forked from EleutherAI/lm-evaluation-harness
-
Notifications
You must be signed in to change notification settings - Fork 47
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
re-evaluate models with jsquad prompt with title
- Loading branch information
Showing
69 changed files
with
470 additions
and
114 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,3 @@ | ||
MODEL_ARGS="pretrained=abeja/gpt-neox-japanese-2.7b,device_map=auto,torch_dtype=auto" | ||
TASK="jsquad-1.2-0.2" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "3" --device "cuda" --output_path "models/abeja-gpt-neox-japanese-2.7b/result.jsquad-1.2.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,3 +1,3 @@ | ||
MODEL_ARGS="pretrained=abeja/gpt-neox-japanese-2.7b" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.1-0.2,xlsum_ja" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.2-0.2,xlsum_ja" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2,3,3,3,1" --device "cuda" --output_path "models/abeja-gpt-neox-japanese-2.7b/result.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
22 changes: 22 additions & 0 deletions
22
models/abeja-gpt-neox-japanese-2.7b/result.jsquad-1.2.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,22 @@ | ||
{ | ||
"results": { | ||
"jsquad-1.2-0.2": { | ||
"exact_match": 15.803692030616839, | ||
"f1": 25.18326978234071 | ||
} | ||
}, | ||
"versions": { | ||
"jsquad-1.2-0.2": 1.2 | ||
}, | ||
"config": { | ||
"model": "hf-causal", | ||
"model_args": "pretrained=abeja/gpt-neox-japanese-2.7b,device_map=auto,torch_dtype=auto", | ||
"num_fewshot": 3, | ||
"batch_size": null, | ||
"device": "cuda", | ||
"no_cache": false, | ||
"limit": null, | ||
"bootstrap_iters": 100000, | ||
"description_dict": {} | ||
} | ||
} |
3 changes: 3 additions & 0 deletions
3
models/cyberagent/cyberagent-open-calm-1b/harness.jsquad-1.2.sh
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-1b,device_map=auto,torch_dtype=auto" | ||
TASK="jsquad-1.2-0.2" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "3" --device "cuda" --output_path "models/cyberagent-open-calm-1b/result.jsquad-1.2.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,3 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-1b" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.1-0.2,xlsum_ja" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.2-0.2,xlsum_ja" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2,3,3,3,1" --device "cuda" --output_path "models/cyberagent-open-calm-1b/result.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
22 changes: 22 additions & 0 deletions
22
models/cyberagent/cyberagent-open-calm-1b/result.jsquad-1.2.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,22 @@ | ||
{ | ||
"results": { | ||
"jsquad-1.2-0.2": { | ||
"exact_match": 39.53174245835209, | ||
"f1": 49.49399460234075 | ||
} | ||
}, | ||
"versions": { | ||
"jsquad-1.2-0.2": 1.2 | ||
}, | ||
"config": { | ||
"model": "hf-causal", | ||
"model_args": "pretrained=cyberagent/open-calm-1b", | ||
"num_fewshot": 3, | ||
"batch_size": null, | ||
"device": "cuda", | ||
"no_cache": false, | ||
"limit": null, | ||
"bootstrap_iters": 100000, | ||
"description_dict": {} | ||
} | ||
} |
3 changes: 3 additions & 0 deletions
3
models/cyberagent/cyberagent-open-calm-3b/harness.jsquad-1.2.sh
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-3b,device_map=auto,torch_dtype=auto" | ||
TASK="jsquad-1.2-0.2" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2" --device "cuda" --output_path "models/cyberagent/cyberagent-open-calm-3b/result.jsquad-1.2.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,3 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-3b" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.1-0.2,jaqket_v2-0.1-0.2,xlsum_ja,xwinograd_ja,mgsm" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.2-0.2,jaqket_v2-0.1-0.2,xlsum_ja,xwinograd_ja,mgsm" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "3,3,3,2,1,1,0,5" --device "cuda" --output_path "models/cyberagent/cyberagent-open-calm-3b/result.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
22 changes: 22 additions & 0 deletions
22
models/cyberagent/cyberagent-open-calm-3b/result.jsquad-1.2.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,22 @@ | ||
{ | ||
"results": { | ||
"jsquad-1.2-0.2": { | ||
"exact_match": 44.529491220171096, | ||
"f1": 56.02141036867636 | ||
} | ||
}, | ||
"versions": { | ||
"jsquad-1.2-0.2": 1.2 | ||
}, | ||
"config": { | ||
"model": "hf-causal", | ||
"model_args": "pretrained=cyberagent/open-calm-3b,device_map=auto,torch_dtype=auto", | ||
"num_fewshot": 2, | ||
"batch_size": null, | ||
"device": "cuda", | ||
"no_cache": false, | ||
"limit": null, | ||
"bootstrap_iters": 100000, | ||
"description_dict": {} | ||
} | ||
} |
3 changes: 3 additions & 0 deletions
3
models/cyberagent/cyberagent-open-calm-7b/harness.jsquad-1.2.sh
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-7b,device_map=auto,torch_dtype=auto" | ||
TASK="jsquad-1.2-0.2" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2" --device "cuda" --output_path "models/cyberagent/cyberagent-open-calm-7b/result.jsquad-1.2.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,3 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-7b" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.1-0.2,jaqket_v2-0.1-0.2,xlsum_ja,xwinograd_ja,mgsm" | ||
TASK="jcommonsenseqa-1.1-0.2,jnli-1.1-0.2,marc_ja-1.1-0.2,jsquad-1.2-0.2,jaqket_v2-0.1-0.2,xlsum_ja,xwinograd_ja,mgsm" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "3,3,3,2,1,1,0,5" --device "cuda" --output_path "models/cyberagent/cyberagent-open-calm-7b/result.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
22 changes: 22 additions & 0 deletions
22
models/cyberagent/cyberagent-open-calm-7b/result.jsquad-1.2.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,22 @@ | ||
{ | ||
"results": { | ||
"jsquad-1.2-0.2": { | ||
"exact_match": 48.10895992796038, | ||
"f1": 60.90961937230767 | ||
} | ||
}, | ||
"versions": { | ||
"jsquad-1.2-0.2": 1.2 | ||
}, | ||
"config": { | ||
"model": "hf-causal", | ||
"model_args": "pretrained=cyberagent/open-calm-7b,device_map=auto,torch_dtype=auto", | ||
"num_fewshot": 2, | ||
"batch_size": null, | ||
"device": "cuda", | ||
"no_cache": false, | ||
"limit": null, | ||
"bootstrap_iters": 100000, | ||
"description_dict": {} | ||
} | ||
} |
3 changes: 3 additions & 0 deletions
3
models/cyberagent/cyberagent-open-calm-large/harness.jsquad-1.2.sh
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-large,use_fast=True,device_map=auto,torch_dtype=auto" | ||
TASK="jsquad-1.2-0.2" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "3" --device "cuda" --output_path "models/cyberagent-open-calm-large/result.jsquad-1.2.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
22 changes: 22 additions & 0 deletions
22
models/cyberagent/cyberagent-open-calm-large/result.jsquad-1.2.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,22 @@ | ||
{ | ||
"results": { | ||
"jsquad-1.2-0.2": { | ||
"exact_match": 40.4997748761819, | ||
"f1": 51.32160467436942 | ||
} | ||
}, | ||
"versions": { | ||
"jsquad-1.2-0.2": 1.2 | ||
}, | ||
"config": { | ||
"model": "hf-causal", | ||
"model_args": "pretrained=cyberagent/open-calm-large,use_fast=True,device_map=auto,torch_dtype=auto", | ||
"num_fewshot": 3, | ||
"batch_size": null, | ||
"device": "cuda", | ||
"no_cache": false, | ||
"limit": null, | ||
"bootstrap_iters": 100000, | ||
"description_dict": {} | ||
} | ||
} |
3 changes: 3 additions & 0 deletions
3
models/cyberagent/cyberagent-open-calm-medium/harness.jsquad-1.2.sh
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,3 @@ | ||
MODEL_ARGS="pretrained=cyberagent/open-calm-medium,use_fast=True,device_map=auto,torch_dtype=auto" | ||
TASK="jsquad-1.2-0.2" | ||
python main.py --model hf-causal --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "3" --device "cuda" --output_path "models/cyberagent-open-calm-medium/result.jsquad-1.2.json" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
22 changes: 22 additions & 0 deletions
22
models/cyberagent/cyberagent-open-calm-medium/result.jsquad-1.2.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,22 @@ | ||
{ | ||
"results": { | ||
"jsquad-1.2-0.2": { | ||
"exact_match": 29.85141828005403, | ||
"f1": 40.49655778214922 | ||
} | ||
}, | ||
"versions": { | ||
"jsquad-1.2-0.2": 1.2 | ||
}, | ||
"config": { | ||
"model": "hf-causal", | ||
"model_args": "pretrained=cyberagent/open-calm-medium,use_fast=True,device_map=auto,torch_dtype=auto", | ||
"num_fewshot": 3, | ||
"batch_size": null, | ||
"device": "cuda", | ||
"no_cache": false, | ||
"limit": null, | ||
"bootstrap_iters": 100000, | ||
"description_dict": {} | ||
} | ||
} |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,4 @@ | ||
MODEL_ARGS="pretrained=huggyllama/llama-7b,use_accelerate=True,load_in_8bit=True" | ||
TASK="jsquad-1.2-0.3" | ||
python main.py --model hf-causal-experimental --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2" --device "cuda" --output_path "models/llama/llama-7b/result.jsquad-1.2.json" --batch_size 2 | ||
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,3 +1,3 @@ | ||
MODEL_ARGS="pretrained=huggyllama/llama-7b,use_accelerate=True,load_in_8bit=True" | ||
TASK="jsquad-1.1-0.3,jcommonsenseqa-1.1-0.3,jnli-1.1-0.3,marc_ja-1.1-0.3" | ||
TASK="jsquad-1.2-0.3,jcommonsenseqa-1.1-0.3,jnli-1.1-0.3,marc_ja-1.1-0.3" | ||
python main.py --model hf-causal-experimental --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2,3,3,3" --device "cuda" --output_path "models/llama/llama-7b/result.json" --batch_size 2 > models/llama/llama-7b/harness.out 2> models/llama/llama-7b/harness.err |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,22 @@ | ||
{ | ||
"results": { | ||
"jsquad-1.2-0.3": { | ||
"exact_match": 36.24493471409275, | ||
"f1": 50.91625240527312 | ||
} | ||
}, | ||
"versions": { | ||
"jsquad-1.2-0.3": 1.2 | ||
}, | ||
"config": { | ||
"model": "hf-causal-experimental", | ||
"model_args": "pretrained=huggyllama/llama-7b,use_accelerate=True,load_in_8bit=True", | ||
"num_fewshot": 2, | ||
"batch_size": 2, | ||
"device": "cuda", | ||
"no_cache": false, | ||
"limit": null, | ||
"bootstrap_iters": 100000, | ||
"description_dict": {} | ||
} | ||
} |
4 changes: 2 additions & 2 deletions
4
...a2/llama2-2.7b/harness.jsquad-1.1-0.31.sh → .../llama2/llama2-2.7b/harness.jsquad-1.2.sh
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,4 +1,4 @@ | ||
MODEL_ARGS="pretrained=meta-llama/Llama-2-7b-hf,use_accelerate=True,dtype=auto" | ||
TASK="jsquad-1.1-0.31" | ||
python main.py --model hf-causal-experimental --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2" --device "cuda" --output_path "models/llama2/llama2-2.7b/result.jsquad-1.1-0.31.json" --batch_size 2 | ||
TASK="jsquad-1.2-0.3" | ||
python main.py --model hf-causal-experimental --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2" --device "cuda" --output_path "models/llama2/llama2-2.7b/result.jsquad-1.2.json" --batch_size 2 | ||
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,5 +1,5 @@ | ||
MODEL_ARGS="pretrained=meta-llama/Llama-2-7b-hf,use_accelerate=True" | ||
TASK="jsquad-1.1-0.3,jcommonsenseqa-1.1-0.3,jnli-1.1-0.3,marc_ja-1.1-0.3" | ||
TASK="jsquad-1.2-0.3,jcommonsenseqa-1.1-0.3,jnli-1.1-0.3,marc_ja-1.1-0.3" | ||
python main.py --model hf-causal-experimental --model_args $MODEL_ARGS --tasks $TASK --num_fewshot "2,3,3,3" --device "cuda" --output_path "models/llama2/llama2-2.7b/result.json" --batch_size 2 > models/llama2/llama2-2.7b/harness.out 2> models/llama2/llama2-2.7b/harness.err | ||
|
||
|
Oops, something went wrong.