Add llama-unsloth to zero-shot eval

data/experiments/predict/pl-court-instruct/.gitignore

@@ -7,3 +7,4 @@
 /outputs_Unsloth-Llama-3-8B-Instruct-fine-tuned.json
 /outputs_Unsloth-Mistral-7B-Instruct-v0.3.json
 /outputs_Unsloth-Mistral-7B-Instruct-v0.3-fine-tuned.json
+/outputs_Unsloth-Llama-3-8B-Instruct.json

data/experiments/predict/pl-court-instruct/metrics_Unsloth-Llama-3-8B-Instruct.json

@@ -0,0 +1,12 @@
+{
+	"full_text_chrf": 0.4385761320590973,
+	"field_chrf": {
+		"court_name": 0.8789530396461487,
+		"date": 0.9822721481323242,
+		"department_name": 0.9057374000549316,
+		"judges": 0.9149863123893738,
+		"legal_bases": 0.42645466327667236,
+		"recorder": 0.7640316486358643,
+		"signature": 0.7549777626991272
+	}
+}

data/experiments/predict/pl-court-instruct/metrics_summary.md

@@ -3,6 +3,7 @@
 | Meta-Llama-3-8B-Instruct                    |            0.247 |        0.862 |  0.971 |             0.833 |    0.882 |         0.287 |      0.805 |       0.778 |
 | Mistral-7B-Instruct-v0.2                    |            0.432 |        0.839 |  0.922 |             0.850 |    0.879 |         0.333 |      0.837 |       0.145 |
 | Mistral-7B-Instruct-v0.2-fine-tuned         |            0.772 |        0.987 |  0.990 |             0.965 |    0.952 |         0.600 |      0.979 |       0.972 |
+| Unsloth-Llama-3-8B-Instruct                 |            0.439 |        0.879 |  0.982 |             0.906 |    0.915 |         0.426 |      0.764 |       0.755 |
 | Unsloth-Llama-3-8B-Instruct-fine-tuned      |            0.828 |        0.995 |  0.989 |             0.986 |    0.977 |         0.601 |      0.993 |       0.994 |
 | Unsloth-Mistral-7B-Instruct-v0.3            |            0.477 |        0.830 |  0.987 |             0.900 |    0.870 |         0.419 |      0.943 |       0.567 |
 | Unsloth-Mistral-7B-Instruct-v0.3-fine-tuned |            0.798 |        0.995 |  0.988 |             0.986 |    0.967 |         0.608 |      0.987 |       0.976 |

dvc.lock

@@ -320,8 +320,8 @@ stages:
     outs:
     - path: data/experiments/predict/pl-court-instruct/metrics_summary.md
       hash: md5
-      md5: f6ef483c69060955ee54526323083b4e
-      size: 1319
+      md5: b4e10d25a74b97bc3f93bc72d6055015
+      size: 1484
   [email protected]:
     cmd: PYTHONPATH=. python scripts/sft/evaluate.py  --output-file 
       data/experiments/predict/pl-court-instruct/outputs_Unsloth-Mistral-7B-Instruct-v0.3.json
@@ -381,3 +381,43 @@ stages:
       hash: md5
       md5: 6e6330c14b645d1a65115546a8127484
       size: 306
+  predict@Unsloth-Llama-3-8B-Instruct:
+    cmd: PYTHONPATH=. python scripts/sft/predict.py model=Unsloth-Llama-3-8B-Instruct
+    deps:
+    - path: configs/model/Unsloth-Llama-3-8B-Instruct.yaml
+      hash: md5
+      md5: e97bb2e6bf39f75edea7714d6ba58b77
+      size: 160
+    - path: configs/predict.yaml
+      hash: md5
+      md5: e6b047cf62e612a32381d6221eb99b4e
+      size: 416
+    - path: scripts/sft/predict.py
+      hash: md5
+      md5: 69e4844a715c9c5c75e1127a06472ad4
+      size: 3148
+    outs:
+    - path: 
+        data/experiments/predict/pl-court-instruct/outputs_Unsloth-Llama-3-8B-Instruct.json
+      hash: md5
+      md5: df2f1d464152f87737c8ebb5b0673854
+      size: 2179383
+  evaluate@Unsloth-Llama-3-8B-Instruct:
+    cmd: PYTHONPATH=. python scripts/sft/evaluate.py  --output-file 
+      data/experiments/predict/pl-court-instruct/outputs_Unsloth-Llama-3-8B-Instruct.json
+    deps:
+    - path: 
+        data/experiments/predict/pl-court-instruct/outputs_Unsloth-Llama-3-8B-Instruct.json
+      hash: md5
+      md5: df2f1d464152f87737c8ebb5b0673854
+      size: 2179383
+    - path: scripts/sft/evaluate.py
+      hash: md5
+      md5: 5ee442a9a3525af7596bf24c3d724a1d
+      size: 570
+    outs:
+    - path: 
+        data/experiments/predict/pl-court-instruct/metrics_Unsloth-Llama-3-8B-Instruct.json
+      hash: md5
+      md5: 521a731cc2c45d3eda0656a8e69d505b
+      size: 307

dvc.yaml

@@ -58,6 +58,7 @@ stages:
         - Meta-Llama-3-8B-Instruct
         - Mistral-7B-Instruct-v0.2
         - Mistral-7B-Instruct-v0.2-fine-tuned
+        - Unsloth-Llama-3-8B-Instruct
         - Unsloth-Llama-3-8B-Instruct-fine-tuned
         - Unsloth-Mistral-7B-Instruct-v0.3
         - Unsloth-Mistral-7B-Instruct-v0.3-fine-tuned
@@ -73,9 +74,7 @@ stages:
   evaluate:
     matrix:
       model:
-        - Meta-Llama-3-8B-Instruct
-        - Mistral-7B-Instruct-v0.2
-        - Mistral-7B-Instruct-v0.2-fine-tuned
+        - Unsloth-Llama-3-8B-Instruct
         - Unsloth-Llama-3-8B-Instruct-fine-tuned
         - Unsloth-Mistral-7B-Instruct-v0.3
         - Unsloth-Mistral-7B-Instruct-v0.3-fine-tuned