul_auc": 0.6843260029366593, "noul_acc": 0.672, "noul_f1": 0.7759562841530054, "noul_nll": 0.6250243791341782, "noul_ece": 0.11322395074367522, "noul_brier": 0.21671475337394724, "noul_sharp": 0.12332789599895477, "entropy": 0.6514794230461121} +{"time": "14:56:48", "event": "step", "step": 110, "rows_seen": 3520, "lr": 9.128677399521752e-05, "tok_per_s": 818.6033910951888, "loss": 0.6246248260140419, "entropy": 0.604729138314724, "sharp": 0.19723656587302685} +{"time": "15:00:34", "event": "step", "step": 120, "rows_seen": 3840, "lr": 8.943056365728335e-05, "tok_per_s": 820.7401742662969, "loss": 0.7470029867254198, "entropy": 0.5473670903593302, "sharp": 0.24039341136813164} +{"time": "15:04:22", "event": "step", "step": 130, "rows_seen": 4160, "lr": 8.741945370109202e-05, "tok_per_s": 822.128574168893, "loss": 0.6280730292201042, "entropy": 0.6206810176372528, "sharp": 0.16349924262613058} +{"time": "15:08:05", "event": "step", "step": 140, "rows_seen": 4480, "lr": 8.526187942070997e-05, "tok_per_s": 823.5858737424818, "loss": 0.6877431944012642, "entropy": 0.6785900443792343, "sharp": 0.07349966000765562} +{"time": "15:11:57", "event": "step", "step": 150, "rows_seen": 4800, "lr": 8.296689043247915e-05, "tok_per_s": 825.5068292890887, "loss": 0.7086129002273083, "entropy": 0.6279721036553383, "sharp": 0.14614077657461166} +{"time": "15:15:50", "event": "step", "step": 160, "rows_seen": 5120, "lr": 8.054411271778132e-05, "tok_per_s": 827.0023994783346, "loss": 0.7685284856706858, "entropy": 0.6500500217080116, "sharp": 0.1163442237302661} +{"time": "15:19:40", "event": "step", "step": 170, "rows_seen": 5440, "lr": 7.800370824832674e-05, "tok_per_s": 828.4397478955465, "loss": 0.754140667617321, "entropy": 0.6031205579638481, "sharp": 0.1930545661598444} +{"time": "15:23:33", "event": "step", "step": 180, "rows_seen": 5760, "lr": 7.535633236331324e-05, "tok_per_s": 829.5254864991339, "loss": 0.6428478360176086, "entropy": 0.6708729267120361, "sharp": 0.08859098423272371} +{"time": "15:27:20", "event": "step", "step": 190, "rows_seen": 6080, "lr": 7.26130890772308e-05, "tok_per_s": 830.864614360502, "loss": 0.5738391559571028, "entropy": 0.6217141449451447, "sharp": 0.15805615298449993} +{"time": "15:31:14", "event": "step", "step": 200, "rows_seen": 6400, "lr": 6.978548450576568e-05, "tok_per_s": 831.7710687381879, "loss": 0.6900174766778946, "entropy": 0.6694989874958992, "sharp": 0.0823324890807271} +{"time": "15:32:43", "event": "val", "step": 200, "rows_seen": 6400, "n": 500, "dropped_long": 0, "noul_ce": 0.6128379106521606, "noul_T": 0.47963188571230614, "noul_auc": 0.7305193161296856, "noul_acc": 0.7, "noul_f1": 0.7572815533980582, "noul_nll": 0.6143106191158295, "noul_ece": 0.1173224647641182, "noul_brier": 0.21178017520018558, "noul_sharp": 0.10002126544713974, "entropy": 0.6633838415145874} +{"time": "15:36:35", "event": "step", "step": 210, "rows_seen": 6720, "lr": 6.688537860515266e-05, "tok_per_s": 818.3244384803786, "loss": 0.6299922671169043, "entropy": 0.6626281999051571, "sharp": 0.10336934588849545} +{"time": "15:40:22", "event": "step", "step": 220, "rows_seen": 7040, "lr": 6.392493542739706e-05, "tok_per_s": 819.8278409393922, "loss": 0.627080044709146, "entropy": 0.6587131060659885, "sharp": 0.11120752524584532} +{"time": "15:44:19", "event": "step", "step": 230, "rows_seen": 7360, "lr": 6.091657210001309e-05, "tok_per_s": 820.8217398785365, "loss": 0.5289444080553949, "entropy": 0.6182603435590863, "sharp": 0.16458167880773544} +{"time": "15:48:16", "event": "step", "step": 240, "rows_seen": 7680, "lr": 5.7872906744276015e-05, "tok_per_s": 821.7753717182602, "loss": 0.668922908604145, "entropy": 0.5956772342324257, "sharp": 0.19893265888094902} +{"time": "15:52:13", "event": "step", "step": 250, "rows_seen": 8000, "lr": 5.48067055504363e-05, "tok_per_s": 822.6517675023122, "loss": 0.6773939281702042, "entropy": 0.6701566651463509, "sharp": 0.08436406124383211} +{"time": "15:55:57", "event": "step", "step": 260, "rows_seen": 8320, "lr": 5.173082923188119e-05, "tok_per_s": 823.7591338250224, "loss": 0.6251176791265607, "entropy": 0.6630509197711945, "sharp": 0.10360329411923885} +{"time": "15:59:46", "event": "step", "step": 270, "rows_seen": 8640, "lr": 4.865817908283331e-05, "tok_per_s": 825.0068217712911, "loss": 0.698808241635561, "entropy": 0.5815245062112808, "sharp": 0.21828553080558777} +{"time": "16:03:35", "event": "step", "step": 280, "rows_seen": 8960, "lr": 4.560164286583922e-05, "tok_per_s": 826.2237881602903, "loss": 0.7049896717071533, "entropy": 0.654325544834137, "sharp": 0.09251557663083076} +{"time": "16:07:26", "event": "step", "step": 290, "rows_seen": 9280, "lr": 4.257404075601443e-05, "tok_per_s": 827.1292792775448, "loss": 0.688546247780323, "entropy": 0.6717874817550182, "sharp": 0.08574923686683178} +{"time": "16:11:15", "event": "step", "step": 300, "rows_seen": 9600, "lr": 3.958807156877371e-05, "tok_per_s": 827.914890733714, "loss": 0.49946608766913414, "entropy": 0.5843364968895912, "sharp": 0.20784184522926807} +{"time": "16:12:44", "event": "val", "step": 300, "rows_seen": 9600, "n": 500, "dropped_long": 0, "noul_ce": 0.605241060256958, "noul_T": 1.2048853041749858, "noul_auc": 0.7253042142748646, "noul_acc": 0.636, "noul_f1": 0.7672634271099744, "noul_nll": 0.6115475974157453, "noul_ece": 0.13372548961639405, "noul_brier": 0.21162126575926385, "noul_sharp": 0.2091650366783142, "entropy": 0.5707269906997681} +{"time": "16:16:32", "event": "step", "step": 310, "rows_seen": 9920, "lr": 3.665625949658603e-05, "tok_per_s": 818.8158304626734, "loss": 0.5467460304498672, "entropy": 0.5993260815739632, "sharp": 0.18954907543957233} +{"time": "16:20:26", "event": "step", "step": 320, "rows_seen": 10240, "lr": 3.379090157815887e-05, "tok_per_s": 820.0792956300551, "loss": 0.6160920187830925, "entropy": 0.6452987492084503, "sharp": 0.13316628709435463} +{"time": "16:24:14", "event": "step", "step": 330, "rows_seen": 10560, "lr": 3.10040161203844e-05, "tok_per_s": 821.1997374291061, "loss": 0.5811580680310726, "entropy": 0.6088825017213821, "sharp": 0.16268229950219393} +{"time": "16:28:05", "event": "step", "step": 340, "rows_seen": 10880, "lr": 2.8307292289383776e-05, "tok_per_s": 821.9665063572719, "loss": 0.6118759140372276, "entropy": 0.6484171822667122, "sharp": 0.12860039994120598} +{"time": "16:31:58", "event": "step", "step": 350, "rows_seen": 11200, "lr": 2.571204108208194e-05, "tok_per_s": 822.97676320785, "loss": 0.7243249639868736, "entropy": 0.6476022303104401, "sharp": 0.1273578805848956} +{"time": "16:35:44", "event": "step", "step": 360, "rows_seen": 11520, "lr": 2.322914788395595e-05, "tok_per_s": 823.7959610682416, "loss": 0.6667574434541166, "entropy": 0.6401418969035149, "sharp": 0.11859891004860401} +{"time": "16:39:31", "event": "step", "step": 370, "rows_seen": 11840, "lr": 2.0869026811945157e-05, "tok_per_s": 824.6030474215052, "loss": 0.6770539954304695, "entropy": 0.6548423916101456, "sharp": 0.11251895409077406} +{"time": "16:43:23", "event": "step", "step": 380, "rows_seen": 12160, "lr": 1.8641577034025627e-05, "tok_per_s": 825.3579685095436, "loss": 0.683322167955339, "entropy": 0.6304250992834568, "sharp": 0.15285257250070572} +{"time": "16:47:17", "event": "step", "step": 390, "rows_seen": 12480, "lr": 1.6556141248659472e-05, "tok_per_s": 825.9141744084313, "loss": 0.5915361074730754, "entropy": 0.6303008385002613, "sharp": 0.1420151386409998} +{"time": "16:51:08", "event": "step", "step": 400, "rows_seen": 12800, "lr": 1.4621466498270773e-05, "tok_per_s": 826.6632978600414, "loss": 0.5277743395417929, "entropy": 0.637439914047718, "sharp": 0.13507275190204382} +{"time": "16:52:37", "event": "val", "step": 400, "rows_seen": 12800, "n": 500, "dropped_long": 0, "noul_ce": 0.5858119130134583, "noul_T": 0.7325064249704112, "noul_auc": 0.7344348618588715, "noul_acc": 0.702, "noul_f1": 0.7732115677321156, "noul_nll": 0.5899210872650147, "noul_ece": 0.06367846584320071, "noul_brier": 0.20153352968871333, "noul_sharp": 0.14930002391338348, "entropy": 0.6234035491943359} +{"time": "16:56:25", "event": "step", "step": 410, "rows_seen": 13120, "lr": 1.284566748110963e-05, "tok_per_s": 819.9214470641089, "loss": 0.589809961616993, "entropy": 0.6252590082585812, "sharp": 0.14650190621614456} +{"time": "17:00:14", "event": "step", "step": 420, "rows_seen": 13440, "lr": 1.1236192515387821e-05, "tok_per_s": 820.7593532438311, "loss": 0.5925923585891724, "entropy": 0.6398867517709732, "sharp": 0.11890413705259562} +{"time": "17:04:05", "event": "step", "step": 430, "rows_seen": 13760, "lr": 9.799792298443728e-06, "tok_per_s": 821.6220787132297, "loss": 0.6462609097361565, "entropy": 0.6125414073467255, "sharp": 0.16461621969938278} +{"time": "17:07:57", "event": "step", "step": 440, "rows_seen": 14080, "lr": 8.542491591971554e-06, "tok_per_s": 822.3114539874771, "loss": 0.6334951892495155, "entropy": 0.6172429770231247, "sharp": 0.15678761340677738} +{"time": "17:11:47", "event": "step", "step": 450, "rows_seen": 14400, "lr": 7.469563952077155e-06, "tok_per_s": 823.0525671531353, "loss": 0.7070290148258209, "entropy": 0.6357797607779503, "sharp": 0.1409710068255663} +{"time": "17:15:34", "event": "step", "step": 460, "rows_seen": 14720, "lr": 6.585509610150972e-06, "tok_per_s": 823.6646668729059, "loss": 0.6082311794161797, "entropy": 0.6516525633633137, "sharp": 0.1177465096116066} +{"time": "17:19:30", "event": "step", "step": 470, "rows_seen": 15040, "lr": 5.894036597333646e-06, "tok_per_s": 824.2039480742526, "loss": 0.633880028501153, "entropy": 0.6386780217289925, "sharp": 0.13634685520082712} +{"time": "17:23:27", "event": "step", "step": 480, "rows_seen": 15360, "lr": 5.398045191744809e-06, "tok_per_s": 824.8682535596918, "loss": 0.5927850492298603, "entropy": 0.6280360221862793, "sharp": 0.1544718537479639} +{"time": "17:27:15", "event": "step", "step": 490, "rows_seen": 15680, "lr": 5.099615753708697e-06, "tok_per_s": 825.4752101102926, "loss": 0.6072592251002789, "entropy": 0.6435235626995564, "sharp": 0.12449884880334139} +{"time": "17:31:12", "event": "step", "step": 500, "rows_seen": 16000, "lr": 5e-06, "tok_per_s": 825.8721817850699, "loss": 0.5903935097157955, "entropy": 0.6376286670565605, "sharp": 0.14049519877880812} +{"time": "17:32:41", "event": "val", "step": 500, "rows_seen": 16000, "n": 500, "dropped_long": 0, "noul_ce": 0.5839874148368835, "noul_T": 0.7941656377356624, "noul_auc": 0.7346711447908051, "noul_acc": 0.706, "noul_f1": 0.7699530516431925, "noul_nll": 0.5867154191732407, "noul_ece": 0.0668563278913498, "noul_brier": 0.2000565286770292, "noul_sharp": 0.1619034707546234, "entropy": 0.6146537661552429} +{"time": "17:35:55", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "replay_seen": 842, "step": 500, "tokens": 9814579, "tok_per_s": 806.9924313028692, "rows_per_s": 1.384814012135178, "val": {"n": 1200, "dropped_long": 0, "noul_ce": 0.5691902041435242, "noul_T": 0.8002108628973789, "choice_ce": 0.6705418825149536, "choice_T": 1.0728412258162334, "score_ce": 1.0455740690231323, "score_T": 0.9491235278105069, "noul_auc": 0.7222741333333333, "noul_acc": 0.698, "noul_f1": 0.7739520958083832, "noul_nll": 0.5880829917564988, "noul_ece": 0.04662516665458681, "noul_brier": 0.20118704083871727, "noul_sharp": 0.1676119565963745, "entropy": 0.6107596158981323}, "dropped_long": 0, "peak_gib": 18.236003875732422} diff --git a/e33/S1-E33-M20P40-n16k-syn/DONE b/e33/S1-E33-M20P40-n16k-syn/DONE new file mode 100644 index 0000000000000000000000000000000000000000..edbf1ac57713212dd1c79ea60203e2fab1be72de --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/DONE @@ -0,0 +1 @@ +2026-10-23 11:35:25 \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/README.md b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/README.md new file mode 100644 index 0000000000000000000000000000000000000000..57167079ef3f181aacdbd678a733d71697d18f63 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/README.md @@ -0,0 +1,206 @@ +--- +base_model: /workspace/jev_kt/models/hf/JEV-9B +library_name: peft +tags: +- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **Developed by:** [More Information Needed] +- **Funded by [optional]:** [More Information Needed] +- **Shared by [optional]:** [More Information Needed] +- **Model type:** [More Information Needed] +- **Language(s) (NLP):** [More Information Needed] +- **License:** [More Information Needed] +- **Finetuned from model [optional]:** [More Information Needed] + +### Model Sources [optional] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. + +## How to Get Started with the Model + +Use the code below to get started with the model. + +[More Information Needed] + +## Training Details + +### Training Data + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). + +- **Hardware Type:** [More Information Needed] +- **Hours used:** [More Information Needed] +- **Cloud Provider:** [More Information Needed] +- **Compute Region:** [More Information Needed] +- **Carbon Emitted:** [More Information Needed] + +## Technical Specifications [optional] + +### Model Architecture and Objective + +[More Information Needed] + +### Compute Infrastructure + +[More Information Needed] + +#### Hardware + +[More Information Needed] + +#### Software + +[More Information Needed] + +## Citation [optional] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[More Information Needed] + +## More Information [optional] + +[More Information Needed] + +## Model Card Authors [optional] + +[More Information Needed] + +## Model Card Contact + +[More Information Needed] +### Framework versions + +- PEFT 0.21.2 \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/adapter_config.json b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a9b85cb7ab38f99dfec817ce989ca40bd46b3cd6 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/adapter_config.json @@ -0,0 +1,57 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 32, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 16, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "out_proj", + "v_proj", + "in_proj_qkv", + "up_proj", + "in_proj_z", + "down_proj", + "gate_proj", + "o_proj", + "q_proj", + "k_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..81a79af0d147f4537504a2943c433d28aa4ac6ea --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1998c229ad01987725960ed3c5763f1da62efb5d123aeb99b59b9792b7010bce +size 160486456 diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d093d03334d1bcd8511a3867adb084f382fd3c5 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json @@ -0,0 +1,58 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": null, + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64.0, + "lora_bias": false, + "lora_dropout": 0.0, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "lm_head", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0c92c9c802c5d0cda298062cf021eef7bf1181d1 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe86a6ba16b8fe092a27422e15e8cbd5f782a3055c61c93e1fb7faf2660766fc +size 176641832 diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json new file mode 100644 index 0000000000000000000000000000000000000000..dc91c6eb66f068ee65817ce10bfe89f18c037fe3 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json @@ -0,0 +1,99 @@ +{ + "bias": [ + 0.00027095075347460806, + -0.0002712052664719522, + -0.00215821317397058, + -0.0012642444344237447, + -0.00037909665843471885, + 0.0022430443204939365, + 0.0014937476953491569, + -0.0026326365768909454, + -0.0004677560064010322, + -0.00013563339598476887, + 0.0005219306913204491, + 0.0010486760875210166, + 0.0001400769833708182, + -0.0034511531703174114, + -0.004273152444511652, + -0.005068799946457148, + -0.005938724149018526, + -0.004366480745375156, + -0.004930523689836264, + -0.004939631093293428, + -0.00384484208188951, + -0.00335189257748425, + -0.0056142439134418964, + -0.003465337445959449 + ], + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature" +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/calibration.json b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/calibration.json new file mode 100644 index 0000000000000000000000000000000000000000..7a03c9d277ae883ceb924fa163a066a6ca8d6d41 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/calibration.json @@ -0,0 +1,29 @@ +{ + "version": 1, + "per_kind": { + "noul": 0.7536078969778258, + "choice": 1.0369940375159572, + "score": 0.9641858015101192 + }, + "fit": { + "noul": { + "n": 1087, + "T": 0.7536078969778258, + "ce_before": 0.5733757615089417, + "ce_after": 0.5677251815795898 + }, + "choice": { + "n": 60, + "T": 1.0369940375159572, + "ce_before": 0.6717521548271179, + "ce_after": 0.6714757680892944 + }, + "score": { + "n": 53, + "T": 0.9641858015101192, + "ce_before": 1.0550035238265991, + "ce_after": 1.0546045303344727 + } + }, + "source": "models/e33/S1-E33-M20P40-n16k-syn/ckpt-16000" +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/head.safetensors b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/head.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2e29a105fb1ac687b16e35c45f77c27813728c48 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/head.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:225c17a0e7d3ab4fb832923dcfcc3ccfd1b9835dde3b2f1198f257d7cd21c78e +size 393464 diff --git a/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/judge_config.json b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/judge_config.json new file mode 100644 index 0000000000000000000000000000000000000000..9ecf4bdbde54e03dc7b5abad3414c79d19eeb5a3 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/ckpt-16000/judge_config.json @@ -0,0 +1,164 @@ +{ + "base_model_path": "/root/models/Qwen3.5-9B", + "hidden_size": 4096, + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "kinds": [ + "noul", + "choice", + "score" + ], + "model_name": "jev-judge-qwen35-9b", + "model_version": "0.8.0", + "lm_head_included": true, + "weights_mode": "unmerged", + "adapter_subfolder": "adapter", + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ] + }, + "trained_stage": "e16", + "source_checkpoint": "models/e33/S1-E33-M20P40-n16k-syn/ckpt-16000", + "softcap": null, + "e16": { + "data": "data/e33/E33-M20P40-n16k", + "data_manifest": { + "arm": "E33-M20P40-n16k", + "name": null, + "train_sha256": "0c998c032f8987adac4577ffcddbc6a65fc3ba6b81ba469c1767c02644b903fb", + "n_train": 16000 + }, + "init": "released", + "objective": "s-log", + "sigma": 0.5, + "K": 8, + "reward": "brier", + "clip": 0.2, + "passes": 1, + "lr": 0.0001, + "lr_head": 0.0002, + "rows_per_step": 32, + "train_seed": 0, + "replay": "data/e16/replay/train.jsonl", + "replay_frac": 0.05, + "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", + "max_len": 3072, + "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", + "total_rows": 16000, + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05 + }, + "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", + "event": "ckpt", + "tag": "ckpt-16000", + "rows_seen": 16000, + "replay_seen": 842, + "step": 500, + "tokens": 9814579, + "tok_per_s": 808.7513072822612, + "rows_per_s": 1.3878322760350466, + "val": { + "n": 1200, + "dropped_long": 0, + "noul_ce": 0.5733757615089417, + "noul_T": 0.7536078969778258, + "choice_ce": 0.6717521548271179, + "choice_T": 1.0369940375159572, + "score_ce": 1.0550035238265991, + "score_T": 0.9641858015101192, + "noul_auc": 0.7241642666666666, + "noul_acc": 0.685, + "noul_f1": 0.7608200455580866, + "noul_nll": 0.590823968693614, + "noul_ece": 0.05668055501580238, + "noul_brier": 0.2023056829963619, + "noul_sharp": 0.15657421946525574, + "entropy": 0.6214356422424316 + }, + "dropped_long": 0, + "peak_gib": 18.236003875732422 + } +} \ No newline at end of file diff --git a/e33/S1-E33-M20P40-n16k-syn/train_log.jsonl b/e33/S1-E33-M20P40-n16k-syn/train_log.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..f1eca6ffb09e3c401c567d46eba9fc42a5c301a2 --- /dev/null +++ b/e33/S1-E33-M20P40-n16k-syn/train_log.jsonl @@ -0,0 +1,57 @@ +{"time": "08:13:07", "event": "start", "data": "data/e33/E33-M20P40-n16k", "data_manifest": {"arm": "E33-M20P40-n16k", "name": null, "train_sha256": "0c998c032f8987adac4577ffcddbc6a65fc3ba6b81ba469c1767c02644b903fb", "n_train": 16000}, "init": "released", "objective": "s-log", "sigma": 0.5, "K": 8, "reward": "brier", "clip": 0.2, "passes": 1, "lr": 0.0001, "lr_head": 0.0002, "rows_per_step": 32, "train_seed": 0, "replay": "data/e16/replay/train.jsonl", "replay_frac": 0.05, "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", "max_len": 3072, "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", "total_rows": 16000, "lora": {"r": 16, "lora_alpha": 32, "lora_dropout": 0.05}, "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", "device": "cuda", "load_s": 186.75728130340576} +{"time": "08:16:55", "event": "step", "step": 10, "rows_seen": 320, "lr": 7.333333333333333e-05, "tok_per_s": 834.6921166229316, "loss": 0.6334675438702106, "entropy": 0.6528529115021229, "sharp": 0.11347178183495998} +{"time": "08:20:44", "event": "step", "step": 20, "rows_seen": 640, "lr": 9.997508952966197e-05, "tok_per_s": 845.8693691529938, "loss": 0.6971936989575624, "entropy": 0.6627442203462124, "sharp": 0.10100687853991985} +{"time": "08:24:40", "event": "step", "step": 30, "rows_seen": 960, "lr": 9.977596250541417e-05, "tok_per_s": 848.5050223983193, "loss": 0.7298796325922012, "entropy": 0.6808525398373604, "sharp": 0.06191548611968756} +{"time": "08:28:35", "event": "step", "step": 40, "rows_seen": 1280, "lr": 9.937854366485982e-05, "tok_per_s": 850.7955858157326, "loss": 0.7005446255207062, "entropy": 0.639897458255291, "sharp": 0.1383969560265541} +{"time": "08:32:16", "event": "step", "step": 50, "rows_seen": 1600, "lr": 9.8784499920729e-05, "tok_per_s": 850.5229789882324, "loss": 0.6737376227974892, "entropy": 0.6729620024561882, "sharp": 0.08500486146658659} +{"time": "08:36:13", "event": "step", "step": 60, "rows_seen": 1920, "lr": 9.79963228989294e-05, "tok_per_s": 850.9770813230041, "loss": 0.6547166109085083, "entropy": 0.643814604729414, "sharp": 0.12674631737172604} +{"time": "08:40:00", "event": "step", "step": 70, "rows_seen": 2240, "lr": 9.701731848780134e-05, "tok_per_s": 851.2325512337642, "loss": 0.6519396528601646, "entropy": 0.6633624210953712, "sharp": 0.10033598821610212} +{"time": "08:43:52", "event": "step", "step": 80, "rows_seen": 2560, "lr": 9.585159297207315e-05, "tok_per_s": 850.5616989075138, "loss": 0.6236812435090542, "entropy": 0.6104888617992401, "sharp": 0.17594678606837988} +{"time": "08:47:39", "event": "step", "step": 90, "rows_seen": 2880, "lr": 9.450403580967612e-05, "tok_per_s": 849.9966186161708, "loss": 0.6535216812044382, "entropy": 0.6427956651896238, "sharp": 0.13958603981882334} +{"time": "08:51:31", "event": "step", "step": 100, "rows_seen": 3200, "lr": 9.298029912365855e-05, "tok_per_s": 850.7578188501235, "loss": 0.6206827238202095, "entropy": 0.6642791256308556, "sharp": 0.09940087143331766} +{"time": "08:53:00", "event": "val", "step": 100, "rows_seen": 3200, "n": 500, "dropped_long": 0, "noul_ce": 0.6004648208618164, "noul_T": 0.7332869469703953, "noul_auc": 0.720477291522506, "noul_acc": 0.708, "noul_f1": 0.7739938080495357, "noul_nll": 0.6005643621385097, "noul_ece": 0.05962640810012819, "noul_brier": 0.20563077495026427, "noul_sharp": 0.14495421946048737, "entropy": 0.6349033117294312} +{"time": "08:56:37", "event": "step", "step": 110, "rows_seen": 3520, "lr": 9.128677399521752e-05, "tok_per_s": 821.824708277265, "loss": 0.6502102166414261, "entropy": 0.6602606698870659, "sharp": 0.10468737687915564} +{"time": "09:00:21", "event": "step", "step": 120, "rows_seen": 3840, "lr": 8.943056365728335e-05, "tok_per_s": 824.2019704334859, "loss": 0.7216617409139872, "entropy": 0.5636089574545622, "sharp": 0.220951359719038} +{"time": "09:04:07", "event": "step", "step": 130, "rows_seen": 4160, "lr": 8.741945370109202e-05, "tok_per_s": 826.0923282324308, "loss": 0.6275180354714394, "entropy": 0.6667738035321236, "sharp": 0.09454494621604681} +{"time": "09:07:50", "event": "step", "step": 140, "rows_seen": 4480, "lr": 8.526187942070997e-05, "tok_per_s": 827.2654550017238, "loss": 0.67619464173913, "entropy": 0.6331211924552917, "sharp": 0.14430847484618425} +{"time": "09:11:43", "event": "step", "step": 150, "rows_seen": 4800, "lr": 8.296689043247915e-05, "tok_per_s": 828.795658712477, "loss": 0.6780807822942734, "entropy": 0.6458774507045746, "sharp": 0.13158006686717272} +{"time": "09:15:35", "event": "step", "step": 160, "rows_seen": 5120, "lr": 8.054411271778132e-05, "tok_per_s": 830.1566643961453, "loss": 0.7391501776874065, "entropy": 0.6528471745550632, "sharp": 0.11345870047807693} +{"time": "09:19:25", "event": "step", "step": 170, "rows_seen": 5440, "lr": 7.800370824832674e-05, "tok_per_s": 831.4080472090963, "loss": 0.6620825305581093, "entropy": 0.6246006488800049, "sharp": 0.16433628741651773} +{"time": "09:23:19", "event": "step", "step": 180, "rows_seen": 5760, "lr": 7.535633236331324e-05, "tok_per_s": 832.2049488245029, "loss": 0.6418737918138504, "entropy": 0.6729150265455246, "sharp": 0.07942454144358635} +{"time": "09:27:07", "event": "step", "step": 190, "rows_seen": 6080, "lr": 7.26130890772308e-05, "tok_per_s": 833.2000802393027, "loss": 0.6191496253013611, "entropy": 0.6573431305587292, "sharp": 0.11025342717766762} +{"time": "09:30:59", "event": "step", "step": 200, "rows_seen": 6400, "lr": 6.978548450576568e-05, "tok_per_s": 834.2327115681361, "loss": 0.6632450968027115, "entropy": 0.6655780747532845, "sharp": 0.08871013019233942} +{"time": "09:32:28", "event": "val", "step": 200, "rows_seen": 6400, "n": 500, "dropped_long": 0, "noul_ce": 0.598476767539978, "noul_T": 0.7084120180784974, "noul_auc": 0.7335741168925418, "noul_acc": 0.692, "noul_f1": 0.7441860465116279, "noul_nll": 0.5992745052427053, "noul_ece": 0.0875435711145401, "noul_brier": 0.2056747248190986, "noul_sharp": 0.1401194930076599, "entropy": 0.6354633569717407} +{"time": "09:36:20", "event": "step", "step": 210, "rows_seen": 6720, "lr": 6.688537860515266e-05, "tok_per_s": 820.6502771891819, "loss": 0.6477267909795046, "entropy": 0.66724544018507, "sharp": 0.09406618867069483} +{"time": "09:40:07", "event": "step", "step": 220, "rows_seen": 7040, "lr": 6.392493542739706e-05, "tok_per_s": 822.0828850419384, "loss": 0.6350954649969935, "entropy": 0.645612420514226, "sharp": 0.13076961133629084} +{"time": "09:44:04", "event": "step", "step": 230, "rows_seen": 7360, "lr": 6.091657210001309e-05, "tok_per_s": 823.0521579520484, "loss": 0.6120151542127132, "entropy": 0.634583511389792, "sharp": 0.14943482168018818} +{"time": "09:48:00", "event": "step", "step": 240, "rows_seen": 7680, "lr": 5.7872906744276015e-05, "tok_per_s": 824.0820127062603, "loss": 0.6576330363750458, "entropy": 0.6363804861903191, "sharp": 0.1506327036768198} +{"time": "09:51:54", "event": "step", "step": 250, "rows_seen": 8000, "lr": 5.48067055504363e-05, "tok_per_s": 825.1990614839201, "loss": 0.6741576343774796, "entropy": 0.6751580238342285, "sharp": 0.07021459378302097} +{"time": "09:55:37", "event": "step", "step": 260, "rows_seen": 8320, "lr": 5.173082923188119e-05, "tok_per_s": 826.3026104339746, "loss": 0.6422720225527883, "entropy": 0.6670975536108017, "sharp": 0.09885596670210361} +{"time": "09:59:28", "event": "step", "step": 270, "rows_seen": 8640, "lr": 4.865817908283331e-05, "tok_per_s": 827.375040384621, "loss": 0.6690735965967178, "entropy": 0.6420081779360771, "sharp": 0.1370291942730546} +{"time": "10:03:17", "event": "step", "step": 280, "rows_seen": 8960, "lr": 4.560164286583922e-05, "tok_per_s": 828.4143308609, "loss": 0.7405221462249756, "entropy": 0.6498714536428452, "sharp": 0.10969929583370686} +{"time": "10:07:08", "event": "step", "step": 290, "rows_seen": 9280, "lr": 4.257404075601443e-05, "tok_per_s": 829.3279071091144, "loss": 0.6922129094600677, "entropy": 0.6504716016352177, "sharp": 0.12393020559102297} +{"time": "10:10:56", "event": "step", "step": 300, "rows_seen": 9600, "lr": 3.958807156877371e-05, "tok_per_s": 830.0992317733071, "loss": 0.6168073639273643, "entropy": 0.6522897556424141, "sharp": 0.11487219575792551} +{"time": "10:12:25", "event": "val", "step": 300, "rows_seen": 9600, "n": 500, "dropped_long": 0, "noul_ce": 0.5929418206214905, "noul_T": 0.6742968264197762, "noul_auc": 0.7332703245514842, "noul_acc": 0.71, "noul_f1": 0.7672552166934189, "noul_nll": 0.5955979643315077, "noul_ece": 0.084287614107132, "noul_brier": 0.20386416261754284, "noul_sharp": 0.13998760282993317, "entropy": 0.6352567672729492} +{"time": "10:16:13", "event": "step", "step": 310, "rows_seen": 9920, "lr": 3.665625949658603e-05, "tok_per_s": 820.8849739366888, "loss": 0.6018178090453148, "entropy": 0.6624333262443542, "sharp": 0.10367251187562943} +{"time": "10:20:06", "event": "step", "step": 320, "rows_seen": 10240, "lr": 3.379090157815887e-05, "tok_per_s": 822.1702387415982, "loss": 0.6465166732668877, "entropy": 0.6494022682309151, "sharp": 0.1233661575242877} +{"time": "10:23:53", "event": "step", "step": 330, "rows_seen": 10560, "lr": 3.10040161203844e-05, "tok_per_s": 823.4054480792602, "loss": 0.5945486165583134, "entropy": 0.6228913813829422, "sharp": 0.15881608240306377} +{"time": "10:27:40", "event": "step", "step": 340, "rows_seen": 10880, "lr": 2.8307292289383776e-05, "tok_per_s": 824.4701371797404, "loss": 0.6590227223932743, "entropy": 0.6720576547086239, "sharp": 0.08543357811868191} +{"time": "10:31:31", "event": "step", "step": 350, "rows_seen": 11200, "lr": 2.571204108208194e-05, "tok_per_s": 825.54706766103, "loss": 0.6911755874752998, "entropy": 0.6557943299412727, "sharp": 0.11337604280561209} +{"time": "10:35:16", "event": "step", "step": 360, "rows_seen": 11520, "lr": 2.322914788395595e-05, "tok_per_s": 826.4094436168443, "loss": 0.672113474458456, "entropy": 0.6336255744099617, "sharp": 0.13487344607710838} +{"time": "10:39:03", "event": "step", "step": 370, "rows_seen": 11840, "lr": 2.0869026811945157e-05, "tok_per_s": 827.1306496519461, "loss": 0.6700553707778454, "entropy": 0.6551447063684464, "sharp": 0.11386012751609087} +{"time": "10:42:56", "event": "step", "step": 380, "rows_seen": 12160, "lr": 1.8641577034025627e-05, "tok_per_s": 827.8117141715628, "loss": 0.6640688851475716, "entropy": 0.6413109470158815, "sharp": 0.13535328581929207} +{"time": "10:46:47", "event": "step", "step": 390, "rows_seen": 12480, "lr": 1.6556141248659472e-05, "tok_per_s": 828.5516928774163, "loss": 0.6268397979438305, "entropy": 0.632170895114541, "sharp": 0.14578355848789215} +{"time": "10:50:38", "event": "step", "step": 400, "rows_seen": 12800, "lr": 1.4621466498270773e-05, "tok_per_s": 829.2252069550974, "loss": 0.5890467502176762, "entropy": 0.644784465432167, "sharp": 0.12795494962483644} +{"time": "10:52:07", "event": "val", "step": 400, "rows_seen": 12800, "n": 500, "dropped_long": 0, "noul_ce": 0.5862334966659546, "noul_T": 0.7606483978429152, "noul_auc": 0.7381647567129669, "noul_acc": 0.7, "noul_f1": 0.7706422018348624, "noul_nll": 0.5874892692416906, "noul_ece": 0.0580228523015976, "noul_brier": 0.20039999223207658, "noul_sharp": 0.1556634157896042, "entropy": 0.6206153035163879} +{"time": "10:55:54", "event": "step", "step": 410, "rows_seen": 13120, "lr": 1.284566748110963e-05, "tok_per_s": 822.4939761646747, "loss": 0.6346560120582581, "entropy": 0.6383396536111832, "sharp": 0.1300529083237052} +{"time": "10:59:44", "event": "step", "step": 420, "rows_seen": 13440, "lr": 1.1236192515387821e-05, "tok_per_s": 823.1133792883301, "loss": 0.6326565071940422, "entropy": 0.6483790576457977, "sharp": 0.10434569697827101} +{"time": "11:03:35", "event": "step", "step": 430, "rows_seen": 13760, "lr": 9.799792298443728e-06, "tok_per_s": 823.9389462869989, "loss": 0.6603165343403816, "entropy": 0.6323894076049328, "sharp": 0.141101255081594} +{"time": "11:07:27", "event": "step", "step": 440, "rows_seen": 14080, "lr": 8.542491591971554e-06, "tok_per_s": 824.6522983284796, "loss": 0.6576901823282242, "entropy": 0.6554438918828964, "sharp": 0.10777606535702944} +{"time": "11:11:16", "event": "step", "step": 450, "rows_seen": 14400, "lr": 7.469563952077155e-06, "tok_per_s": 825.4080300941106, "loss": 0.6637649387121201, "entropy": 0.6494643464684486, "sharp": 0.12167353462427855} +{"time": "11:15:02", "event": "step", "step": 460, "rows_seen": 14720, "lr": 6.585509610150972e-06, "tok_per_s": 826.0346436111113, "loss": 0.6434320770204067, "entropy": 0.6648931764066219, "sharp": 0.09884033910930157} +{"time": "11:19:00", "event": "step", "step": 470, "rows_seen": 15040, "lr": 5.894036597333646e-06, "tok_per_s": 826.328165731913, "loss": 0.6828350089490414, "entropy": 0.6499973721802235, "sharp": 0.12077272217720747} +{"time": "11:23:00", "event": "step", "step": 480, "rows_seen": 15360, "lr": 5.398045191744809e-06, "tok_per_s": 826.7948135441493, "loss": 0.6283932998776436, "entropy": 0.641958087682724, "sharp": 0.13573688734322786} +{"time": "11:26:51", "event": "step", "step": 490, "rows_seen": 15680, "lr": 5.099615753708697e-06, "tok_per_s": 827.1983496490172, "loss": 0.6231705062091351, "entropy": 0.6532550044357777, "sharp": 0.11139336042106152} +{"time": "11:30:45", "event": "step", "step": 500, "rows_seen": 16000, "lr": 5e-06, "tok_per_s": 827.6767047919681, "loss": 0.6067280508577824, "entropy": 0.6390356905758381, "sharp": 0.14126633014529943} +{"time": "11:32:14", "event": "val", "step": 500, "rows_seen": 16000, "n": 500, "dropped_long": 0, "noul_ce": 0.5895894765853882, "noul_T": 0.7579464023116482, "noul_auc": 0.737270257042075, "noul_acc": 0.696, "noul_f1": 0.7610062893081762, "noul_nll": 0.5895842367261648, "noul_ece": 0.05830827236175538, "noul_brier": 0.20119680086413858, "noul_sharp": 0.15372470021247864, "entropy": 0.6235913038253784} +{"time": "11:35:25", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "replay_seen": 842, "step": 500, "tokens": 9814579, "tok_per_s": 808.7513072822612, "rows_per_s": 1.3878322760350466, "val": {"n": 1200, "dropped_long": 0, "noul_ce": 0.5733757615089417, "noul_T": 0.7536078969778258, "choice_ce": 0.6717521548271179, "choice_T": 1.0369940375159572, "score_ce": 1.0550035238265991, "score_T": 0.9641858015101192, "noul_auc": 0.7241642666666666, "noul_acc": 0.685, "noul_f1": 0.7608200455580866, "noul_nll": 0.590823968693614, "noul_ece": 0.05668055501580238, "noul_brier": 0.2023056829963619, "noul_sharp": 0.15657421946525574, "entropy": 0.6214356422424316}, "dropped_long": 0, "peak_gib": 18.236003875732422} diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/DONE b/e33/S1-E33-M40P40-n16k-a100-syn/DONE new file mode 100644 index 0000000000000000000000000000000000000000..1495071b366064ea733b93135496d9e1938fbc71 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/DONE @@ -0,0 +1 @@ +2026-10-24 06:50:10 \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/README.md b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/README.md new file mode 100644 index 0000000000000000000000000000000000000000..57167079ef3f181aacdbd678a733d71697d18f63 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/README.md @@ -0,0 +1,206 @@ +--- +base_model: /workspace/jev_kt/models/hf/JEV-9B +library_name: peft +tags: +- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **Developed by:** [More Information Needed] +- **Funded by [optional]:** [More Information Needed] +- **Shared by [optional]:** [More Information Needed] +- **Model type:** [More Information Needed] +- **Language(s) (NLP):** [More Information Needed] +- **License:** [More Information Needed] +- **Finetuned from model [optional]:** [More Information Needed] + +### Model Sources [optional] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. + +## How to Get Started with the Model + +Use the code below to get started with the model. + +[More Information Needed] + +## Training Details + +### Training Data + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). + +- **Hardware Type:** [More Information Needed] +- **Hours used:** [More Information Needed] +- **Cloud Provider:** [More Information Needed] +- **Compute Region:** [More Information Needed] +- **Carbon Emitted:** [More Information Needed] + +## Technical Specifications [optional] + +### Model Architecture and Objective + +[More Information Needed] + +### Compute Infrastructure + +[More Information Needed] + +#### Hardware + +[More Information Needed] + +#### Software + +[More Information Needed] + +## Citation [optional] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[More Information Needed] + +## More Information [optional] + +[More Information Needed] + +## Model Card Authors [optional] + +[More Information Needed] + +## Model Card Contact + +[More Information Needed] +### Framework versions + +- PEFT 0.21.2 \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..54b9ad43b862d1847270e01700d702377e8598cf --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/adapter_config.json @@ -0,0 +1,57 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 32, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 16, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "o_proj", + "in_proj_z", + "k_proj", + "up_proj", + "v_proj", + "gate_proj", + "in_proj_qkv", + "out_proj", + "q_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..713ed1174dfa34cb5739d20046b9c608edfa1146 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c39fecddf4eb407c1d1f604409131c1052549d8d610d0b6ee185e15b8a1e3980 +size 160486456 diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d093d03334d1bcd8511a3867adb084f382fd3c5 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_config.json @@ -0,0 +1,58 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": null, + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64.0, + "lora_bias": false, + "lora_dropout": 0.0, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "lm_head", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1dcb3347af9daeb2d3f9786c86c003603a5df3b5 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0779ab950735755fe718df27c1dbad733a41375c4fac3bdbb3157d77c5df16f8 +size 176641832 diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json new file mode 100644 index 0000000000000000000000000000000000000000..39533b484328cf437280af5832b3f852a40fb7b9 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/adapter_vllm/decision_head.json @@ -0,0 +1,99 @@ +{ + "bias": [ + 0.00030024704756215215, + -0.0003005006001330912, + -0.002452700864523649, + -0.0019857671577483416, + -0.0007780624437145889, + 0.001768098445609212, + 0.000989338499493897, + -0.002353379735723138, + -0.00040715502109378576, + -0.0008366759866476059, + 2.670622052391991e-07, + 0.0002804692485369742, + -0.0006767124868929386, + -0.0033653108403086662, + -0.003846945706754923, + -0.005079030059278011, + -0.005940228700637817, + -0.004343521315604448, + -0.004926910158246756, + -0.004934332333505154, + -0.003751315176486969, + -0.003340932074934244, + -0.005651351995766163, + -0.003458593273535371 + ], + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature" +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/calibration.json b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/calibration.json new file mode 100644 index 0000000000000000000000000000000000000000..8917a763db187b12b767142be443c2cf4aaacad6 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/calibration.json @@ -0,0 +1,29 @@ +{ + "version": 1, + "per_kind": { + "noul": 0.734008934482168, + "choice": 1.034794503470482, + "score": 0.927287545986635 + }, + "fit": { + "noul": { + "n": 1087, + "T": 0.734008934482168, + "ce_before": 0.5796747207641602, + "ce_after": 0.5732983350753784 + }, + "choice": { + "n": 60, + "T": 1.034794503470482, + "ce_before": 0.6689075231552124, + "ce_after": 0.6686466336250305 + }, + "score": { + "n": 53, + "T": 0.927287545986635, + "ce_before": 1.0485594272613525, + "ce_after": 1.0468835830688477 + } + }, + "source": "models/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000" +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/head.safetensors b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/head.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..dbbdf6b8c6ecdac7e287e128b7643ea3f3e0b241 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/head.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f732994d66b8cf5b243183313a5ade41c249085dd9c76ec19779be19a865c85 +size 393464 diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/judge_config.json b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/judge_config.json new file mode 100644 index 0000000000000000000000000000000000000000..0571be14cd868562c10e506f167d1a68ba79e2ae --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000/judge_config.json @@ -0,0 +1,164 @@ +{ + "base_model_path": "/root/models/Qwen3.5-9B", + "hidden_size": 4096, + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "kinds": [ + "noul", + "choice", + "score" + ], + "model_name": "jev-judge-qwen35-9b", + "model_version": "0.8.0", + "lm_head_included": true, + "weights_mode": "unmerged", + "adapter_subfolder": "adapter", + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ] + }, + "trained_stage": "e16", + "source_checkpoint": "models/e33/S1-E33-M40P40-n16k-a100-syn/ckpt-16000", + "softcap": null, + "e16": { + "data": "data/e33/E33-M40P40-n16k-a100", + "data_manifest": { + "arm": "E33-M40P40-n16k-a100", + "name": null, + "train_sha256": "6b8fc4a6962a559c74a8afeab09fba2dfe674609cd0ce16fd3be6090368d3f9a", + "n_train": 16000 + }, + "init": "released", + "objective": "s-log", + "sigma": 0.5, + "K": 8, + "reward": "brier", + "clip": 0.2, + "passes": 1, + "lr": 0.0001, + "lr_head": 0.0002, + "rows_per_step": 32, + "train_seed": 0, + "replay": "data/e16/replay/train.jsonl", + "replay_frac": 0.05, + "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", + "max_len": 3072, + "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", + "total_rows": 16000, + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05 + }, + "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", + "event": "ckpt", + "tag": "ckpt-16000", + "rows_seen": 16000, + "replay_seen": 842, + "step": 500, + "tokens": 9805957, + "tok_per_s": 854.5919225528155, + "rows_per_s": 1.4677850570068176, + "val": { + "n": 1200, + "dropped_long": 0, + "noul_ce": 0.5796747207641602, + "noul_T": 0.734008934482168, + "choice_ce": 0.6689075231552124, + "choice_T": 1.034794503470482, + "score_ce": 1.0485594272613525, + "score_T": 0.927287545986635, + "noul_auc": 0.7160405333333334, + "noul_acc": 0.702, + "noul_f1": 0.7880512091038406, + "noul_nll": 0.596515079177916, + "noul_ece": 0.06937007528543473, + "noul_brier": 0.2045756829835012, + "noul_sharp": 0.14921176433563232, + "entropy": 0.6275624632835388 + }, + "dropped_long": 0, + "peak_gib": 18.236003875732422 + } +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-a100-syn/train_log.jsonl b/e33/S1-E33-M40P40-n16k-a100-syn/train_log.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..fba96f0a461280313dcf081d415e319c0d7f95ae --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-a100-syn/train_log.jsonl @@ -0,0 +1,57 @@ +{"time": "03:38:54", "event": "start", "data": "data/e33/E33-M40P40-n16k-a100", "data_manifest": {"arm": "E33-M40P40-n16k-a100", "name": null, "train_sha256": "6b8fc4a6962a559c74a8afeab09fba2dfe674609cd0ce16fd3be6090368d3f9a", "n_train": 16000}, "init": "released", "objective": "s-log", "sigma": 0.5, "K": 8, "reward": "brier", "clip": 0.2, "passes": 1, "lr": 0.0001, "lr_head": 0.0002, "rows_per_step": 32, "train_seed": 0, "replay": "data/e16/replay/train.jsonl", "replay_frac": 0.05, "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", "max_len": 3072, "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", "total_rows": 16000, "lora": {"r": 16, "lora_alpha": 32, "lora_dropout": 0.05}, "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", "device": "cuda", "load_s": 79.5441153049469} +{"time": "03:42:38", "event": "step", "step": 10, "rows_seen": 320, "lr": 7.333333333333333e-05, "tok_per_s": 885.1924198579771, "loss": 0.7263630763627589, "entropy": 0.5674284184351563, "sharp": 0.21886649541556835} +{"time": "03:46:10", "event": "step", "step": 20, "rows_seen": 640, "lr": 9.997508952966197e-05, "tok_per_s": 891.8940374413393, "loss": 0.6402560044080019, "entropy": 0.6566950008273125, "sharp": 0.10619220696389675} +{"time": "03:49:48", "event": "step", "step": 30, "rows_seen": 960, "lr": 9.977596250541417e-05, "tok_per_s": 895.0844717309982, "loss": 0.6926834341138601, "entropy": 0.664935976266861, "sharp": 0.09891294501721859} +{"time": "03:53:26", "event": "step", "step": 40, "rows_seen": 1280, "lr": 9.937854366485982e-05, "tok_per_s": 897.764404973618, "loss": 0.6687077092938125, "entropy": 0.5757963806390762, "sharp": 0.20476821530610323} +{"time": "03:56:59", "event": "step", "step": 50, "rows_seen": 1600, "lr": 9.8784499920729e-05, "tok_per_s": 900.1193983788623, "loss": 0.709688164293766, "entropy": 0.6772790625691414, "sharp": 0.0819168733432889} +{"time": "04:00:34", "event": "step", "step": 60, "rows_seen": 1920, "lr": 9.79963228989294e-05, "tok_per_s": 899.7175227264156, "loss": 0.6411467157304287, "entropy": 0.6822525933384895, "sharp": 0.06333464942872524} +{"time": "04:04:13", "event": "step", "step": 70, "rows_seen": 2240, "lr": 9.701731848780134e-05, "tok_per_s": 899.2395269574124, "loss": 0.625581176020205, "entropy": 0.6701997518539429, "sharp": 0.10188802145421505} +{"time": "04:07:44", "event": "step", "step": 80, "rows_seen": 2560, "lr": 9.585159297207315e-05, "tok_per_s": 898.9562452626217, "loss": 0.6151504591107368, "entropy": 0.6824939623475075, "sharp": 0.06010466627776623} +{"time": "04:11:21", "event": "step", "step": 90, "rows_seen": 2880, "lr": 9.450403580967612e-05, "tok_per_s": 899.7407018724432, "loss": 0.655643668025732, "entropy": 0.564166933298111, "sharp": 0.21731171198189259} +{"time": "04:14:56", "event": "step", "step": 100, "rows_seen": 3200, "lr": 9.298029912365855e-05, "tok_per_s": 900.0971821216384, "loss": 0.587063618004322, "entropy": 0.6144744604825974, "sharp": 0.17060410790145397} +{"time": "04:16:19", "event": "val", "step": 100, "rows_seen": 3200, "n": 500, "dropped_long": 0, "noul_ce": 0.593371570110321, "noul_T": 1.1050007886661863, "noul_auc": 0.726299978059442, "noul_acc": 0.66, "noul_f1": 0.76775956284153, "noul_nll": 0.6029538196176291, "noul_ece": 0.0684438054561615, "noul_brier": 0.2082393859941073, "noul_sharp": 0.1974889636039734, "entropy": 0.5758929252624512} +{"time": "04:19:53", "event": "step", "step": 110, "rows_seen": 3520, "lr": 9.128677399521752e-05, "tok_per_s": 869.7571492048997, "loss": 0.7452945336699486, "entropy": 0.6653457954525948, "sharp": 0.0886428002268076} +{"time": "04:23:24", "event": "step", "step": 120, "rows_seen": 3840, "lr": 8.943056365728335e-05, "tok_per_s": 872.9344415873792, "loss": 0.688051794655621, "entropy": 0.6546291336417198, "sharp": 0.12737714871764183} +{"time": "04:27:08", "event": "step", "step": 130, "rows_seen": 4160, "lr": 8.741945370109202e-05, "tok_per_s": 875.4646623081286, "loss": 0.6162672359496355, "entropy": 0.621730662882328, "sharp": 0.15106641873717308} +{"time": "04:30:43", "event": "step", "step": 140, "rows_seen": 4480, "lr": 8.526187942070997e-05, "tok_per_s": 877.5404731048884, "loss": 0.6722078137099743, "entropy": 0.6648404449224472, "sharp": 0.090326271019876} +{"time": "04:34:28", "event": "step", "step": 150, "rows_seen": 4800, "lr": 8.296689043247915e-05, "tok_per_s": 879.0416802825639, "loss": 0.7514304770156741, "entropy": 0.6476933434605598, "sharp": 0.1341476384550333} +{"time": "04:38:01", "event": "step", "step": 160, "rows_seen": 5120, "lr": 8.054411271778132e-05, "tok_per_s": 880.3960920654483, "loss": 0.7412259690463543, "entropy": 0.6068226769566536, "sharp": 0.17566951271146536} +{"time": "04:41:40", "event": "step", "step": 170, "rows_seen": 5440, "lr": 7.800370824832674e-05, "tok_per_s": 881.5930592656113, "loss": 0.6779031604528427, "entropy": 0.6612563841044903, "sharp": 0.10762536432594061} +{"time": "04:45:18", "event": "step", "step": 180, "rows_seen": 5760, "lr": 7.535633236331324e-05, "tok_per_s": 882.6773865271027, "loss": 0.6398745849728584, "entropy": 0.6841178461909294, "sharp": 0.05769470985978842} +{"time": "04:48:51", "event": "step", "step": 190, "rows_seen": 6080, "lr": 7.26130890772308e-05, "tok_per_s": 883.3578829028257, "loss": 0.6973040346056223, "entropy": 0.6420499980449677, "sharp": 0.14343614503741264} +{"time": "04:52:31", "event": "step", "step": 200, "rows_seen": 6400, "lr": 6.978548450576568e-05, "tok_per_s": 884.2522483779724, "loss": 0.6453092582523823, "entropy": 0.6743573993444443, "sharp": 0.0814631087705493} +{"time": "04:53:55", "event": "val", "step": 200, "rows_seen": 6400, "n": 500, "dropped_long": 0, "noul_ce": 0.6030312180519104, "noul_T": 0.6272285269806894, "noul_auc": 0.7234477055239574, "noul_acc": 0.688, "noul_f1": 0.74, "noul_nll": 0.606294600456953, "noul_ece": 0.08061845314502716, "noul_brier": 0.20880262882127107, "noul_sharp": 0.1256517916917801, "entropy": 0.6458978056907654} +{"time": "04:57:31", "event": "step", "step": 210, "rows_seen": 6720, "lr": 6.688537860515266e-05, "tok_per_s": 869.3091400709204, "loss": 0.7088755257427692, "entropy": 0.6332606598734856, "sharp": 0.1201555784791708} +{"time": "05:01:11", "event": "step", "step": 220, "rows_seen": 7040, "lr": 6.392493542739706e-05, "tok_per_s": 870.9304981042593, "loss": 0.6857896484434605, "entropy": 0.675256609916687, "sharp": 0.067671412602067} +{"time": "05:04:49", "event": "step", "step": 230, "rows_seen": 7360, "lr": 6.091657210001309e-05, "tok_per_s": 871.7896602425201, "loss": 0.6461450196802616, "entropy": 0.6655657701194286, "sharp": 0.10216846689581871} +{"time": "05:08:29", "event": "step", "step": 240, "rows_seen": 7680, "lr": 5.7872906744276015e-05, "tok_per_s": 873.1519880006581, "loss": 0.6373593062162399, "entropy": 0.6440010443329811, "sharp": 0.11958589032292366} +{"time": "05:12:03", "event": "step", "step": 250, "rows_seen": 8000, "lr": 5.48067055504363e-05, "tok_per_s": 874.3116223569153, "loss": 0.6454163044691086, "entropy": 0.6323886066675186, "sharp": 0.15133871231228113} +{"time": "05:15:43", "event": "step", "step": 260, "rows_seen": 8320, "lr": 5.173082923188119e-05, "tok_per_s": 875.4343781471778, "loss": 0.6096611358225346, "entropy": 0.5486396923661232, "sharp": 0.24990964494645596} +{"time": "05:19:20", "event": "step", "step": 270, "rows_seen": 8640, "lr": 4.865817908283331e-05, "tok_per_s": 876.0012864679079, "loss": 0.6716932263225317, "entropy": 0.6735616438090801, "sharp": 0.08131483942270279} +{"time": "05:23:11", "event": "step", "step": 280, "rows_seen": 8960, "lr": 4.560164286583922e-05, "tok_per_s": 877.2890171669583, "loss": 0.7083826716989279, "entropy": 0.6559370756149292, "sharp": 0.10486898943781853} +{"time": "05:26:52", "event": "step", "step": 290, "rows_seen": 9280, "lr": 4.257404075601443e-05, "tok_per_s": 878.2532252369497, "loss": 0.6523172818124294, "entropy": 0.6475921049714088, "sharp": 0.12317295279353857} +{"time": "05:30:28", "event": "step", "step": 300, "rows_seen": 9600, "lr": 3.958807156877371e-05, "tok_per_s": 878.9171748157879, "loss": 0.6850400911644101, "entropy": 0.6319319866597652, "sharp": 0.15251983143389225} +{"time": "05:31:51", "event": "val", "step": 300, "rows_seen": 9600, "n": 500, "dropped_long": 0, "noul_ce": 0.5956318378448486, "noul_T": 0.7513072759938252, "noul_auc": 0.7277683077078868, "noul_acc": 0.698, "noul_f1": 0.7528641571194763, "noul_nll": 0.5975338519960642, "noul_ece": 0.07161584562063217, "noul_brier": 0.20485542972097182, "noul_sharp": 0.14779217541217804, "entropy": 0.6283900737762451} +{"time": "05:35:29", "event": "step", "step": 310, "rows_seen": 9920, "lr": 3.665625949658603e-05, "tok_per_s": 869.0667813801002, "loss": 0.7233850955963135, "entropy": 0.6242437940090895, "sharp": 0.138937434181571} +{"time": "05:39:04", "event": "step", "step": 320, "rows_seen": 10240, "lr": 3.379090157815887e-05, "tok_per_s": 870.2181391408117, "loss": 0.5846858490258455, "entropy": 0.5720607303082943, "sharp": 0.2097584642469883} +{"time": "05:42:40", "event": "step", "step": 330, "rows_seen": 10560, "lr": 3.10040161203844e-05, "tok_per_s": 871.0559290575437, "loss": 0.697181697934866, "entropy": 0.6308042667806149, "sharp": 0.1490570567548275} +{"time": "05:46:17", "event": "step", "step": 340, "rows_seen": 10880, "lr": 2.8307292289383776e-05, "tok_per_s": 871.7878583172628, "loss": 0.6868227832019329, "entropy": 0.6717342380434275, "sharp": 0.08868567738682032} +{"time": "05:49:53", "event": "step", "step": 350, "rows_seen": 11200, "lr": 2.571204108208194e-05, "tok_per_s": 872.624299463724, "loss": 0.6562195718288422, "entropy": 0.6541205197572708, "sharp": 0.10902172978967428} +{"time": "05:53:36", "event": "step", "step": 360, "rows_seen": 11520, "lr": 2.322914788395595e-05, "tok_per_s": 873.4192027887292, "loss": 0.7022134512662888, "entropy": 0.6380065456032753, "sharp": 0.13561051059514284} +{"time": "05:57:05", "event": "step", "step": 370, "rows_seen": 11840, "lr": 2.0869026811945157e-05, "tok_per_s": 873.9917700597257, "loss": 0.6171634159982204, "entropy": 0.6466353200376034, "sharp": 0.125652645714581} +{"time": "06:00:43", "event": "step", "step": 380, "rows_seen": 12160, "lr": 1.8641577034025627e-05, "tok_per_s": 874.7455617210661, "loss": 0.711353525519371, "entropy": 0.6458983868360519, "sharp": 0.13067299872636795} +{"time": "06:04:27", "event": "step", "step": 390, "rows_seen": 12480, "lr": 1.6556141248659472e-05, "tok_per_s": 875.5628721697493, "loss": 0.6505849175155163, "entropy": 0.664426177740097, "sharp": 0.10033649485558271} +{"time": "06:08:01", "event": "step", "step": 400, "rows_seen": 12800, "lr": 1.4621466498270773e-05, "tok_per_s": 876.0892673362345, "loss": 0.6862772032618523, "entropy": 0.6670309081673622, "sharp": 0.09477117843925953} +{"time": "06:09:25", "event": "val", "step": 400, "rows_seen": 12800, "n": 500, "dropped_long": 0, "noul_ce": 0.5938152074813843, "noul_T": 0.6869957716737829, "noul_auc": 0.7297092032201988, "noul_acc": 0.698, "noul_f1": 0.7599364069952306, "noul_nll": 0.5966609976887703, "noul_ece": 0.08797803795337679, "noul_brier": 0.20456294142083767, "noul_sharp": 0.13961169123649597, "entropy": 0.6338917016983032} +{"time": "06:13:06", "event": "step", "step": 410, "rows_seen": 13120, "lr": 1.284566748110963e-05, "tok_per_s": 868.7751916750661, "loss": 0.728324125520885, "entropy": 0.6291781179606915, "sharp": 0.1470704460516572} +{"time": "06:16:39", "event": "step", "step": 420, "rows_seen": 13440, "lr": 1.1236192515387821e-05, "tok_per_s": 869.395038823365, "loss": 0.6821083053946495, "entropy": 0.6421918347477913, "sharp": 0.1251325123012066} +{"time": "06:20:18", "event": "step", "step": 430, "rows_seen": 13760, "lr": 9.799792298443728e-06, "tok_per_s": 870.0177792647041, "loss": 0.655871819704771, "entropy": 0.6480694189667702, "sharp": 0.11198160145431757} +{"time": "06:23:52", "event": "step", "step": 440, "rows_seen": 14080, "lr": 8.542491591971554e-06, "tok_per_s": 870.6776818599245, "loss": 0.6682901065796614, "entropy": 0.6499051600694656, "sharp": 0.11998931784182787} +{"time": "06:27:28", "event": "step", "step": 450, "rows_seen": 14400, "lr": 7.469563952077155e-06, "tok_per_s": 871.3863957204726, "loss": 0.636380510404706, "entropy": 0.6391087155789137, "sharp": 0.14191021583974361} +{"time": "06:31:13", "event": "step", "step": 460, "rows_seen": 14720, "lr": 6.585509610150972e-06, "tok_per_s": 872.1255134582918, "loss": 0.6400385666638613, "entropy": 0.6596169173717499, "sharp": 0.10499859135597944} +{"time": "06:34:55", "event": "step", "step": 470, "rows_seen": 15040, "lr": 5.894036597333646e-06, "tok_per_s": 872.721874818939, "loss": 0.6945435144007206, "entropy": 0.6433812007308006, "sharp": 0.13231699168682098} +{"time": "06:38:33", "event": "step", "step": 480, "rows_seen": 15360, "lr": 5.398045191744809e-06, "tok_per_s": 873.275297459479, "loss": 0.6982832103967667, "entropy": 0.6603386998176575, "sharp": 0.09769300371408463} +{"time": "06:42:13", "event": "step", "step": 490, "rows_seen": 15680, "lr": 5.099615753708697e-06, "tok_per_s": 874.0264061401459, "loss": 0.6273852866142988, "entropy": 0.6676870472729206, "sharp": 0.09229922015219927} +{"time": "06:45:48", "event": "step", "step": 500, "rows_seen": 16000, "lr": 5e-06, "tok_per_s": 874.4697314310247, "loss": 0.6198776587843895, "entropy": 0.6514460258185863, "sharp": 0.11864995211362839} +{"time": "06:47:11", "event": "val", "step": 500, "rows_seen": 16000, "n": 500, "dropped_long": 0, "noul_ce": 0.591821551322937, "noul_T": 0.6996818490956741, "noul_auc": 0.7292197600040506, "noul_acc": 0.71, "noul_f1": 0.7845468053491828, "noul_nll": 0.5957867748588324, "noul_ece": 0.07907445490360261, "noul_brier": 0.2041328460113967, "noul_sharp": 0.14249670505523682, "entropy": 0.6315938830375671} +{"time": "06:50:10", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "replay_seen": 842, "step": 500, "tokens": 9805957, "tok_per_s": 854.5919225528155, "rows_per_s": 1.4677850570068176, "val": {"n": 1200, "dropped_long": 0, "noul_ce": 0.5796747207641602, "noul_T": 0.734008934482168, "choice_ce": 0.6689075231552124, "choice_T": 1.034794503470482, "score_ce": 1.0485594272613525, "score_T": 0.927287545986635, "noul_auc": 0.7160405333333334, "noul_acc": 0.702, "noul_f1": 0.7880512091038406, "noul_nll": 0.596515079177916, "noul_ece": 0.06937007528543473, "noul_brier": 0.2045756829835012, "noul_sharp": 0.14921176433563232, "entropy": 0.6275624632835388}, "dropped_long": 0, "peak_gib": 18.236003875732422} diff --git a/e33/S1-E33-M40P40-n16k-syn/DONE b/e33/S1-E33-M40P40-n16k-syn/DONE new file mode 100644 index 0000000000000000000000000000000000000000..0c8adcb58af34902432a37a54d90f28de8705350 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/DONE @@ -0,0 +1 @@ +2026-10-23 11:24:49 \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/README.md b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/README.md new file mode 100644 index 0000000000000000000000000000000000000000..57167079ef3f181aacdbd678a733d71697d18f63 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/README.md @@ -0,0 +1,206 @@ +--- +base_model: /workspace/jev_kt/models/hf/JEV-9B +library_name: peft +tags: +- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **Developed by:** [More Information Needed] +- **Funded by [optional]:** [More Information Needed] +- **Shared by [optional]:** [More Information Needed] +- **Model type:** [More Information Needed] +- **Language(s) (NLP):** [More Information Needed] +- **License:** [More Information Needed] +- **Finetuned from model [optional]:** [More Information Needed] + +### Model Sources [optional] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. + +## How to Get Started with the Model + +Use the code below to get started with the model. + +[More Information Needed] + +## Training Details + +### Training Data + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). + +- **Hardware Type:** [More Information Needed] +- **Hours used:** [More Information Needed] +- **Cloud Provider:** [More Information Needed] +- **Compute Region:** [More Information Needed] +- **Carbon Emitted:** [More Information Needed] + +## Technical Specifications [optional] + +### Model Architecture and Objective + +[More Information Needed] + +### Compute Infrastructure + +[More Information Needed] + +#### Hardware + +[More Information Needed] + +#### Software + +[More Information Needed] + +## Citation [optional] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[More Information Needed] + +## More Information [optional] + +[More Information Needed] + +## Model Card Authors [optional] + +[More Information Needed] + +## Model Card Contact + +[More Information Needed] +### Framework versions + +- PEFT 0.21.2 \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/adapter_config.json b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baaddb41cda79e92b44198194cc233eff3123910 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/adapter_config.json @@ -0,0 +1,57 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 32, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 16, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "in_proj_z", + "out_proj", + "up_proj", + "q_proj", + "k_proj", + "gate_proj", + "o_proj", + "in_proj_qkv", + "v_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..53c6a15e22d2db2967587593219598fbc9d7e1a9 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0682741f4c2b890a6bfcbcee9ff1a50625a04faf57848700a97f3241812aa47 +size 160486456 diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d093d03334d1bcd8511a3867adb084f382fd3c5 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_config.json @@ -0,0 +1,58 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": { + "base_model_class": "Qwen3_5ForCausalLM", + "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5" + }, + "base_model_name_or_path": null, + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": false, + "inference_mode": true, + "init_lora_weights": true, + "kasa_config": null, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64.0, + "lora_bias": false, + "lora_dropout": 0.0, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "monteclora_config": null, + "peft_type": "LORA", + "peft_version": "0.21.0", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "lm_head", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ], + "target_parameters": null, + "task_type": null, + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false, + "velora_config": null +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c95516076c1413a9956a6d40eadc46bbc3d60454 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3c5da55b78177e8f21d5ba9c9759020acaf1ee2f99f44d16ee83dad50ce0b39 +size 176641832 diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json new file mode 100644 index 0000000000000000000000000000000000000000..3a1ec15c24a496dbdca955820f212fe6b9eecbd3 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/adapter_vllm/decision_head.json @@ -0,0 +1,99 @@ +{ + "bias": [ + 0.000385597231797874, + -0.00038585023139603436, + -0.0025997136253863573, + -0.001556922565214336, + -0.0006185499369166791, + 0.002033072058111429, + 0.0009567081579007208, + -0.003146836068481207, + -0.00024209108960349113, + -0.0005781410727649927, + 9.662006050348282e-05, + 0.00032394504523836076, + -0.00022292108042165637, + -0.0034981975331902504, + -0.0042379372753202915, + -0.005086993332952261, + -0.005952034145593643, + -0.004357457626610994, + -0.004930952563881874, + -0.0049574230797588825, + -0.0037764417938888073, + -0.0033611999824643135, + -0.005689066369086504, + -0.0034674094058573246 + ], + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature" +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/calibration.json b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/calibration.json new file mode 100644 index 0000000000000000000000000000000000000000..5906ed8f479c1086488ca8e6b4d84cb80eb7ae65 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/calibration.json @@ -0,0 +1,29 @@ +{ + "version": 1, + "per_kind": { + "noul": 0.7167783643453142, + "choice": 1.0255042891935702, + "score": 0.991955069195411 + }, + "fit": { + "noul": { + "n": 1087, + "T": 0.7167783643453142, + "ce_before": 0.5752585530281067, + "ce_after": 0.567782461643219 + }, + "choice": { + "n": 60, + "T": 1.0255042891935702, + "ce_before": 0.6692060232162476, + "ce_after": 0.669064462184906 + }, + "score": { + "n": 53, + "T": 0.991955069195411, + "ce_before": 1.0639286041259766, + "ce_after": 1.0639088153839111 + } + }, + "source": "models/e33/S1-E33-M40P40-n16k-syn/ckpt-16000" +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/head.safetensors b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/head.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..48a96f2279481a7f71428f2a673c7ce965507a0d --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/head.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5661f26b162751277b06783aa13a5e380994254200093bd44ae28e4f93bfeb07 +size 393464 diff --git a/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/judge_config.json b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/judge_config.json new file mode 100644 index 0000000000000000000000000000000000000000..067376c22da0a0b77303e52c45771fbfb507df09 --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/ckpt-16000/judge_config.json @@ -0,0 +1,164 @@ +{ + "base_model_path": "/root/models/Qwen3.5-9B", + "hidden_size": 4096, + "slots": { + "num_slots": 24, + "ranges": { + "noul": [ + 0, + 2 + ], + "score": [ + 2, + 8 + ], + "choice": [ + 8, + 24 + ] + }, + "verbalizers": [ + "false", + "true", + "0", + "1", + "2", + "3", + "4", + "5", + "A", + "B", + "C", + "D", + "E", + "F", + "G", + "H", + "I", + "J", + "K", + "L", + "M", + "N", + "O", + "P" + ], + "template_version": "bare-v1" + }, + "verbalizer_ids": [ + 3721, + 1802, + 15, + 16, + 17, + 18, + 19, + 20, + 32, + 33, + 34, + 35, + 36, + 37, + 38, + 39, + 40, + 41, + 42, + 43, + 44, + 45, + 46, + 47 + ], + "kinds": [ + "noul", + "choice", + "score" + ], + "model_name": "jev-judge-qwen35-9b", + "model_version": "0.8.0", + "lm_head_included": true, + "weights_mode": "unmerged", + "adapter_subfolder": "adapter", + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05, + "target_modules": [ + "down_proj", + "gate_proj", + "in_proj_qkv", + "in_proj_z", + "k_proj", + "o_proj", + "out_proj", + "q_proj", + "up_proj", + "v_proj" + ] + }, + "trained_stage": "e16", + "source_checkpoint": "models/e33/S1-E33-M40P40-n16k-syn/ckpt-16000", + "softcap": null, + "e16": { + "data": "data/e33/E33-M40P40-n16k", + "data_manifest": { + "arm": "E33-M40P40-n16k", + "name": null, + "train_sha256": "f8be2a02731a11ebdec6a1f3e7b20732054a909b572f0191f57759a3a631197e", + "n_train": 16000 + }, + "init": "released", + "objective": "s-log", + "sigma": 0.5, + "K": 8, + "reward": "brier", + "clip": 0.2, + "passes": 1, + "lr": 0.0001, + "lr_head": 0.0002, + "rows_per_step": 32, + "train_seed": 0, + "replay": "data/e16/replay/train.jsonl", + "replay_frac": 0.05, + "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", + "max_len": 3072, + "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", + "total_rows": 16000, + "lora": { + "r": 16, + "lora_alpha": 32, + "lora_dropout": 0.05 + }, + "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", + "event": "ckpt", + "tag": "ckpt-16000", + "rows_seen": 16000, + "replay_seen": 842, + "step": 500, + "tokens": 9805957, + "tok_per_s": 853.0247657415953, + "rows_per_s": 1.4650934222487946, + "val": { + "n": 1200, + "dropped_long": 0, + "noul_ce": 0.5752585530281067, + "noul_T": 0.7167783643453142, + "choice_ce": 0.6692060232162476, + "choice_T": 1.0255042891935702, + "score_ce": 1.0639286041259766, + "score_T": 0.991955069195411, + "noul_auc": 0.7213098666666666, + "noul_acc": 0.696, + "noul_f1": 0.7721139430284858, + "noul_nll": 0.5925446359589696, + "noul_ece": 0.05998595729470254, + "noul_brier": 0.20298855838876312, + "noul_sharp": 0.1488448828458786, + "entropy": 0.6269921064376831 + }, + "dropped_long": 0, + "peak_gib": 18.236003875732422 + } +} \ No newline at end of file diff --git a/e33/S1-E33-M40P40-n16k-syn/train_log.jsonl b/e33/S1-E33-M40P40-n16k-syn/train_log.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..f0b630919196e8840d4435dad48868bd7333067d --- /dev/null +++ b/e33/S1-E33-M40P40-n16k-syn/train_log.jsonl @@ -0,0 +1,57 @@ +{"time": "08:13:07", "event": "start", "data": "data/e33/E33-M40P40-n16k", "data_manifest": {"arm": "E33-M40P40-n16k", "name": null, "train_sha256": "f8be2a02731a11ebdec6a1f3e7b20732054a909b572f0191f57759a3a631197e", "n_train": 16000}, "init": "released", "objective": "s-log", "sigma": 0.5, "K": 8, "reward": "brier", "clip": 0.2, "passes": 1, "lr": 0.0001, "lr_head": 0.0002, "rows_per_step": 32, "train_seed": 0, "replay": "data/e16/replay/train.jsonl", "replay_frac": 0.05, "replay_sha256": "9dd2df19851678750cf79d2fdb47581720930ef2f0c6e18e6fbe23ef2181847e", "max_len": 3072, "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", "total_rows": 16000, "lora": {"r": 16, "lora_alpha": 32, "lora_dropout": 0.05}, "trainer_sha256": "a0b9d7552588ffb068565686509162a4f122f6f03392e693fa20088493e2a4f3", "device": "cuda", "load_s": 186.51470518112183} +{"time": "08:16:54", "event": "step", "step": 10, "rows_seen": 320, "lr": 7.333333333333333e-05, "tok_per_s": 877.5466413467042, "loss": 0.701326496899128, "entropy": 0.6085406895726919, "sharp": 0.1839306727051735} +{"time": "08:20:26", "event": "step", "step": 20, "rows_seen": 640, "lr": 9.997508952966197e-05, "tok_per_s": 887.6896671324774, "loss": 0.6841608211398125, "entropy": 0.6104379445314407, "sharp": 0.1718672476708889} +{"time": "08:24:03", "event": "step", "step": 30, "rows_seen": 960, "lr": 9.977596250541417e-05, "tok_per_s": 892.5419895132027, "loss": 0.6867619939148426, "entropy": 0.673102293163538, "sharp": 0.08125011995434761} +{"time": "08:27:42", "event": "step", "step": 40, "rows_seen": 1280, "lr": 9.937854366485982e-05, "tok_per_s": 895.4628734580723, "loss": 0.676459826529026, "entropy": 0.6758796349167824, "sharp": 0.08045716490596533} +{"time": "08:31:15", "event": "step", "step": 50, "rows_seen": 1600, "lr": 9.8784499920729e-05, "tok_per_s": 897.7686169225742, "loss": 0.6884408891201019, "entropy": 0.6762037128210068, "sharp": 0.0816372400149703} +{"time": "08:34:50", "event": "step", "step": 60, "rows_seen": 1920, "lr": 9.79963228989294e-05, "tok_per_s": 897.6880586788025, "loss": 0.636358592659235, "entropy": 0.6442252621054649, "sharp": 0.13726660516113043} +{"time": "08:38:30", "event": "step", "step": 70, "rows_seen": 2240, "lr": 9.701731848780134e-05, "tok_per_s": 897.1675513089461, "loss": 0.6046697190031409, "entropy": 0.6089828992262483, "sharp": 0.17679774016141891} +{"time": "08:42:01", "event": "step", "step": 80, "rows_seen": 2560, "lr": 9.585159297207315e-05, "tok_per_s": 896.9060955596917, "loss": 0.6517690941691399, "entropy": 0.6728371158242226, "sharp": 0.0801616394892335} +{"time": "08:45:38", "event": "step", "step": 90, "rows_seen": 2880, "lr": 9.450403580967612e-05, "tok_per_s": 897.8684786000953, "loss": 0.6420326754450798, "entropy": 0.633781835436821, "sharp": 0.14034037850797176} +{"time": "08:49:13", "event": "step", "step": 100, "rows_seen": 3200, "lr": 9.298029912365855e-05, "tok_per_s": 898.2824442351211, "loss": 0.6331725865602493, "entropy": 0.6289944816380739, "sharp": 0.15755248069763184} +{"time": "08:50:37", "event": "val", "step": 100, "rows_seen": 3200, "n": 500, "dropped_long": 0, "noul_ce": 0.5850945115089417, "noul_T": 0.854966010431862, "noul_auc": 0.7192114901014328, "noul_acc": 0.692, "noul_f1": 0.7680722891566265, "noul_nll": 0.5950000842362643, "noul_ece": 0.03460281646251679, "noul_brier": 0.20397586889466213, "noul_sharp": 0.17110642790794373, "entropy": 0.6072657704353333} +{"time": "08:54:11", "event": "step", "step": 110, "rows_seen": 3520, "lr": 9.128677399521752e-05, "tok_per_s": 868.1237159090424, "loss": 0.7051668204367161, "entropy": 0.6504334770143032, "sharp": 0.12005877401679754} +{"time": "08:57:43", "event": "step", "step": 120, "rows_seen": 3840, "lr": 8.943056365728335e-05, "tok_per_s": 871.3350506495691, "loss": 0.6997609799727798, "entropy": 0.6523835957050323, "sharp": 0.11436803452670574} +{"time": "09:01:27", "event": "step", "step": 130, "rows_seen": 4160, "lr": 8.741945370109202e-05, "tok_per_s": 873.8544416924424, "loss": 0.6477537378668785, "entropy": 0.629639521241188, "sharp": 0.15131608955562115} +{"time": "09:05:02", "event": "step", "step": 140, "rows_seen": 4480, "lr": 8.526187942070997e-05, "tok_per_s": 875.9735754458873, "loss": 0.675873413681984, "entropy": 0.666770450770855, "sharp": 0.10009584948420525} +{"time": "09:08:47", "event": "step", "step": 150, "rows_seen": 4800, "lr": 8.296689043247915e-05, "tok_per_s": 877.4239305410016, "loss": 0.6851847022771835, "entropy": 0.650326780974865, "sharp": 0.11930546164512634} +{"time": "09:12:21", "event": "step", "step": 160, "rows_seen": 5120, "lr": 8.054411271778132e-05, "tok_per_s": 878.7809701199482, "loss": 0.731493353843689, "entropy": 0.6083054915070534, "sharp": 0.16842507757246494} +{"time": "09:16:00", "event": "step", "step": 170, "rows_seen": 5440, "lr": 7.800370824832674e-05, "tok_per_s": 880.073397512603, "loss": 0.6781997233629227, "entropy": 0.6768593862652779, "sharp": 0.06668039597570896} +{"time": "09:19:38", "event": "step", "step": 180, "rows_seen": 5760, "lr": 7.535633236331324e-05, "tok_per_s": 881.1884452982389, "loss": 0.6352675929665565, "entropy": 0.6572296470403671, "sharp": 0.11780781578272581} +{"time": "09:23:11", "event": "step", "step": 190, "rows_seen": 6080, "lr": 7.26130890772308e-05, "tok_per_s": 881.8816407506598, "loss": 0.682921402156353, "entropy": 0.6563410013914108, "sharp": 0.1110598174855113} +{"time": "09:26:52", "event": "step", "step": 200, "rows_seen": 6400, "lr": 6.978548450576568e-05, "tok_per_s": 882.8492312107616, "loss": 0.6532541811466217, "entropy": 0.6616755835711956, "sharp": 0.10394374839961529} +{"time": "09:28:15", "event": "val", "step": 200, "rows_seen": 6400, "n": 500, "dropped_long": 0, "noul_ce": 0.5963744521141052, "noul_T": 0.6831531784794845, "noul_auc": 0.7310931461072387, "noul_acc": 0.7, "noul_f1": 0.7781065088757396, "noul_nll": 0.5970423142313958, "noul_ece": 0.07967041397094728, "noul_brier": 0.2042589849991402, "noul_sharp": 0.13798345625400543, "entropy": 0.6366021633148193} +{"time": "09:31:51", "event": "step", "step": 210, "rows_seen": 6720, "lr": 6.688537860515266e-05, "tok_per_s": 868.0067589312561, "loss": 0.6937913708388805, "entropy": 0.6538265645503998, "sharp": 0.10700359661132097} +{"time": "09:35:31", "event": "step", "step": 220, "rows_seen": 7040, "lr": 6.392493542739706e-05, "tok_per_s": 869.6566944426392, "loss": 0.6824341788887978, "entropy": 0.6751945465803146, "sharp": 0.07583794463425875} +{"time": "09:39:09", "event": "step", "step": 230, "rows_seen": 7360, "lr": 6.091657210001309e-05, "tok_per_s": 870.6165679029849, "loss": 0.665081076323986, "entropy": 0.6792085580527782, "sharp": 0.06292346306145191} +{"time": "09:42:50", "event": "step", "step": 240, "rows_seen": 7680, "lr": 5.7872906744276015e-05, "tok_per_s": 871.9453799522336, "loss": 0.6527600958943367, "entropy": 0.6512566730380058, "sharp": 0.11794712487608194} +{"time": "09:46:24", "event": "step", "step": 250, "rows_seen": 8000, "lr": 5.48067055504363e-05, "tok_per_s": 873.1541322926765, "loss": 0.6435472592711449, "entropy": 0.6326887533068657, "sharp": 0.1492773024365306} +{"time": "09:50:04", "event": "step", "step": 260, "rows_seen": 8320, "lr": 5.173082923188119e-05, "tok_per_s": 874.2865157421535, "loss": 0.6347041428089142, "entropy": 0.6309486702084541, "sharp": 0.1465424858033657} +{"time": "09:53:42", "event": "step", "step": 270, "rows_seen": 8640, "lr": 4.865817908283331e-05, "tok_per_s": 874.8572519359606, "loss": 0.668428199365735, "entropy": 0.670079194009304, "sharp": 0.084346785210073} +{"time": "09:57:33", "event": "step", "step": 280, "rows_seen": 8960, "lr": 4.560164286583922e-05, "tok_per_s": 876.1877894873659, "loss": 0.7080271132290363, "entropy": 0.6305513642728329, "sharp": 0.1306461114436388} +{"time": "10:01:14", "event": "step", "step": 290, "rows_seen": 9280, "lr": 4.257404075601443e-05, "tok_per_s": 877.121558067762, "loss": 0.6656216531991959, "entropy": 0.6559411734342575, "sharp": 0.10965016391128302} +{"time": "10:04:50", "event": "step", "step": 300, "rows_seen": 9600, "lr": 3.958807156877371e-05, "tok_per_s": 877.8182152820489, "loss": 0.670805336907506, "entropy": 0.627765167504549, "sharp": 0.15224727615714073} +{"time": "10:06:13", "event": "val", "step": 300, "rows_seen": 9600, "n": 500, "dropped_long": 0, "noul_ce": 0.5935400128364563, "noul_T": 0.8376717232689267, "noul_auc": 0.7342492109837809, "noul_acc": 0.688, "noul_f1": 0.7364864864864865, "noul_nll": 0.5966843917518854, "noul_ece": 0.07446267390251161, "noul_brier": 0.20503769866799518, "noul_sharp": 0.15500333905220032, "entropy": 0.6151803731918335} +{"time": "10:09:51", "event": "step", "step": 310, "rows_seen": 9920, "lr": 3.665625949658603e-05, "tok_per_s": 867.9574780985163, "loss": 0.6864422373473644, "entropy": 0.6181308822706342, "sharp": 0.14367495104670525} +{"time": "10:13:26", "event": "step", "step": 320, "rows_seen": 10240, "lr": 3.379090157815887e-05, "tok_per_s": 869.1394407152385, "loss": 0.6503309532999992, "entropy": 0.6404871940612793, "sharp": 0.1307089775800705} +{"time": "10:17:04", "event": "step", "step": 330, "rows_seen": 10560, "lr": 3.10040161203844e-05, "tok_per_s": 869.9114240213348, "loss": 0.6424980722367764, "entropy": 0.6509234420955181, "sharp": 0.1187315285205841} +{"time": "10:20:40", "event": "step", "step": 340, "rows_seen": 10880, "lr": 2.8307292289383776e-05, "tok_per_s": 870.6832768599025, "loss": 0.6834957003593445, "entropy": 0.6689800024032593, "sharp": 0.08583842124789953} +{"time": "10:24:17", "event": "step", "step": 350, "rows_seen": 11200, "lr": 2.571204108208194e-05, "tok_per_s": 871.4994333441027, "loss": 0.668251134455204, "entropy": 0.6571514010429382, "sharp": 0.10154504049569368} +{"time": "10:28:00", "event": "step", "step": 360, "rows_seen": 11520, "lr": 2.322914788395595e-05, "tok_per_s": 872.2809385901488, "loss": 0.7033851593732834, "entropy": 0.659178338944912, "sharp": 0.10913206543773413} +{"time": "10:31:29", "event": "step", "step": 370, "rows_seen": 11840, "lr": 2.0869026811945157e-05, "tok_per_s": 872.866050423109, "loss": 0.6349151842296124, "entropy": 0.66221584379673, "sharp": 0.10252324119210243} +{"time": "10:35:08", "event": "step", "step": 380, "rows_seen": 12160, "lr": 1.8641577034025627e-05, "tok_per_s": 873.6073732025222, "loss": 0.7021514661610126, "entropy": 0.651942141354084, "sharp": 0.11681764293462038} +{"time": "10:38:52", "event": "step", "step": 390, "rows_seen": 12480, "lr": 1.6556141248659472e-05, "tok_per_s": 874.3996090147184, "loss": 0.6621464490890503, "entropy": 0.6576134413480759, "sharp": 0.10623965971171856} +{"time": "10:42:27", "event": "step", "step": 400, "rows_seen": 12800, "lr": 1.4621466498270773e-05, "tok_per_s": 874.8924920068696, "loss": 0.6770116165280342, "entropy": 0.6663002818822861, "sharp": 0.09193841274827719} +{"time": "10:43:50", "event": "val", "step": 400, "rows_seen": 12800, "n": 500, "dropped_long": 0, "noul_ce": 0.591166615486145, "noul_T": 0.7227459834517542, "noul_auc": 0.7332196924946415, "noul_acc": 0.712, "noul_f1": 0.7669902912621359, "noul_nll": 0.5939047406762838, "noul_ece": 0.06830801701545715, "noul_brier": 0.2033137971372521, "noul_sharp": 0.1456829309463501, "entropy": 0.6279944181442261} +{"time": "10:47:32", "event": "step", "step": 410, "rows_seen": 13120, "lr": 1.284566748110963e-05, "tok_per_s": 867.5863743979382, "loss": 0.6724479738622904, "entropy": 0.6369229070842266, "sharp": 0.13556503131985664} +{"time": "10:51:06", "event": "step", "step": 420, "rows_seen": 13440, "lr": 1.1236192515387821e-05, "tok_per_s": 868.1674091588366, "loss": 0.6634872686117887, "entropy": 0.6534842401742935, "sharp": 0.10859416890889406} +{"time": "10:54:45", "event": "step", "step": 430, "rows_seen": 13760, "lr": 9.799792298443728e-06, "tok_per_s": 868.7565802681784, "loss": 0.6799826323986053, "entropy": 0.6522127091884613, "sharp": 0.1109787030145526} +{"time": "10:58:20", "event": "step", "step": 440, "rows_seen": 14080, "lr": 8.542491591971554e-06, "tok_per_s": 869.3948165742845, "loss": 0.6743068546056747, "entropy": 0.6556941941380501, "sharp": 0.11242342181503773} +{"time": "11:01:57", "event": "step", "step": 450, "rows_seen": 14400, "lr": 7.469563952077155e-06, "tok_per_s": 870.0264906246681, "loss": 0.6403181254863739, "entropy": 0.652459166944027, "sharp": 0.12286585383117199} +{"time": "11:05:43", "event": "step", "step": 460, "rows_seen": 14720, "lr": 6.585509610150972e-06, "tok_per_s": 870.7372262320418, "loss": 0.653562530875206, "entropy": 0.6596773974597454, "sharp": 0.10167846642434597} +{"time": "11:09:25", "event": "step", "step": 470, "rows_seen": 15040, "lr": 5.894036597333646e-06, "tok_per_s": 871.3085341646346, "loss": 0.7055892683565617, "entropy": 0.6569633148610592, "sharp": 0.11131461337208748} +{"time": "11:13:05", "event": "step", "step": 480, "rows_seen": 15360, "lr": 5.398045191744809e-06, "tok_per_s": 871.8195047892456, "loss": 0.6841177884489298, "entropy": 0.665019553154707, "sharp": 0.09070733468979597} +{"time": "11:16:46", "event": "step", "step": 490, "rows_seen": 15680, "lr": 5.099615753708697e-06, "tok_per_s": 872.5281754984467, "loss": 0.6354478728026152, "entropy": 0.6697030067443848, "sharp": 0.08762179035693407} +{"time": "11:20:21", "event": "step", "step": 500, "rows_seen": 16000, "lr": 5e-06, "tok_per_s": 872.9076742972147, "loss": 0.6421502344310284, "entropy": 0.6548953130841255, "sharp": 0.1104683019220829} +{"time": "11:21:45", "event": "val", "step": 500, "rows_seen": 16000, "n": 500, "dropped_long": 0, "noul_ce": 0.590094804763794, "noul_T": 0.701236751591683, "noul_auc": 0.7329496548581458, "noul_acc": 0.71, "noul_f1": 0.7716535433070866, "noul_nll": 0.5929140857309103, "noul_ece": 0.0695128221511841, "noul_brier": 0.20282698300716095, "noul_sharp": 0.14450830221176147, "entropy": 0.6300867199897766} +{"time": "11:24:49", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "replay_seen": 842, "step": 500, "tokens": 9805957, "tok_per_s": 853.0247657415953, "rows_per_s": 1.4650934222487946, "val": {"n": 1200, "dropped_long": 0, "noul_ce": 0.5752585530281067, "noul_T": 0.7167783643453142, "choice_ce": 0.6692060232162476, "choice_T": 1.0255042891935702, "score_ce": 1.0639286041259766, "score_T": 0.991955069195411, "noul_auc": 0.7213098666666666, "noul_acc": 0.696, "noul_f1": 0.7721139430284858, "noul_nll": 0.5925446359589696, "noul_ece": 0.05998595729470254, "noul_brier": 0.20298855838876312, "noul_sharp": 0.1488448828458786, "entropy": 0.6269921064376831}, "dropped_long": 0, "peak_gib": 18.236003875732422}