wehub/microsoft--unilm
1
0
Fork 0
Code Issues Pull Requests Actions Deployments Models Agent Notes Packages Projects Releases Wiki Activity

chore: import upstream snapshot with attribution

Browse Source
This commit is contained in:
wehub-resource-sync
2026-07-13 13:24:13 +08:00
commit 1037506f2e
6050 changed files with 1731598 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
PFPO/conf/api/vllm/magicoder/llama3/test_case_input_gen_v1_0.yaml
+112
View File
@@ -0,0 +1,112 @@
defaults:
- hydra: default
- _self_
hydra:
searchpath:
- file://conf/
data_path_prefix: /mnt/fangkai_blob/share/
model_path_prefix: /mnt/fangkai_blob/share/models
output_path_prefix: /mnt/fangkai_blob/reward_modeling/
train_file: ${data_path_prefix}/dataset/magicoder/data-oss_instruct-decontaminated-python.json
dev_file: ${train_file}
test_file: ${train_file}
port: 6000
model:
sampling_params:
_target_: vllm.SamplingParams
n: 1
temperature: 0.0
top_p: 1.0
stop: [ "</s>", "\n\n\n\n", "Context:\n", "Thought 42:", "<|end_of_text|>", "<|eot_id|>, <|EOT|>" ]
max_tokens: 1024
tem: ${sampling_params.temperature}
n: ${sampling_params.n}
split_size: 2
split_id: 0
max_num_seqs: 64
suffix: ${split_id}-of-${split_size}
output_file: ${output_dir}/apps-test-inputs-gen/${eval_sub_path}/sub_dev.0shot.tem${tem}.n${n}.${suffix}.v1.0.json
flush_file: ${output_file}l
apply_chat_template: True
add_generation_prompt: True
chat_prefix:
chat_connect:
chat_suffix:
prompt:
_target_: data.input_utils.read_text
file_path: prompts/apps/test_input_gen_2shot_v2.1.txt
# Data loading
read_tensor:
_target_: data.combine_dataset.ResponseAlignDataset
template:
_target_: data.input_utils.compose_template
units:
prompt: ${prompt}
composition: "{prompt}"
instruction:
replacement:
"[[Question]]": "problem"
index_field: index
service_based: False
split_size: ${split_size}
split_id: ${split_id}
service_processor:
_target_: data.vllm.VLLMRequestGenerator
api_url: http://0.0.0.0:${port}/v1/completions
max_tokens: 4096
model: ${model}
stop: ${sampling_params.stop}
temperature: ${sampling_params.temperature}
n: ${sampling_params.n}
max_data_num: -1
flush_file: ${flush_file}
exp_name:
save_best: False
eval_sub_path: ""
output_dir: ${model_path_prefix}//Meta-Llama-3.1-70B-Instruct/
# Dataloader
num_workers: 32
prefetch_factor: 2
dp_size:
tp_size: 1
pp_size: 1
post_process:
_target_: post_processors.openai_api_callback.SaveOnlyCallBack
output_file: ${output_file}
answer_clean:
index_field: index
resume: True
# Training hyper-parameters
per_gpu_train_batch_size: 1
per_gpu_eval_batch_size: 1
ddp_eval: False
no_cuda: False
seed: 42
local_rank: -1
# Temporary variables
fp16: True
fp16_bfloat16: True
n_gpu: 1
device:
train_batch_size:
eval_batch_size:
world_size:
PFPO/conf/api/vllm/magicoder/mistral/func_head_extract_v1_0.yaml
+112
View File
@@ -0,0 +1,112 @@
defaults:
- hydra: default
- _self_
hydra:
searchpath:
- file://conf/
data_path_prefix: /mnt/fangkai_blob/share/
model_path_prefix: /mnt/fangkai_blob/share/models
output_path_prefix: /mnt/fangkai_blob/reward_modeling/
train_file: ${data_path_prefix}/dataset/magicoder/data-oss_instruct-decontaminated-python.json
dev_file: ${train_file}
test_file: ${train_file}
port: 6000
model:
sampling_params:
_target_: vllm.SamplingParams
n: 1
temperature: 0.0
top_p: 1.0
stop: [ "</s>", "\n\n\n\n", "Context:\n", "Thought 42:", "<|end_of_text|>", "<|eot_id|>, <|EOT|>" ]
max_tokens: 512
tem: ${sampling_params.temperature}
n: ${sampling_params.n}
split_size: 2
split_id: 0
max_num_seqs: 64
max_model_len: 4096
suffix: ${split_id}-of-${split_size}
output_file: ${output_dir}/apps-test-inputs-gen/${eval_sub_path}/oss_instruct_python.func_head_extract.tem${tem}.n${n}.${suffix}.v1.0.json
flush_file: ${output_file}
apply_chat_template: True
add_generation_prompt: True
chat_prefix:
chat_connect:
chat_suffix:
prompt:
_target_: data.input_utils.read_text
file_path: prompts/magicoder/oss_has_function_head_v1_0.txt
# Data loading
read_tensor:
_target_: data.combine_dataset.ResponseAlignDataset
template:
_target_: data.input_utils.compose_template
units:
prompt: ${prompt}
composition: "{prompt}"
instruction:
replacement:
"[[Question]]": "problem"
index_field: index
service_based: False
split_size: ${split_size}
split_id: ${split_id}
service_processor:
_target_: data.vllm.VLLMRequestGenerator
api_url: http://0.0.0.0:${port}/v1/completions
max_tokens: 4096
model: ${model}
stop: ${sampling_params.stop}
temperature: ${sampling_params.temperature}
n: ${sampling_params.n}
max_data_num: -1
flush_file: ${flush_file}
exp_name:
save_best: False
eval_sub_path: ""
output_dir: ${model_path_prefix}/Mistral-Large-Instruct-2407/
# Dataloader
num_workers: 32
prefetch_factor: 2
dp_size:
tp_size: 1
pp_size: 1
post_process:
_target_: post_processors.openai_api_callback.SaveOnlyCallBack
output_file: ${output_file}
answer_clean:
index_field: index
resume: True
# Training hyper-parameters
per_gpu_train_batch_size: 1
per_gpu_eval_batch_size: 1
ddp_eval: False
no_cuda: False
seed: 42
local_rank: -1
# Temporary variables
fp16: True
fp16_bfloat16: True
n_gpu: 1
device:
train_batch_size:
eval_batch_size:
world_size:
PFPO/conf/api/vllm/magicoder/mistral/test_case_input_gen_v1_0.yaml
+112
View File
@@ -0,0 +1,112 @@
defaults:
- hydra: default
- _self_
hydra:
searchpath:
- file://conf/
data_path_prefix: /mnt/fangkai_blob/share/
model_path_prefix: /mnt/fangkai_blob/share/models
output_path_prefix: /mnt/fangkai_blob/reward_modeling/
train_file: ${data_path_prefix}/dataset/magicoder/data-oss_instruct-decontaminated-python.json
dev_file: ${train_file}
test_file: ${train_file}
port: 6000
model:
sampling_params:
_target_: vllm.SamplingParams
n: 1
temperature: 0.0
top_p: 1.0
stop: [ "</s>", "\n\n\n\n", "Context:\n", "Thought 42:", "<|end_of_text|>", "<|eot_id|>, <|EOT|>" ]
max_tokens: 4096
tem: ${sampling_params.temperature}
n: ${sampling_params.n}
split_size: 2
split_id: 0
max_num_seqs: 64
max_model_len: 4096
suffix: ${split_id}-of-${split_size}
output_file: ${output_dir}/apps-test-inputs-gen/${eval_sub_path}/sub_dev.0shot.tem${tem}.n${n}.${suffix}.v1.0.json
flush_file: ${output_file}
apply_chat_template: True
add_generation_prompt: True
chat_prefix:
chat_connect:
chat_suffix:
prompt:
_target_: data.input_utils.read_text
file_path: prompts/apps/test_input_gen_2shot_v2.1.txt
# Data loading
read_tensor:
_target_: data.combine_dataset.ResponseAlignDataset
template:
_target_: data.input_utils.compose_template
units:
prompt: ${prompt}
composition: "{prompt}"
instruction:
replacement:
"[[Question]]": "problem"
index_field: index
service_based: False
split_size: ${split_size}
split_id: ${split_id}
service_processor:
_target_: data.vllm.VLLMRequestGenerator
api_url: http://0.0.0.0:${port}/v1/completions
max_tokens: 4096
model: ${model}
stop: ${sampling_params.stop}
temperature: ${sampling_params.temperature}
n: ${sampling_params.n}
max_data_num: -1
flush_file: ${flush_file}
exp_name:
save_best: False
eval_sub_path: ""
output_dir: ${model_path_prefix}/Mistral-Large-Instruct-2407/
# Dataloader
num_workers: 32
prefetch_factor: 2
dp_size:
tp_size: 1
pp_size: 1
post_process:
_target_: post_processors.openai_api_callback.SaveOnlyCallBack
output_file: ${output_file}
answer_clean:
index_field: index
resume: True
# Training hyper-parameters
per_gpu_train_batch_size: 1
per_gpu_eval_batch_size: 1
ddp_eval: False
no_cuda: False
seed: 42
local_rank: -1
# Temporary variables
fp16: True
fp16_bfloat16: True
n_gpu: 1
device:
train_batch_size:
eval_batch_size:
world_size: