Training in progress, step 200

adapter_config.json

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

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:53281bd7b359d3310f6ea45b1405573e0a5bf41ac9e9df912712d706c2adbe8f
+oid sha256:4c0c4efebd4a76a1a424ffbc4649470e5c2cdaf06d9f3c6510552423f88b0b7f
 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": [
-    "k_proj",
-    "down_proj",
-    "o_proj",
-    "q_proj",
     "gate_proj",
     "up_proj",
-    "v_proj"
+    "o_proj",
+    "k_proj",
+    "v_proj",
+    "q_proj",
+    "down_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

checkpoint-200/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:940420e857b16d9d460804b2fc6046f7c2306729caa496cd3c3574ad0be0fa3b
+oid sha256:4c0c4efebd4a76a1a424ffbc4649470e5c2cdaf06d9f3c6510552423f88b0b7f
 size 319977229

checkpoint-200/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5dca35ae333f259f26be71eebfab93f0c9b67e7b6a6554d0ce2dbf638abb966b
+oid sha256:b00ea9d55f649fa8c2c6320b045750babf5317c933d12463aa968c2e4a588f40
 size 1279539525

checkpoint-200/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3a791886a5bd3544204ccb1a2a0ac20c168c0f6b19331526cd745b6faff90e9c
+oid sha256:d90095d424eaa9891e7bcd0d220b52ea79e5466d6fd1503d15c42729f5f70f56
 size 14511

checkpoint-200/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.7692526578903198,
+  "best_metric": 0.7692113518714905,
   "best_model_checkpoint": "experts/expert-26/checkpoint-200",
   "epoch": 0.05355469272995046,
   "global_step": 200,
@@ -10,25 +10,25 @@
     {
       "epoch": 0.0,
       "learning_rate": 0.0002,
-      "loss": 0.8329,
+      "loss": 0.8327,
       "step": 10
     },
     {
       "epoch": 0.01,
       "learning_rate": 0.0002,
-      "loss": 0.8048,
+      "loss": 0.805,
       "step": 20
     },
     {
       "epoch": 0.01,
       "learning_rate": 0.0002,
-      "loss": 0.8108,
+      "loss": 0.811,
       "step": 30
     },
     {
       "epoch": 0.01,
       "learning_rate": 0.0002,
-      "loss": 0.7864,
+      "loss": 0.7862,
       "step": 40
     },
     {
@@ -40,13 +40,13 @@
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.7645,
+      "loss": 0.7647,
       "step": 60
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.757,
+      "loss": 0.7572,
       "step": 70
     },
     {
@@ -58,25 +58,25 @@
     {
       "epoch": 0.02,
       "learning_rate": 0.0002,
-      "loss": 0.7777,
+      "loss": 0.7775,
       "step": 90
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.7482,
+      "loss": 0.7483,
       "step": 100
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.7402,
+      "loss": 0.7404,
       "step": 110
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0002,
-      "loss": 0.7442,
+      "loss": 0.7444,
       "step": 120
     },
     {
@@ -94,83 +94,83 @@
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.7632,
+      "loss": 0.763,
       "step": 150
     },
     {
       "epoch": 0.04,
       "learning_rate": 0.0002,
-      "loss": 0.7882,
+      "loss": 0.7881,
       "step": 160
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.7759,
+      "loss": 0.7757,
       "step": 170
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.8138,
+      "loss": 0.8139,
       "step": 180
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.7874,
+      "loss": 0.7873,
       "step": 190
     },
     {
       "epoch": 0.05,
       "learning_rate": 0.0002,
-      "loss": 0.7547,
+      "loss": 0.7542,
       "step": 200
     },
     {
       "epoch": 0.05,
-      "eval_loss": 0.7692526578903198,
-      "eval_runtime": 85.9008,
-      "eval_samples_per_second": 11.641,
-      "eval_steps_per_second": 5.821,
+      "eval_loss": 0.7692113518714905,
+      "eval_runtime": 63.4219,
+      "eval_samples_per_second": 15.767,
+      "eval_steps_per_second": 7.884,
       "step": 200
     },
     {
       "epoch": 0.05,
-      "mmlu_eval_accuracy": 0.4937716133877423,
+      "mmlu_eval_accuracy": 0.49597563739434053,
       "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": 0.6363636363636364,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
       "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
       "mmlu_eval_accuracy_college_biology": 0.5,
       "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": 0.4090909090909091,
+      "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": 0.4230769230769231,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "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": 0.36585365853658536,
-      "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
       "mmlu_eval_accuracy_global_facts": 0.6,
       "mmlu_eval_accuracy_high_school_biology": 0.375,
-      "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
+      "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": 0.4418604651162791,
       "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
-      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
       "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
-      "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
-      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
-      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
       "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
-      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
       "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
       "mmlu_eval_accuracy_international_law": 0.8461538461538461,
       "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
@@ -180,22 +180,22 @@
       "mmlu_eval_accuracy_marketing": 0.92,
       "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
       "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
-      "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
-      "mmlu_eval_accuracy_moral_scenarios": 0.23,
-      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_moral_disputes": 0.39473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.21,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
       "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
-      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
       "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
       "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
-      "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
-      "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
+      "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
+      "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
       "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
       "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
       "mmlu_eval_accuracy_sociology": 0.7272727272727273,
       "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
-      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
       "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
-      "mmlu_loss": 1.2817495326382684,
+      "mmlu_loss": 1.286679525755093,
       "step": 200
     }
   ],

checkpoint-200/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1fdce65f27e1b27599ffadb916e306ae3c8008257aa650aa283cc44ffd79ab85
+oid sha256:fe4dd9a1fee85d1d6fa31030cdb57a7342525f029464f99168fb3a0bf75117ad
 size 5819

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1fdce65f27e1b27599ffadb916e306ae3c8008257aa650aa283cc44ffd79ab85
+oid sha256:fe4dd9a1fee85d1d6fa31030cdb57a7342525f029464f99168fb3a0bf75117ad
 size 5819