Upload 10 files

generative/config.json

@@ -0,0 +1,238 @@
+{
+    "datasets": {
+        "xnli-eu-native": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-eu-native.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        },
+        "xnli-eu-var": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-eu-var.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        },
+        "xnli-es-native": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/es/xnli-eu2es-native.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-es-var": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/es/xnli-es-var.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        },
+        "xnli-en": {
+            "data_path": "/tartalo01/users/jbengoetxea004/phd/xnli-paraphrasing/xnli-var-decoders/scripts/parquet-con/xnli-en-test.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-es": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/es/xnli-es-original.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        },
+        "xnli-eu": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-eu-original.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-eu-var-no-rep": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-native-var-eu-NO-REPETITION.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        },
+        "xnli-eu-var-less-gip": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-native-var-eu-less-gip.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-eu-var-less-biz": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/eu/xnli-native-var-eu-less-biz.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        },
+        "xnli-es-var-no-rep": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/es/xnli-native-var-es-no-rep.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-eu-biz": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-test-bizkaiera-done.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-eu-gip": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-test-gipuzkera-done.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-eu-naf": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-test-nafar-lapurtera-done.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        },
+        "xnli-eu-nat-biz": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-native-bizkaieraz-done.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-eu-nat-gip": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-native-gipuzkera-done.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }, 
+        "xnli-eu-nat-naf": {
+            "data_path": "/scratch/jbengoetxea/phd/XNLIvar/data/test_expanded/xnli-eu-native-nafar-lapurtera-done.tsv",
+            "data_path_paraphrase": "",
+            "prem_col": "premise",
+            "hyp_col": "hypothesis",
+            "label_col": "label",
+            "prompts": ["trilabel", "qa-zero", "qa-few"]
+        }
+    },
+    "models": {
+        "llama3instruct8": "meta-llama/Meta-Llama-3-8B-Instruct", 
+        "llama3instruct70": "meta-llama/Meta-Llama-3-70B-Instruct", 
+        "gemmainstruct9": "google/gemma-2-9b-it", 
+        "gemmainstruct27": "google/gemma-2-27b-it", 
+        "latxainstruct70": "HiTZ/Latxa-Llama-3.1-70B-Instruct", 
+        "llama3base70": "meta-llama/Meta-Llama-3.1-70B"
+    },
+    "prompts": {
+        "trilabel": {
+            "nli-zero": {
+                "preffix": "Please, answer in one word, with one of the following labels: <entailment>, <contradiction> or <neutral> Use exactly one of these three labels.",
+                "label_mapping": {
+                    "entailment": "entailment",
+                    "contradiction": "contradiction",
+                    "neutral": "neutral"
+                }
+            },
+            "nli-few": {
+                "preffix": "Say which is the inference relationship between these two sentences. Please, answer in one word, with one of the following labels: <entailment>, <contradiction> or <neutral> Use exactly one of these three labels. Here you have some examples: Postal Service were to reduce delivery frequency ->  The postal service could deliver less frequently: <entailment>. This elegant spa town on the edge of the Lac du Bourget has offered cures for rheumatism and other ailments for centuries -> The town was only established in the past fifty years: <contradiction>. And while we allow people to give a kidney to their child , we do not allow them to donate their heart -> You can't always donate organs to your child: <neutral>. " ,
+                "label_mapping": {
+                    "entailment": "entailment",
+                    "contradiction": "contradiction",
+                    "neutral": "neutral"
+                }
+            },
+            "qa-zero": {
+                "preffix": "Are these two sentences entailed, contradicted or undetermined to each other? Please, answer in one word, with one of the following labels: <entailment>, <contradiction> or <neutral> Use exactly one of these three labels.",
+                "label_mapping": {
+                    "entailment": "entailment",
+                    "contradiction": "contradiction",
+                    "neutral": "neutral"
+                }
+            },
+            "qa-few": {
+                "preffix": "Are these two sentences entailed, contradicted or undetermined to each other? Please, answer in one word, with one of the following labels: <entailment>, <contradiction> or <neutral> Use exactly one of these three labels. Here you have some examples: Postal Service were to reduce delivery frequency -> The postal service could deliver less frequently: <entailment>. This elegant spa town on the edge of the Lac du Bourget has offered cures for rheumatism and other ailments for centuries -> The town was only established in the past fifty years: <contradiction>. And while we allow people to give a kidney to their child , we do not allow them to donate their heart -> You can't always donate organs to your child: <neutral>.",
+                "label_mapping": {
+                    "entailment": "entailment",
+                    "contradiction": "contradiction",
+                    "neutral": "neutral"
+                }
+            },
+            "chain": {
+                "preffix": "You are an expert linguist and your task is to annotate sentences for the task of Natural Language Inference. This task consists in determining if a first sentence (premise) entails, contradicts or does not entail nor contradict the second sentence (hypothesis). Please, answer in one word, with one of the following labels: <entailment>, <contradiction> or <neutral> \n Use exactly one of these three labels \n Here you have a few examples:\n Premise: Postal Service were to reduce delivery frequency. \n Hypothesis: The postal service could deliver less frequently. \n Answer: <entailment> \n Premise: This elegant spa town on the edge of the Lac du Bourget has offered cures for rheumatism and other ailments for centuries. \n Hypothesis: The town was only established in the past fifty years. \n Answer: <contradiction> \n Premise: And while we allow people to give a kidney to their child , we do not allow them to donate their heart. \n Hypothesis: You can't always donate organs to your child. \n Answer: <neutral>",
+                "label_mapping": {
+                    "entailment": "entailment",
+                    "contradiction": "contradiction",
+                    "neutral": "neutral"
+                }
+            }
+        }, 
+        "qa-zero": {
+            "entailment": {
+                "preffix": "Are these two sentences entailed? Please, answer between <yes> or <no>.",
+                "label_mapping": {
+                    "yes": "entailment",
+                    "no": "not_entailment"
+                }
+            },
+            "contradiction": {
+                "preffix": "Are these two sentences contradictions? Please, answer between <yes> or <no>.",
+                "label_mapping": {
+                    "yes": "contradiction",
+                    "no": "not_contradiction"
+                }
+            },
+            "neutral": {
+                "preffix": "Are these two sentences unrelated? Please, answer between <yes> or <no>.",
+                "label_mapping": {
+                    "yes": "neutral",
+                    "no": "not_neutral"
+                }
+            }
+        },
+        "qa-few": {
+            "entailment": {
+                "preffix": "Are these two sentences entailed? Please, answer between <yes> or <no>. Here you have some examples: Postal Service were to reduce delivery frequency -> The postal service could deliver less frequently: <yes>. This elegant spa town on the edge of the Lac du Bourget has offered cures for rheumatism and other ailments for centuries -> The town was only established in the past fifty years: <no>. And while we allow people to give a kidney to their child , we do not allow them to donate their heart -> You can't always donate organs to your child: <no>.",
+                "label_mapping": {
+                    "yes": "entailment",
+                    "no": "not_entailment"
+                }
+            },
+            "contradiction": {
+                "preffix": "Are these two sentences contradictions? Please, answer between <yes> or <no>. Here you have some examples: Postal Service were to reduce delivery frequency -> The postal service could deliver less frequently: <no>. This elegant spa town on the edge of the Lac du Bourget has offered cures for rheumatism and other ailments for centuries -> The town was only established in the past fifty years: <yes>. And while we allow people to give a kidney to their child , we do not allow them to donate their heart -> You can't always donate organs to your child: <no>.",
+                "label_mapping": {
+                    "yes": "contradiction",
+                    "no": "not_contradiction"
+                }
+            },
+            "neutral": {
+                "preffix": "Are these two sentences unrelated? Please, answer between <yes> or <no>. Here you have some examples: Postal Service were to reduce delivery frequency -> The postal service could deliver less frequently: <no>. This elegant spa town on the edge of the Lac du Bourget has offered cures for rheumatism and other ailments for centuries -> The town was only established in the past fifty years: <no>. And while we allow people to give a kidney to their child , we do not allow them to donate their heart -> You can't always donate organs to your child: <yes>.",
+                "label_mapping": {
+                    "yes": "neutral",
+                    "no": "not_neutral"
+                }
+            }
+        }
+    }
+}

generative/requirements.txt

@@ -0,0 +1,49 @@
+accelerate==1.6.0
+certifi==2025.4.26
+charset-normalizer==3.4.2
+filelock==3.18.0
+fsspec==2025.3.2
+huggingface-hub==0.31.2
+idna==3.10
+Jinja2==3.1.3
+joblib==1.5.0
+MarkupSafe==2.1.5
+mpmath==1.3.0
+networkx==3.3
+numpy==2.2.5
+nvidia-cublas-cu11==11.11.3.6
+nvidia-cuda-cupti-cu11==11.8.87
+nvidia-cuda-nvrtc-cu11==11.8.89
+nvidia-cuda-runtime-cu11==11.8.89
+nvidia-cudnn-cu11==9.1.0.70
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.3.0.86
+nvidia-cusolver-cu11==11.4.1.48
+nvidia-cusparse-cu11==11.7.5.86
+nvidia-nccl-cu11==2.21.5
+nvidia-nvtx-cu11==11.8.86
+packaging==25.0
+pandas==2.2.3
+pillow==11.0.0
+psutil==7.0.0
+python-dateutil==2.9.0.post0
+pytz==2025.2
+PyYAML==6.0.2
+regex==2024.11.6
+requests==2.32.3
+safetensors==0.5.3
+scikit-learn==1.6.1
+scipy==1.15.3
+six==1.17.0
+sympy==1.13.3
+threadpoolctl==3.6.0
+tokenizers==0.21.1
+torch==2.7.0+cu118
+torchaudio==2.7.0+cu118
+torchvision==0.22.0+cu118
+tqdm==4.67.1
+transformers @ git+https://github.com/huggingface/transformers@b311a3f50697c9602cc5d13a5faf7f6059c392ca
+triton==3.3.0
+typing_extensions==4.13.2
+tzdata==2025.2
+urllib3==2.4.0

generative/scripts/array-gemma.sh

@@ -0,0 +1,50 @@
+#!/usr/bin/env bash
+#SBATCH --qos=regular
+#SBATCH --job-name=xnli_gemmainstruct27
+#SBATCH --cpus-per-task=4
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:4
+#SBATCH --constraint=a100-sxm4
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli_gemmainstruct27_%a.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli_gemmainstruct27_%a.err
+#SBATCH --time=01:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+#SBATCH --array=0-8%2
+
+source /scratch/jbengoetxea/phd/.gemma_env/bin/activate
+
+export TRANSFORMERS_CACHE="/scratch/jbengoetxea/.cache"
+
+# Values for the 2 loops:
+# DATASET_VALUES=(xnli-eu-var xnli-eu-native xnli-eu xnli-es-var xnli-es-native xnli-es)  
+# PROMPT_TYPE_VALUES=(chain nli-zero nli-few qa-zero qa-few)         
+
+
+DATASET_VALUES=(xnli-eu-biz xnli-eu-gip xnli-eu-naf)  
+PROMPT_TYPE_VALUES=(chain nli-zero nli-few)
+
+
+N=${#PROMPT_TYPE_VALUES[@]}  # Number of items in the second level (VALUES2)
+
+# Decode SLURM_ARRAY_TASK_ID to get the two indices
+IDX1=$((SLURM_ARRAY_TASK_ID / N))
+IDX2=$((SLURM_ARRAY_TASK_ID % N))
+
+# Use IDX1 and IDX2 for your two-level loops
+DATASET="${DATASET_VALUES[${IDX1}]}"
+PROMPT_TYPE="${PROMPT_TYPE_VALUES[${IDX2}]}"
+
+
+TASK=trilabel
+MODEL=gemmainstruct27
+OUTPUT=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/results/$DATASET/$MODEL
+
+python3 /scratch/jbengoetxea/phd/XNLIvar/scripts/generative/scripts/zero_shot.py \
+    --dataset "${DATASET}" \
+    --model $MODEL \
+    --output_dir $OUTPUT \
+    --task $TASK \
+    --prompt_type "${PROMPT_TYPE}" 

generative/scripts/array-latxa.sh

@@ -0,0 +1,46 @@
+#!/usr/bin/env bash
+#SBATCH --qos=regular
+#SBATCH --job-name=xnli_latxainstruct70
+#SBATCH --cpus-per-task=2
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:4
+#SBATCH --constraint=a100-sxm4
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-eu-var_latxainstruct70_%a.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-eu-var_latxainstruct70_%a.err
+#SBATCH --time=01:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+#SBATCH --array=0-5%2 
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate
+
+export TRANSFORMERS_CACHE="/scratch/jbengoetxea/.cache"
+
+# Values for the 2 loops:
+DATASET_VALUES=(xnli-eu-nat-biz xnli-eu-nat-gip xnli-eu-nat-naf)  
+PROMPT_TYPE_VALUES=(nli-few nli-zero)
+   
+
+N=${#PROMPT_TYPE_VALUES[@]}  # Number of items in the second level (VALUES2)
+
+# Decode SLURM_ARRAY_TASK_ID to get the two indices
+IDX1=$((SLURM_ARRAY_TASK_ID / N))
+IDX2=$((SLURM_ARRAY_TASK_ID % N))
+
+# Use IDX1 and IDX2 for your two-level loops
+DATASET="${DATASET_VALUES[${IDX1}]}"
+PROMPT_TYPE="${PROMPT_TYPE_VALUES[${IDX2}]}"
+
+
+TASK=trilabel
+MODEL=latxainstruct70
+OUTPUT=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/results/$DATASET/$MODEL
+
+python3 /scratch/jbengoetxea/phd/XNLIvar/scripts/generative/scripts/zero_shot.py \
+    --dataset "${DATASET}" \
+    --model $MODEL \
+    --output_dir $OUTPUT \
+    --task $TASK \
+    --prompt_type "${PROMPT_TYPE}" 

generative/scripts/array-llama.sh

@@ -0,0 +1,46 @@
+#!/usr/bin/env bash
+#SBATCH --qos=regular
+#SBATCH --job-name=xnli_llamainstruct70
+#SBATCH --cpus-per-task=2
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:4
+#SBATCH --constraint=a100-sxm4
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-llamainstruct70_%a.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-llamainstruct70_%a.err
+#SBATCH --time=01:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+#SBATCH --array=0-5%2 
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate
+
+export TRANSFORMERS_CACHE="/scratch/jbengoetxea/.cache"
+
+# Values for the 2 loops:
+DATASET_VALUES=(xnli-eu-nat-biz xnli-eu-nat-gip xnli-eu-nat-naf)  
+PROMPT_TYPE_VALUES=(nli-few nli-zero)
+   
+
+N=${#PROMPT_TYPE_VALUES[@]}  # Number of items in the second level (VALUES2)
+
+# Decode SLURM_ARRAY_TASK_ID to get the two indices
+IDX1=$((SLURM_ARRAY_TASK_ID / N))
+IDX2=$((SLURM_ARRAY_TASK_ID % N))
+
+# Use IDX1 and IDX2 for your two-level loops
+DATASET="${DATASET_VALUES[${IDX1}]}"
+PROMPT_TYPE="${PROMPT_TYPE_VALUES[${IDX2}]}"
+
+
+TASK=trilabel
+MODEL=llama3instruct70
+OUTPUT=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/results/$DATASET/$MODEL
+
+python3 /scratch/jbengoetxea/phd/XNLIvar/scripts/generative/scripts/zero_shot.py \
+    --dataset "${DATASET}" \
+    --model $MODEL \
+    --output_dir $OUTPUT \
+    --task $TASK \
+    --prompt_type "${PROMPT_TYPE}" 

generative/scripts/get_accuracy.py

@@ -0,0 +1,66 @@
+import pandas as pd
+from sklearn.metrics import accuracy_score
+import argparse
+import os
+
+def process_tsv_files(parent_dir, output_csv, is_qa=None):
+    results = []
+    
+    # Walk through all subdirectories
+    for root, _, files in os.walk(parent_dir):
+        for file in files:
+            if file.endswith(".tsv"):  # Check for TSV files
+                file_path = os.path.join(root, file)
+
+                try:
+                    df = pd.read_csv(file_path, sep="\t")
+                    # Evaluate non-qa inference results
+                    if is_qa is None: 
+                        if "gold_label" in df.columns and "prediction" in df.columns:
+                            y_gold = df["gold_label"].tolist()
+                            y_pred = df["prediction"].tolist()
+                            
+                            accuracy = accuracy_score(y_gold, y_pred)
+                            results.append([file_path, accuracy])
+                        else:
+                            print(f"Skipping {file_path}: Required columns not found.")
+
+                    # Evaluate qa inference results
+                    elif is_qa == "y":
+                        for neg_value in df["prediction"].unique(): 
+                            if neg_value.startswith("not_"): 
+                                    evaluating_label = neg_value.replace("not_", "")
+
+                                    # Replace all other labels in gold_label with the negated version
+                                    df["gold_label"] = df["gold_label"].apply(
+                                        lambda x: x if x == evaluating_label else neg_value
+                                    )
+
+                        if "gold_label" in df.columns and "prediction" in df.columns:
+                            y_gold = df["gold_label"].tolist()
+                            y_pred = df["prediction"].tolist()
+                            
+                            accuracy = accuracy_score(y_gold, y_pred)
+                            results.append([file_path, accuracy])
+                        else:
+                            print(f"Skipping {file_path}: Required columns not found.")
+
+                except Exception as e:
+                    print(f"Error processing {file_path}: {e}")
+    
+    # Save results to a CSV
+    results_df = pd.DataFrame(results, columns=["File Path", "Accuracy"])
+    results_df.to_csv(output_csv, index=False)
+    print(f"Results saved to {output_csv}")
+
+def main(): 
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--parent_dir", type=str, required=True, help="Path to the parent folder")
+    parser.add_argument("--output_csv", type=str, required=True, help="Path to the output CSV file")
+    parser.add_argument("--is_qa", type=str, help="Are we evaluating qa inference?")
+    args = parser.parse_args()
+    
+    process_tsv_files(args.parent_dir, args.output_csv, args.is_qa)
+
+if __name__ == "__main__":
+    main()

generative/scripts/qa-array-gemma.sh

@@ -0,0 +1,50 @@
+#!/usr/bin/env bash
+#SBATCH --partition=hitz-exclusive
+#SBATCH --account=hitz-exclusive
+#SBATCH --job-name=xnli_gemmainstruct27
+#SBATCH --cpus-per-task=2
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:2
+#SBATCH --constraint=a100-sxm4
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-gemmainstruct27_%a.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-gemmainstruct27_%a.err
+#SBATCH --time=01:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+#SBATCH --array=0-35%2 
+
+
+export TRANSFORMERS_CACHE="/scratch/jbengoetxea/.cache"
+
+# Values
+DATASET_VALUES=(xnli-eu-var xnli-eu-native xnli-eu xnli-es-var xnli-es-native xnli-es)  
+PROMPT_TYPE_VALUES=(contradiction entailment neutral)
+TASK_VALUES=(qa-zero qa-few)
+
+# Get job array working
+D=${#DATASET_VALUES[@]}
+P=${#PROMPT_TYPE_VALUES[@]}
+T=${#TASK_VALUES[@]}
+
+TASK_ID=$SLURM_ARRAY_TASK_ID
+
+IDX_D=$((TASK_ID / (P * T)))
+IDX_P=$(((TASK_ID / T) % P))
+IDX_T=$((TASK_ID % T))
+
+DATASET="${DATASET_VALUES[$IDX_D]}"
+PROMPT_TYPE="${PROMPT_TYPE_VALUES[$IDX_P]}"
+TASK="${TASK_VALUES[$IDX_T]}"
+
+# Final values and run script
+MODEL=gemmainstruct27
+OUTPUT=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/results/$DATASET/$MODEL/$TASK
+
+TORCHDYNAMO_DISABLE=1 python3 /scratch/jbengoetxea/phd/XNLIvar/scripts/generative/scripts/zero_shot.py  \
+    --dataset "${DATASET}" \
+    --model $MODEL \
+    --output_dir $OUTPUT \
+    --task $TASK \
+    --prompt_type "${PROMPT_TYPE}" 

generative/scripts/qa-array-latxa.sh

@@ -0,0 +1,51 @@
+#!/usr/bin/env bash
+#SBATCH --partition=hitz-exclusive
+#SBATCH --account=hitz-exclusive
+#SBATCH --job-name=xnli_latxainstruct70
+#SBATCH --cpus-per-task=2
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:2
+#SBATCH --constraint=a100-sxm4
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-latxainstruct70_%a.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-latxainstruct70_%a.err
+#SBATCH --time=01:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+#SBATCH --array=0-10%2 
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate
+
+export TRANSFORMERS_CACHE="/scratch/jbengoetxea/.cache"
+
+# Values
+DATASET_VALUES=(xnli-eu-biz xnli-eu-gip xnli-eu-naf xnli-eu-nat-biz xnli-eu-nat-gip xnli-eu-nat-naf)  
+PROMPT_TYPE_VALUES=(contradiction entailment neutral)
+TASK_VALUES=(qa-zero qa-few)
+
+# Get job array working
+D=${#DATASET_VALUES[@]}
+P=${#PROMPT_TYPE_VALUES[@]}
+T=${#TASK_VALUES[@]}
+
+TASK_ID=$SLURM_ARRAY_TASK_ID
+
+IDX_D=$((TASK_ID / (P * T)))
+IDX_P=$(((TASK_ID / T) % P))
+IDX_T=$((TASK_ID % T))
+
+DATASET="${DATASET_VALUES[$IDX_D]}"
+PROMPT_TYPE="${PROMPT_TYPE_VALUES[$IDX_P]}"
+TASK="${TASK_VALUES[$IDX_T]}"
+
+# Final values and run script
+MODEL=latxainstruct70
+OUTPUT=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/results/$DATASET/$MODEL/$TASK
+
+python3 /scratch/jbengoetxea/phd/XNLIvar/scripts/generative/scripts/zero_shot.py \
+    --dataset "${DATASET}" \
+    --model $MODEL \
+    --output_dir $OUTPUT \
+    --task $TASK \
+    --prompt_type "${PROMPT_TYPE}" 

generative/scripts/qa-array-llama.sh

@@ -0,0 +1,50 @@
+#!/usr/bin/env bash
+#SBATCH --qos=regular
+#SBATCH --job-name=xnli_llamainstruct70
+#SBATCH --cpus-per-task=2
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --mem=64GB
+#SBATCH --gres=gpu:4
+#SBATCH --constraint=a100-sxm4
+#SBATCH --output=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-llamainstruct70_%a.log
+#SBATCH --error=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/logs/xnli-llamainstruct70_%a.err
+#SBATCH --time=01:00:00 #ee-hh:mm:ss
+#SBATCH --mail-type=REQUEUE
+#SBATCH --mail-user=jaione.bengoetxea@ehu.eus   
+#SBATCH --array=0-10%2 
+
+source /scratch/jbengoetxea/phd/.phd_venv_new/bin/activate
+
+export TRANSFORMERS_CACHE="/scratch/jbengoetxea/.cache"
+
+# Values
+DATASET_VALUES=(xnli-eu-biz xnli-eu-gip xnli-eu-naf xnli-eu-nat-biz xnli-eu-nat-gip xnli-eu-nat-naf)  
+PROMPT_TYPE_VALUES=(contradiction entailment neutral)
+TASK_VALUES=(qa-zero qa-few)
+
+# Get job array working
+D=${#DATASET_VALUES[@]}
+P=${#PROMPT_TYPE_VALUES[@]}
+T=${#TASK_VALUES[@]}
+
+TASK_ID=$SLURM_ARRAY_TASK_ID
+
+IDX_D=$((TASK_ID / (P * T)))
+IDX_P=$(((TASK_ID / T) % P))
+IDX_T=$((TASK_ID % T))
+
+DATASET="${DATASET_VALUES[$IDX_D]}"
+PROMPT_TYPE="${PROMPT_TYPE_VALUES[$IDX_P]}"
+TASK="${TASK_VALUES[$IDX_T]}"
+
+# Final values and run script
+MODEL=llama3instruct70
+OUTPUT=/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/results/$DATASET/$MODEL/$TASK
+
+python3 /scratch/jbengoetxea/phd/XNLIvar/scripts/generative/scripts/zero_shot.py \
+    --dataset "${DATASET}" \
+    --model $MODEL \
+    --output_dir $OUTPUT \
+    --task $TASK \
+    --prompt_type "${PROMPT_TYPE}" 

generative/scripts/zero_shot.py

@@ -0,0 +1,310 @@
+from transformers import AutoTokenizer, AutoModelForCausalLM, set_seed
+import torch
+from huggingface_hub import login
+import re
+import sys
+from sklearn.metrics import accuracy_score
+import argparse
+import json
+import pathlib
+from typing import List, Dict
+import numpy as np
+import pandas as pd
+import logging
+from datetime import datetime
+import os
+
+MAX_NEW_TOKENS = 5
+TEMPERATURE = 0.3
+
+
+with open("/scratch/jbengoetxea/phd/XNLIvar/scripts/generative/config.json", "r") as f:
+    config = json.load(f)
+
+def parse_args():
+    #os.environ['TRANSFORMERS_CACHE'] = '/XXXX-7/users/XXXX-1/metaphor_LLMs/paraphrase_gen/.cache/huggingface/hub'
+
+    parser = argparse.ArgumentParser(
+            description="Finetune a transformers model on a text classification task"
+        )
+
+    parser.add_argument(
+    "--dataset",
+    type=str,
+    default=None,
+    required=True,
+    help="Name of the dataset to predict gold_labels",
+    choices=["xnli-eu-native", "xnli-eu-var", "xnli-es-native", "xnli-es-var", "xnli-en", "xnli-es", "xnli-eu", "xnli-es-var-no-rep", "xnli-eu-var-no-rep", "xnli-eu-var-less-biz", "xnli-eu-var-less-gip", "xnli-eu-biz", "xnli-eu-gip", "xnli-eu-naf", "xnli-eu-nat-biz", "xnli-eu-nat-gip", "xnli-eu-nat-naf"]
+    )
+    
+    parser.add_argument(
+    "--model",
+    type=str,
+    default=None,
+    required=True,
+    help="Model name in config", 
+    choices=["llama3instruct8", "llama3instruct70", "gemmainstruct9", "gemmainstruct27", "latxainstruct70", "llama3base70"]
+    )
+    
+    parser.add_argument(
+    "--output_dir",
+    type=str,
+    default=None,
+    required=True,
+    help="Output path to dump predictions"
+    )
+    
+    parser.add_argument(
+    "--task",
+    type=str,
+    default=None,
+    required=True,
+    help="Type of task formulation",
+    choices=["binary", "trilabel", "qa-zero", "qa-few"]
+    )
+    
+    parser.add_argument(
+    "--prompt_type",
+    type=str,
+    default=None,
+    required=True,
+    help="Type of prompt"
+    )
+    
+    parser.add_argument(
+    "--paraphrases",
+    action="store_true",
+    required=False,
+    help="Dataset with paraphrases generated automatically"
+    )
+    
+    parser.add_argument(
+    "--paraphrase_source",
+    type=str,
+    default=None,
+    required=False,
+    help="Model used to generate paraphrases"
+    )
+    
+    args = parser.parse_args()
+    
+    return args
+
+def load_dataset(data_path: str) -> pd.DataFrame:
+    df = None
+    extension = pathlib.Path(data_path).suffix
+    if extension.endswith("json"):
+        df = pd.read_json(data_path)
+    elif extension.endswith("jsonl"):
+        df = pd.read_json(data_path, lines=True)
+    elif extension.endswith("tsv"):
+        df = pd.read_csv(data_path, sep="\t")
+    else:
+        df = pd.read_csv(data_path)
+        
+    return df
+
+def dump_predictions(out_path: str, premises: List, hypotheses: List, gold_labels: List, predictions: List, paraphrased_sents=None):
+    if paraphrased_sents:
+        with open(out_path, "w") as o:
+            o.write("premise\thypothesis\tgold_label\tprediction\tparaphrased_sentence\n") 
+            for p, h, g, pr, paraph in zip(premises, hypotheses, gold_labels, predictions, paraphrased_sents):
+                o.write(f"{p}\t{h}\t{g}\t{pr}\t{paraph}\n")
+    else:
+        with open(out_path, "w") as o:
+            o.write("premise\thypothesis\tgold_label\tprediction\n") 
+            for p, h, g, pr in zip(premises, hypotheses, gold_labels, predictions):
+                o.write(f"{p}\t{h}\t{g}\t{pr}\n")
+            
+    print(f"{len(predictions)} Predictions stored in {out_path}")
+    
+def map_labels(predictions: List[str], label_mapping: Dict):
+    predictions_clean = [pred.strip("<>.,") for pred in predictions.lower().split()]
+    for pred in predictions_clean:
+        for label in label_mapping:
+            label_lower = label.lower()
+            # Allow partial matching in both directions
+            if pred in label_lower or label_lower in pred:
+                return label_mapping[label]
+    return "unk"
+        
+def get_column_values(df, col_id):
+        return df[col_id].tolist()
+
+
+def map_labels_to_string(labels: List):
+    label_strings = []
+    for label in labels:
+        if label == 0: 
+            label_strings.append("entailment")
+        elif label == 1: 
+            label_strings.append("neutral")
+        else: 
+            label_strings.append("contradiction")
+        
+    return label_strings
+
+def main():
+    
+    args = parse_args()
+
+    if not os.path.exists(args.output_dir):
+        os.makedirs(args.output_dir, exist_ok=True)
+        
+    logger_path = os.path.join(args.output_dir, f"{args.prompt_type}_{args.paraphrase_source+'_' if args.paraphrase_source else ''}{datetime.now().strftime('%d-%m-%Y_%H_%M_%S')}.log")
+    
+    logger = logging.getLogger(__name__)
+    logging.basicConfig(filename=os.path.join(logger_path), encoding='utf-8', level=logging.INFO)
+
+    # Disable compilation (to avoid recompile_limit errors)
+    #torch._dynamo.disable()
+    #torch._dynamo.config.suppress_errors = True
+    #torch._dynamo.config.recompile_limit = 100 
+
+    login(token='LOGIN_TOKEN') # Add hf login token
+    model_id = config.get("models", {}).get(args.model, "")
+    logger.info(f"Model used: {model_id}")
+    logger.info(f"Prompt task: {args.task}")
+    logger.info(f"Dataset with paraphrases: {args.paraphrases}")
+    logger.info(f"Prompt config: {args.prompt_type}")
+    device=torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+    logger.info(f"Device in use: {device}")
+    
+
+    # TORCH_LOGS=recompiles
+
+    datasets_config = config.get("datasets", {})
+    prompt_config = config.get("prompts", {}).get(args.task, {})
+    
+    print(args.task)
+    print(datasets_config.get(args.dataset, {}).get("prompts", []))
+    
+    # Ensure trilabel setup only for Meta4XNLI 
+    assert args.task in datasets_config.get(args.dataset, {}).get("prompts", [])
+    
+    if args.paraphrases:
+        data_path =  datasets_config.get(args.dataset, {}).get("data_path_paraphrase", "")
+    else:
+        data_path =  datasets_config.get(args.dataset, {}).get("data_path", "")
+
+    logger.info(f"Dataset loaded from: {data_path}")
+    df = load_dataset(data_path)
+    logger.info(f"Loaded samples: {len(df)}")
+    premises = get_column_values(df, datasets_config.get(args.dataset, "").get("prem_col", ""))
+    hypotheses = get_column_values(df, datasets_config.get(args.dataset, "").get("hyp_col", ""))
+    if args.paraphrases:
+        gold_labels = get_column_values(df, "gold_label")
+    else:
+        gold_labels = [l for l in get_column_values(df, datasets_config.get(args.dataset, "").get("label_col", ""))]
+    
+    print(gold_labels)
+
+    gold_labels = map_labels_to_string(gold_labels)
+    print(gold_labels)
+
+    labels = list(set(gold_labels))
+
+    set_seed(5)
+
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    
+
+    print("MODEL ID:", model_id)
+
+
+    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.bfloat16)
+
+    print("here i am")
+    
+    tokenizer.pad_token_id = tokenizer.eos_token_id
+
+    predictions = []
+    for p, h, l in zip(premises, hypotheses, gold_labels):
+        preffix_prompt = prompt_config.get(args.prompt_type, {}).get("preffix", "")
+        print(preffix_prompt)
+        print(args.prompt_type)
+        if args.prompt_type == "chain":
+            prompt = preffix_prompt + f"\n Premise: {p}\n Hypothesis: {h}\n Answer: "
+            logger.info(f"Prompt: {prompt}")
+        else:
+            prompt = preffix_prompt + f" {p} -> {h}: "
+            logger.info(f"Prompt: {prompt}")
+        
+        
+        label_mappings = prompt_config.get(args.prompt_type, {}).get("label_mapping")
+        
+        logger.info(f"Label mappings: {label_mappings}")
+        
+        inputs = tokenizer([prompt], return_tensors="pt").to(device)
+        
+        logger.info(f"{p}\t{h}\t{l}")
+                
+    
+        
+
+
+        # #################
+        # input_text = "Write me a poem about Machine Learning."
+        # input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")
+
+        # outputs = model.generate(**input_ids, max_new_tokens=32)
+        # print(tokenizer.decode(outputs[0]))
+        # #################
+
+
+
+        
+        outputs = model.generate(**inputs, max_new_tokens=MAX_NEW_TOKENS, return_dict_in_generate=True, output_scores=True, temperature=TEMPERATURE)
+    
+        transition_scores = model.compute_transition_scores(
+            outputs.sequences, outputs.scores, normalize_logits=True
+        )
+        logger.info(f"{outputs.sequences}\t{outputs.scores}")
+        
+        
+        #print(f"transition scores: {transition_scores}", flush=True)
+
+        #print(f"transition scores: {transition_scores}", flush=True)
+        # input_length is the length of the input prompt for decoder-only models, like the GPT family, and 1 for
+        # encoder-decoder models, like BART or T5.
+        input_length = 1 if model.config.is_encoder_decoder else inputs.input_ids.shape[1]
+        generated_tokens = outputs.sequences[:, input_length:]
+        for tok, score in zip(generated_tokens[0], transition_scores[0]):
+            # | token | token string | log probability | probability
+            logger.info(f"| {tok:5d} | {tokenizer.decode(tok):8s} | {score}")
+            #logger.info(f"| {tok:5d} | {tokenizer.decode(tok):8s} | {score.numpy():.3f} | {np.exp(score.numpy()):.2%}")
+            #o.write(f"| {tok:5d} | {tokenizer.decode(tok):8s} | {score.numpy():.3f} | {np.exp(score.numpy()):.2%}")
+
+
+        answers = tokenizer.decode(generated_tokens[0], skip_special_tokens=True)
+        logger.info(f"Answers: {answers}, split: {answers.split()}")
+        logger.info(f"Mapped label: {map_labels(answers, label_mappings)}")
+        predictions.append(map_labels(answers, label_mappings))
+        logger.info("Label added to predictions.")
+        
+            
+    logger.debug(gold_labels[:5], predictions[:5], flush=True)
+    assert len(gold_labels) == len(predictions)
+    logger.info(f"Gold: {len(gold_labels)}, Pred: {len(predictions)}")
+    
+   
+    predictions_path = os.path.join(args.output_dir, f"{args.prompt_type}_{args.paraphrase_source+'_' if args.paraphrase_source else ''}{datetime.now().strftime('%d-%m-%Y_%H_%M_%S')}.tsv")
+    
+    if args.paraphrases:
+        paraphrased_sents = df.iloc[:, -1].tolist()
+        logger.info(f"Dumping predictions with paraphrased sentences, met location: {list(df.columns)[-1]}")
+        dump_predictions(predictions_path, premises, hypotheses, gold_labels, predictions, paraphrased_sents)    
+    else:
+        dump_predictions(predictions_path, premises, hypotheses, gold_labels, predictions)
+    
+    logger.info(f"Predictions dumped to {predictions_path}")
+    
+    
+    
+    accuracy = accuracy_score(gold_labels, predictions, normalize=True)
+    logger.info(f"Accuracy {len(gold_labels)}, {len(predictions)}: {accuracy}\n")
+    
+
+
+if __name__ == "__main__":
+    main()