Training in progress, step 200

adapter_config.json

@@ -14,13 +14,13 @@
   "r": 64,
   "revision": null,
   "target_modules": [
-    "q_proj",
+    "v_proj",
     "down_proj",
+    "o_proj",
     "k_proj",
     "up_proj",
-    "o_proj",
-    "v_proj",
-    "gate_proj"
+    "gate_proj",
+    "q_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ec6edef86ecafc02b075bf1c1669bf7c4c2005d3547d98a3b1722a931967bcc
+oid sha256:854d9775c36d691131435da90a76e8c4b814ee81dcbe15efd3d242c7247dc3a7
 size 319977229

checkpoint-200/README.md

@@ -4,6 +4,17 @@ library_name: peft
 ## Training procedure
 
 
+The following `bitsandbytes` quantization config was used during training:
+- load_in_8bit: False
+- load_in_4bit: True
+- llm_int8_threshold: 6.0
+- llm_int8_skip_modules: None
+- llm_int8_enable_fp32_cpu_offload: False
+- llm_int8_has_fp16_weight: False
+- bnb_4bit_quant_type: nf4
+- bnb_4bit_use_double_quant: True
+- bnb_4bit_compute_dtype: bfloat16
+
 The following `bitsandbytes` quantization config was used during training:
 - load_in_8bit: False
 - load_in_4bit: True
@@ -16,5 +27,6 @@ The following `bitsandbytes` quantization config was used during training:
 - bnb_4bit_compute_dtype: bfloat16
 ### Framework versions
 
+- PEFT 0.4.0
 
 - PEFT 0.4.0

checkpoint-200/adapter_config.json

@@ -14,13 +14,13 @@
   "r": 64,
   "revision": null,
   "target_modules": [
-    "q_proj",
+    "v_proj",
     "down_proj",
+    "o_proj",
     "k_proj",
     "up_proj",
-    "o_proj",
-    "v_proj",
-    "gate_proj"
+    "gate_proj",
+    "q_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

checkpoint-200/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0075a06493c9d9a0f18d9ad6329effee1c7f8b6f8ee3b5692cae0a760ef1f224
+oid sha256:854d9775c36d691131435da90a76e8c4b814ee81dcbe15efd3d242c7247dc3a7
 size 319977229

checkpoint-200/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c6d60e63ee8cc0554e3a0586fdcd38dad48835b970995c3a185a94fb6495ab4
+oid sha256:73e0d06ba0cdde29b7199f4c0324389e562b8aa84731ad3ad18d266a98b751d1
 size 1279539525

checkpoint-200/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:76be2ab508564538497ab975a1084cf08fe26aab573d835d93a3eb59cbd9df91
+oid sha256:713ec39e0b730d1fa767bc8cdd9bb6284e517b29dcf47393cf5e8ff5feafa73a
 size 14511

checkpoint-200/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.7774001359939575,
+  "best_metric": 0.7779108881950378,
   "best_model_checkpoint": "experts/expert-16/checkpoint-200",
   "epoch": 0.06337135614702155,
   "global_step": 200,
@@ -10,7 +10,7 @@
     {
       "epoch": 0.0,
       "learning_rate": 0.0002,
-      "loss": 0.8331,
+      "loss": 0.8339,
       "step": 10
     },
     {
@@ -22,7 +22,7 @@
     {
       "epoch": 0.01,
       "learning_rate": 0.0002,
-      "loss": 0.9038,
+      "loss": 0.9041,
       "step": 30
     },
     {
@@ -34,13 +34,13 @@
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.8155,
+      "loss": 0.8151,
       "step": 50
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.7896,
+      "loss": 0.79,
       "step": 60
     },
     {
@@ -52,105 +52,105 @@
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.8827,
+      "loss": 0.8831,
       "step": 80
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.861,
+      "loss": 0.8607,
       "step": 90
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.7879,
+      "loss": 0.7876,
       "step": 100
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.803,
+      "loss": 0.8031,
       "step": 110
     },
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.8212,
+      "loss": 0.8207,
       "step": 120
     },
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.8075,
+      "loss": 0.807,
       "step": 130
     },
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.9263,
+      "loss": 0.9262,
       "step": 140
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.7969,
+      "loss": 0.7964,
       "step": 150
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.7883,
+      "loss": 0.7879,
       "step": 160
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.7582,
+      "loss": 0.7587,
       "step": 170
     },
     {
       "epoch": 0.06,
       "learning_rate": 0.0002,
-      "loss": 0.8095,
+      "loss": 0.8091,
       "step": 180
     },
     {
       "epoch": 0.06,
       "learning_rate": 0.0002,
-      "loss": 0.8614,
+      "loss": 0.8615,
       "step": 190
     },
     {
       "epoch": 0.06,
       "learning_rate": 0.0002,
-      "loss": 0.8674,
+      "loss": 0.8672,
       "step": 200
     },
     {
       "epoch": 0.06,
-      "eval_loss": 0.7774001359939575,
-      "eval_runtime": 149.8878,
-      "eval_samples_per_second": 6.672,
-      "eval_steps_per_second": 3.336,
+      "eval_loss": 0.7779108881950378,
+      "eval_runtime": 110.9863,
+      "eval_samples_per_second": 9.01,
+      "eval_steps_per_second": 4.505,
       "step": 200
     },
     {
       "epoch": 0.06,
-      "mmlu_eval_accuracy": 0.4759538024775667,
+      "mmlu_eval_accuracy": 0.4744171116325413,
       "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
       "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
       "mmlu_eval_accuracy_astronomy": 0.4375,
       "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
-      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
       "mmlu_eval_accuracy_college_biology": 0.4375,
-      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
       "mmlu_eval_accuracy_college_computer_science": 0.18181818181818182,
       "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
       "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
       "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
       "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
-      "mmlu_eval_accuracy_conceptual_physics": 0.3076923076923077,
+      "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
       "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
       "mmlu_eval_accuracy_electrical_engineering": 0.25,
       "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
@@ -159,16 +159,16 @@
       "mmlu_eval_accuracy_high_school_biology": 0.375,
       "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
       "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
-      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
       "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
       "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
       "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
       "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
-      "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.3076923076923077,
       "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
       "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
-      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
-      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
       "mmlu_eval_accuracy_high_school_world_history": 0.7692307692307693,
       "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
       "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
@@ -177,25 +177,25 @@
       "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
       "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
       "mmlu_eval_accuracy_management": 0.6363636363636364,
-      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_marketing": 0.88,
       "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
-      "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
       "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
-      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_moral_scenarios": 0.27,
       "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
       "mmlu_eval_accuracy_philosophy": 0.5,
       "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
       "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
-      "mmlu_eval_accuracy_professional_law": 0.3058823529411765,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
       "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
-      "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
+      "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
       "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
       "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
-      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
       "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
       "mmlu_eval_accuracy_virology": 0.5555555555555556,
-      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
-      "mmlu_loss": 1.619096915108728,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 1.5868234255450824,
       "step": 200
     }
   ],

checkpoint-200/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d55ea0037adc27e709c2f87c727ac3fd55b607c5afed1c76d4262111f780937e
+oid sha256:277dab2895cbf8d8b713e5f0339719c5d50d193c9afc1399ad953aa4910a2fab
 size 5819

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d55ea0037adc27e709c2f87c727ac3fd55b607c5afed1c76d4262111f780937e
+oid sha256:277dab2895cbf8d8b713e5f0339719c5d50d193c9afc1399ad953aa4910a2fab
 size 5819