Training in progress, step 200

adapter_config.json

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

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:71906cb7cd37c700161633e12f19362e227d9357485fedea7fef715889cb6657
+oid sha256:9b304c75d2fe950dd03ec74af4fdec1e8843d6732e5fe7bf3700ba7e03fcf3be
 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": [
-    "v_proj",
-    "down_proj",
-    "gate_proj",
     "up_proj",
+    "q_proj",
     "k_proj",
-    "o_proj",
-    "q_proj"
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "o_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

checkpoint-200/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:02f5290587d3eefd596838b4bfefb444cae13257dedb81baa7bdd283eb97d590
+oid sha256:9b304c75d2fe950dd03ec74af4fdec1e8843d6732e5fe7bf3700ba7e03fcf3be
 size 319977229

checkpoint-200/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d29f27fb930230ffd6ea6f0a1639297c854e9531ee60eacd2b8b4c0ef4047dbd
+oid sha256:615c84ce32cd736413f4264f6ca7835a688c96be8de0e506a24dd2ade9ea58aa
 size 1279539525

checkpoint-200/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f82cc9eb0b5433c918393971f9b0774b51dae507d7a4f06a2e26ae2620a7d09
+oid sha256:10f4993ff2506f03c0bd800f239dd4aaf62ed8cad781e7d41a9f1cbb4fb66002
 size 14511

checkpoint-200/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.8420786261558533,
+  "best_metric": 0.8420215845108032,
   "best_model_checkpoint": "experts/expert-22/checkpoint-200",
   "epoch": 0.0769897026272736,
   "global_step": 200,
@@ -10,13 +10,13 @@
     {
       "epoch": 0.0,
       "learning_rate": 0.0002,
-      "loss": 0.9402,
+      "loss": 0.9409,
       "step": 10
     },
     {
       "epoch": 0.01,
       "learning_rate": 0.0002,
-      "loss": 0.9159,
+      "loss": 0.9155,
       "step": 20
     },
     {
@@ -28,25 +28,25 @@
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.8795,
+      "loss": 0.8794,
       "step": 40
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.9281,
+      "loss": 0.9277,
       "step": 50
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.8639,
+      "loss": 0.864,
       "step": 60
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.9537,
+      "loss": 0.953,
       "step": 70
     },
     {
@@ -64,138 +64,138 @@
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.8651,
+      "loss": 0.8647,
       "step": 100
     },
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.9332,
+      "loss": 0.9335,
       "step": 110
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.8883,
+      "loss": 0.8879,
       "step": 120
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.8945,
+      "loss": 0.8953,
       "step": 130
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.8675,
+      "loss": 0.868,
       "step": 140
     },
     {
       "epoch": 0.06,
       "learning_rate": 0.0002,
-      "loss": 0.8583,
+      "loss": 0.8585,
       "step": 150
     },
     {
       "epoch": 0.06,
       "learning_rate": 0.0002,
-      "loss": 0.8992,
+      "loss": 0.8998,
       "step": 160
     },
     {
       "epoch": 0.07,
       "learning_rate": 0.0002,
-      "loss": 0.8667,
+      "loss": 0.8672,
       "step": 170
     },
     {
       "epoch": 0.07,
       "learning_rate": 0.0002,
-      "loss": 0.8367,
+      "loss": 0.8373,
       "step": 180
     },
     {
       "epoch": 0.07,
       "learning_rate": 0.0002,
-      "loss": 0.8495,
+      "loss": 0.8492,
       "step": 190
     },
     {
       "epoch": 0.08,
       "learning_rate": 0.0002,
-      "loss": 0.8338,
+      "loss": 0.8342,
       "step": 200
     },
     {
       "epoch": 0.08,
-      "eval_loss": 0.8420786261558533,
-      "eval_runtime": 104.6029,
-      "eval_samples_per_second": 9.56,
-      "eval_steps_per_second": 4.78,
+      "eval_loss": 0.8420215845108032,
+      "eval_runtime": 77.6159,
+      "eval_samples_per_second": 12.884,
+      "eval_steps_per_second": 6.442,
       "step": 200
     },
     {
       "epoch": 0.08,
-      "mmlu_eval_accuracy": 0.49890909250031423,
+      "mmlu_eval_accuracy": 0.5020690882678942,
       "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
       "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
       "mmlu_eval_accuracy_astronomy": 0.4375,
       "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
-      "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
-      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.5,
       "mmlu_eval_accuracy_college_chemistry": 0.25,
       "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
       "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
       "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
       "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
       "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
-      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
       "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
-      "mmlu_eval_accuracy_electrical_engineering": 0.25,
-      "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
       "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
       "mmlu_eval_accuracy_global_facts": 0.6,
-      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.53125,
       "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
       "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
-      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "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": 0.47619047619047616,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
       "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
       "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
-      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
       "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
-      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8833333333333333,
       "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
       "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
       "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
       "mmlu_eval_accuracy_human_aging": 0.782608695652174,
       "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
       "mmlu_eval_accuracy_international_law": 0.8461538461538461,
-      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
       "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
       "mmlu_eval_accuracy_machine_learning": 0.45454545454545453,
       "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": 0.6627906976744186,
-      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
       "mmlu_eval_accuracy_moral_scenarios": 0.21,
-      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
-      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
       "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
-      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
-      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
       "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
-      "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
+      "mmlu_eval_accuracy_professional_psychology": 0.4782608695652174,
       "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
       "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
       "mmlu_eval_accuracy_sociology": 0.7272727272727273,
       "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
       "mmlu_eval_accuracy_virology": 0.5,
       "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
-      "mmlu_loss": 1.4345150615621174,
+      "mmlu_loss": 1.42281840854463,
       "step": 200
     }
   ],

checkpoint-200/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:68347c5c0bc67308405eaf249524c55462e787942d6d74b927105955b931cf02
+oid sha256:376563167fb75ef06e6c8600f9d035897ffea2363fc5fd45e427161c33ec50e3
 size 5819

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:68347c5c0bc67308405eaf249524c55462e787942d6d74b927105955b931cf02
+oid sha256:376563167fb75ef06e6c8600f9d035897ffea2363fc5fd45e427161c33ec50e3
 size 5819