Model download: https://huggingface.co/unit-mesh/autodev-deepseek-6.7b-finetunes

# Merge and shuffle records from different files
merge_jsonl(
    output_file=merged_file,
    input_files=[oss_instruction, python_instruction, 'code_bugfix_cleaned_5K.json', 'codeGPT_CN_cleaned_20K.json',
                 'code_summarization_CN_cleaned_10K.json', 'code_generation_CN_cleaned_5K.json', 'summary.jsonl'],
    lines_per_file=[4000, 4000, 4000, 15000, 8000, 4000, 25000]
)

Full Changelog: unit-mesh/unit-eval@v0.3.1...v0.3.2

Datasets -> 44k

# Merge and shuffle records from different files
merge_jsonl(
    output_file=merged_file,
    input_files=[oss_instruction, 'code_bugfix_cleaned_5K.json', 'codeGPT_CN_cleaned_20K.json',
                 'code_summarization_CN_cleaned_10K.json', 'java-code-completion.jsonl', 'java-test-gen.jsonl',
                 'kotlin-completion-11929.jsonl', 'kotlin-java-comments-3715.jsonl'],
    lines_per_file=[4000, 4000, 15000, 8000, 5000, 3000, 3000, 2000 ]
)

0.3.2 (2023-12-30)

Bug Fixes

• fix error worker manager (93aeb6b)
• fix not working togther issue #1 (cc9e5b3)
• fix tests (0c2d62e)
• kotlin: add seperate type for class and method test #1 (edb76d9)
• kotlin: fix kotlin test method issue #1 (0ecb7b8)
• kotlin: fix naming issue #1 (a0185f8)
• kotlin: try to resolve lost test gen cases #1 (67275f6)

Features

• add support for java comments (8f0b49d)
• comment: add handle for kotlin comment #1 (ebd5610)
• comment: align comment to Ins #1 (d9d33d4)
• comment: make kotlin comment support works #1 (1e411a7)
• comment: make parse kotlin comments work #1 (61f99ab)
• comments: init comments for #1 (1d60f8d)
• doc: init builder for build in Kotlin (e03dd14)
• kotlin: fix for display issues #1 (d9a12b0)
• kotlin: init test services (5d26c16)
• kotlin: try to handle for call tree (03644a5)
• kotlin: try to parse for method test in Kotlin #1 (464773d)

Latest OpenBayes Link: https://openbayes.com/console/phodal/models/rCmer1KQSgp/5/overview

0.3.0 (2023-12-24)

Bug Fixes

• add lost super classes (ebcd6d4)
• fix after block issue & write test for after block (f7afbc0)
• fix blank issue (dbcca15)
• fix empty issue (c8706eb)
• fix empty tests issue (9b916e0)
• fix for JavaWorker in error issue (897b288)
• fix lfs issue (aa49873)
• fix option issue (a56c9c2)
• fix origin code error issues (f342793)
• fix related code format to instructions issues (a66d8b1)
• fix rename issue (3cf8175)
• fix simliar chunk issue (7aefcae)
• fix simliar chunk length issue (0d0d121)
• fix tests (5291a00)
• fix tests (46e9cad)
• fix tests (2f8a472)
• fix tests (6ce7891)
• fix typos (8e0e911)
• fix typos (f3c4024)
• try to parse deps (123f904)

Features

• add add check naming style function name (c52bc1c)
• add api server examples (4ab3ce0)
• add first version deepseek code finetune script (188bf6d)
• add inline block support (4acd852)
• add limit for each completion type (87df4b0)
• add line as quality threshold (88898ab)
• add shuffle in final output for better display (6fa98c2)
• add simple instruction for render in output (e36ac67)
• add simple line count strategy (744847e)
• add simple server examples (b7dde3e)
• add simple test for java coder (b2efc31)
• add simplie completion block filter (e370bf2)
• disti: init module (f41c95a)
• fintune: update for batch size for 4090 (19a9eb7)
• init support lang design (280ee34)
• make change history (dbdbc49)
• make file in cmpletion realy works (619b83b)
• make instruction in same output for deepseek (c41c418)
• make option working in output (2c4f502)
• parepare for library as context (30ffcec)
• server: add for timeout (1f9e53c)
• server: try to saimple response (46f9863)
• test: add basic rules for generate data (2e9549a)
• test: init first version for class test generate (33ef46f)
• unitest: add ins quality threshold for test size (8591675)
• unitest: add test support (1ebea0e)
• unitest: make it works for instructions (1d5fd62)
• unittest: add identify to frameworks (8c8c327)
• unittest: add simple way to get test framework (2f5c7dc)
• unittest: add support for one package include multiple results (4bc3814)
• unittest: enable get SCA from context (5c64dcc)

Datasets:

2000 Codgen + 2000 TestGen + 2000 OSS Instruct