Training in progress, step 200

adapter_config.json

@@ -14,10 +14,10 @@
   "r": 64,
   "revision": null,
   "target_modules": [
-    "o_proj",
+    "v_proj",
     "q_proj",
+    "o_proj",
     "down_proj",
-    "v_proj",
     "k_proj",
     "gate_proj",
     "up_proj"

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:efdc2a905cceb1a9c3bfe194c1b9885e9dfdbbe4afd906a34a44a8c1628ee9e5
+oid sha256:a2b10da90a336f67a541e93e611dc8cf785e6594bf22c830698a608b4695e8f1
 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

@@ -14,10 +14,10 @@
   "r": 64,
   "revision": null,
   "target_modules": [
-    "o_proj",
+    "v_proj",
     "q_proj",
+    "o_proj",
     "down_proj",
-    "v_proj",
     "k_proj",
     "gate_proj",
     "up_proj"

checkpoint-200/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:951f0410a5b1b895cd7f4ac9717b56d77a66ac1e7f064443ac97a17ac881c0c0
+oid sha256:a2b10da90a336f67a541e93e611dc8cf785e6594bf22c830698a608b4695e8f1
 size 319977229

checkpoint-200/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4e14849e6ad46b1dc5832e2fd13fe921afb7ac60c46b23a430a6801ffc5761ea
+oid sha256:0b9d08d7292e7785801f823e4c80ca66bf9349510885d5641c81dd2c8402310f
 size 1279539525

checkpoint-200/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d440b345926143ca36c3e807d6f422c49a3f1e63c7f15c261f9777fbc7a2bbc4
+oid sha256:494626a050cceac1c0023f61e6deff68d6953b3cd5f8b62db8ab0c97dd59334e
 size 14511

checkpoint-200/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.9700731039047241,
+  "best_metric": 0.970292329788208,
   "best_model_checkpoint": "experts/expert-14/checkpoint-200",
   "epoch": 0.06151480199923107,
   "global_step": 200,
@@ -10,13 +10,13 @@
     {
       "epoch": 0.0,
       "learning_rate": 0.0002,
-      "loss": 1.0634,
+      "loss": 1.064,
       "step": 10
     },
     {
       "epoch": 0.01,
       "learning_rate": 0.0002,
-      "loss": 1.0357,
+      "loss": 1.0358,
       "step": 20
     },
     {
@@ -28,43 +28,43 @@
     {
       "epoch": 0.01,
       "learning_rate": 0.0002,
-      "loss": 0.9858,
+      "loss": 0.9861,
       "step": 40
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 1.0149,
+      "loss": 1.0148,
       "step": 50
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 1.0292,
+      "loss": 1.0293,
       "step": 60
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.9947,
+      "loss": 0.9945,
       "step": 70
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 1.0068,
+      "loss": 1.0069,
       "step": 80
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 1.015,
+      "loss": 1.0147,
       "step": 90
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 1.0055,
+      "loss": 1.0056,
       "step": 100
     },
     {
@@ -76,37 +76,37 @@
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 1.0218,
+      "loss": 1.0214,
       "step": 120
     },
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.9645,
+      "loss": 0.9642,
       "step": 130
     },
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 1.0059,
+      "loss": 1.0051,
       "step": 140
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 1.0041,
+      "loss": 1.004,
       "step": 150
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.9626,
+      "loss": 0.9631,
       "step": 160
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.9682,
+      "loss": 0.9674,
       "step": 170
     },
     {
@@ -118,74 +118,74 @@
     {
       "epoch": 0.06,
       "learning_rate": 0.0002,
-      "loss": 1.0029,
+      "loss": 1.0023,
       "step": 190
     },
     {
       "epoch": 0.06,
       "learning_rate": 0.0002,
-      "loss": 0.9873,
+      "loss": 0.9879,
       "step": 200
     },
     {
       "epoch": 0.06,
-      "eval_loss": 0.9700731039047241,
-      "eval_runtime": 90.2222,
-      "eval_samples_per_second": 11.084,
-      "eval_steps_per_second": 5.542,
+      "eval_loss": 0.970292329788208,
+      "eval_runtime": 67.4255,
+      "eval_samples_per_second": 14.831,
+      "eval_steps_per_second": 7.416,
       "step": 200
     },
     {
       "epoch": 0.06,
-      "mmlu_eval_accuracy": 0.4896808303512203,
+      "mmlu_eval_accuracy": 0.49145352624106475,
       "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
-      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
       "mmlu_eval_accuracy_astronomy": 0.4375,
       "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
-      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
       "mmlu_eval_accuracy_college_biology": 0.3125,
       "mmlu_eval_accuracy_college_chemistry": 0.25,
       "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
       "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
       "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
       "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
-      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
       "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
       "mmlu_eval_accuracy_econometrics": 0.25,
       "mmlu_eval_accuracy_electrical_engineering": 0.3125,
       "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
-      "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
-      "mmlu_eval_accuracy_global_facts": 0.6,
-      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
       "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_geography": 0.9090909090909091,
       "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
       "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
-      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
       "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
       "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
-      "mmlu_eval_accuracy_high_school_psychology": 0.85,
-      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
-      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
-      "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
-      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
       "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
       "mmlu_eval_accuracy_international_law": 0.8461538461538461,
       "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
       "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
       "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
       "mmlu_eval_accuracy_management": 0.6363636363636364,
-      "mmlu_eval_accuracy_marketing": 0.88,
+      "mmlu_eval_accuracy_marketing": 0.84,
       "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
       "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
-      "mmlu_eval_accuracy_moral_disputes": 0.39473684210526316,
-      "mmlu_eval_accuracy_moral_scenarios": 0.23,
-      "mmlu_eval_accuracy_nutrition": 0.696969696969697,
-      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
       "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
-      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
       "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
       "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
       "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
@@ -195,7 +195,7 @@
       "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
       "mmlu_eval_accuracy_virology": 0.4444444444444444,
       "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
-      "mmlu_loss": 1.6262522289242507,
+      "mmlu_loss": 1.6139244634862357,
       "step": 200
     }
   ],

checkpoint-200/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:132dc0489795dd2e62e3dfed60f8fa6c2da9f513c3d6724e00171983c97e1c5b
+oid sha256:6c1f3b4e86a18f376afc61102db182028a2b670075c0fd60d0651751e2c0b01a
 size 5819

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:132dc0489795dd2e62e3dfed60f8fa6c2da9f513c3d6724e00171983c97e1c5b
+oid sha256:6c1f3b4e86a18f376afc61102db182028a2b670075c0fd60d0651751e2c0b01a
 size 5819