Merge pull request #48 from lehors/update_models2

Refresh models
lfai · Nov 14, 2024 · b7c4f40 · b7c4f40
2 parents c8e5b77 + 7fa4f41
commit b7c4f40
Show file tree

Hide file tree

Showing 128 changed files with 546 additions and 383 deletions.
diff --git a/models/Amber.yml b/models/Amber.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: Amber
   version: 7B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: Amber
   producer: LLM360
   contact: ''
+  github: 'https://github.com/LLM360/amber-train'
+  huggingface: 'https://huggingface.co/LLM360/Amber'
   components:
     -
       name: 'Model architecture'

diff --git a/models/Arctic-Base.yml b/models/Arctic-Base.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: Arctic-Base
   version: 480B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: Arctic-Base
   producer: Snowflake
   contact: ''
+  huggingface: 'https://huggingface.co/Snowflake/snowflake-arctic-base'
   components:
     -
       name: 'Model architecture'

diff --git a/models/Arctic-Instruct.yml b/models/Arctic-Instruct.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: Arctic-Instruct
   version: 480B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: Arctic-Base
   producer: Snowflake
   contact: ''
+  huggingface: 'https://huggingface.co/Snowflake/snowflake-arctic-instruct'
   components:
     -
       name: 'Model architecture'

diff --git a/models/BLOOM.yml b/models/BLOOM.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: BLOOM
   version: 176B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ''
   producer: BigScience
   contact: ''
+  huggingface: 'https://huggingface.co/bigscience/bloom'
   components:
     -
       name: 'Model architecture'

diff --git a/models/BTLM-3B-8k-base-.yml b/models/BTLM-3B-8k-base-.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: BTLM-3B-8k-base-
   version: 3B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ''
   producer: Cerebras
   contact: ''
+  huggingface: 'https://huggingface.co/cerebras/btlm-3b-8k-base'
   components:
     -
       name: 'Model architecture'

diff --git a/models/Baichuan-1-13B.yml b/models/Baichuan-1-13B.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: Baichuan-1-13B
   version: 13B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: Baichuan-1
   producer: 'Baichuan AI'
   contact: ''
+  github: 'https://github.com/baichuan-inc/Baichuan-13B'
+  huggingface: 'https://huggingface.co/baichuan-inc/Baichuan-13B'
   components:
     -
       name: 'Model architecture'

diff --git a/models/Baichuan-1-7B.yml b/models/Baichuan-1-7B.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: Baichuan-1-7B
   version: 7B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: Baichuan-1
   producer: 'Baichuan AI'
   contact: ''
+  github: 'https://github.com/baichuan-inc/Baichuan-7B/tree/main'
+  huggingface: 'https://huggingface.co/baichuan-inc/Baichuan-7B'
   components:
     -
       name: 'Model architecture'

diff --git a/models/Bilingual-gpt-neox-4b.yml b/models/Bilingual-gpt-neox-4b.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: Bilingual-gpt-neox-4b
   version: 3.8B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: GPT-NeoX
   producer: Rinna
   contact: ''
+  huggingface: 'https://huggingface.co/rinna/bilingual-gpt-neox-4b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/BioMedGPT.yml b/models/BioMedGPT.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: BioMedGPT
   version: 2.7B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ''
   producer: 'Stanford CRFM'
   contact: ''
+  huggingface: 'https://huggingface.co/stanford-crfm/BioMedLM'
   components:
     -
       name: 'Model architecture'

diff --git a/models/BlackMamba-2.8B.yml b/models/BlackMamba-2.8B.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: BlackMamba-2.8B
   version: 2.8B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: ''
   origin: ''
   producer: Zyphra
   contact: ''
+  huggingface: 'https://huggingface.co/Zyphra/BlackMamba-2.8B'
   components:
     -
       name: 'Model architecture'

diff --git a/models/ByT5.yml b/models/ByT5.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: ByT5
   version: 13B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer encoder-decoder'
   origin: ''
   producer: Google
   contact: ''
+  huggingface: 'https://huggingface.co/google/byt5-xxl'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CerebrasGPT.yml b/models/CerebrasGPT.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: CerebrasGPT
   version: 13B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ''
   producer: Cerebras
   contact: ''
+  huggingface: 'https://huggingface.co/cerebras/Cerebras-GPT-13B'
   components:
     -
       name: 'Model architecture'

diff --git a/models/ChatGLM.yml b/models/ChatGLM.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: ChatGLM
   version: 6B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ''
   producer: 'Tsinghua University'
   contact: ''
+  huggingface: 'https://huggingface.co/THUDM/chatglm-6b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/ChatGLM2.yml b/models/ChatGLM2.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: ChatGLM2
   version: 6B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ChatGLM3-6B-Base
   producer: 'Tsinghua University'
   contact: ''
+  huggingface: 'https://huggingface.co/THUDM/chatglm2-6b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/ChatGLM3.yml b/models/ChatGLM3.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: ChatGLM3
   version: 6B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ChatGLM3-6B-Base
   producer: 'Tsinghua University'
   contact: ''
+  github: 'https://github.com/THUDM/ChatGLM3'
+  huggingface: 'https://huggingface.co/THUDM/chatglm3-6b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/Chinchilla.yml b/models/Chinchilla.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: Chinchilla
   version: 70B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: language
   architecture: 'transformer decoder'
   origin: ''
   producer: DeepMind
   contact: ''
+  github: 'https://github.com/jonnii/chinchilla'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeFuse.yml b/models/CodeFuse.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: CodeFuse
   version: 13B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer decoder'
   origin: ''
   producer: 'Ant Group'
   contact: ''
+  github: 'https://github.com/codefuse-ai'
+  huggingface: 'https://huggingface.co/codefuse-ai/CodeFuse-13B'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeGemma-2B.yml b/models/CodeGemma-2B.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: CodeGemma-2B
   version: 2B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer decoder'
   origin: Gemma
   producer: Google
   contact: ''
+  huggingface: 'https://huggingface.co/google/codegemma-7b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeGemma-7B-Instruct.yml b/models/CodeGemma-7B-Instruct.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: CodeGemma-7B-Instruct
   version: 7B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer decoder'
   origin: Gemma
   producer: Google
   contact: ''
+  huggingface: 'https://huggingface.co/google/codegemma-7b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeGemma-7B.yml b/models/CodeGemma-7B.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: CodeGemma-7B
   version: 7B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer decoder'
   origin: Gemma
   producer: Google
   contact: ''
+  huggingface: 'https://huggingface.co/google/codegemma-2b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeT5+-16B.yml b/models/CodeT5+-16B.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: CodeT5+-16B
   version: 16B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer encoder-decoder'
   origin: ''
   producer: Salesforce
   contact: ''
+  github: 'https://github.com/salesforce/CodeT5/tree/main/CodeT5%2B'
+  huggingface: 'https://huggingface.co/Salesforce/codet5p-16b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeT5+-220M.yml b/models/CodeT5+-220M.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: CodeT5+-220M
   version: 220M
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer encoder-decoder'
   origin: ''
   producer: Salesforce
   contact: ''
+  github: 'https://github.com/salesforce/CodeT5/tree/main/CodeT5%2B'
+  huggingface: 'https://huggingface.co/Salesforce/codet5p-16b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeT5+-2B.yml b/models/CodeT5+-2B.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: CodeT5+-2B
   version: 2B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer encoder-decoder'
   origin: ''
   producer: Salesforce
   contact: ''
+  github: 'https://github.com/salesforce/CodeT5/tree/main/CodeT5%2B'
+  huggingface: 'https://huggingface.co/Salesforce/codet5p-16b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeT5+-6B.yml b/models/CodeT5+-6B.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: CodeT5+-6B
   version: 6B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer encoder-decoder'
   origin: ''
   producer: Salesforce
   contact: ''
+  github: 'https://github.com/salesforce/CodeT5/tree/main/CodeT5%2B'
+  huggingface: 'https://huggingface.co/Salesforce/codet5p-16b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeT5+-770M.yml b/models/CodeT5+-770M.yml
@@ -5,12 +5,14 @@ framework:
 release:
   name: CodeT5+-770M
   version: 770M
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer encoder-decoder'
   origin: ''
   producer: Salesforce
   contact: ''
+  github: 'https://github.com/salesforce/CodeT5/tree/main/CodeT5%2B'
+  huggingface: 'https://huggingface.co/Salesforce/codet5p-16b'
   components:
     -
       name: 'Model architecture'

diff --git a/models/CodeT5.yml b/models/CodeT5.yml
@@ -5,12 +5,13 @@ framework:
 release:
   name: CodeT5
   version: 16B
-  date: '2024-10-03'
+  date: '2024-11-14'
   type: code
   architecture: 'transformer encoder-decoder'
   origin: ''
   producer: Salesforce
   contact: ''
+  github: 'https://github.com/salesforce/CodeT5/tree/main'
   components:
     -
       name: 'Model architecture'