update hugchat

HugAILab · May 4, 2023 · 935c977 · 935c977
1 parent 16210a2
commit 935c977
Showing 1 changed file with 7 additions and 6 deletions.
diff --git a/documents/instruction_prompting/generative_instruction_tuning.md b/documents/instruction_prompting/generative_instruction_tuning.md
@@ -121,19 +121,20 @@ bash ./application/instruction_prompting/HugChat/supervised_finetuning/run_causa
 ## Pre-built HugChat Models
 
 We design HugChat application based on generative instruction-tuning.
-We have trained following models, and release the weights about HugChat:
+We have trained following models based on SFT, and release the weights at Huggingface:
 
 | Backbone | Size | Corpora | Config | Progress | Script | HuggingFace Model Link
 | --- | --- | --- | --- | --- | --- | --- |
 | GPT-2 | base (0.3B) | English | V100 8*32G | Finish | [run_causal_instruction_gpt2.sh](../../applications/instruction_prompting/HugChat/supervised_finetuning/run_casual_instruction_gpt2.sh) | [wjn1996/hugnlp-hugchat-gpt2](https://huggingface.co/wjn1996/hugnlp-hugchat-gpt2)
-| GPT-2 | large (0.8B) | English | V100 8*32G | Finish | [run_causal_instruction_gpt2.sh](../../applications/instruction_prompting/HugChat/supervised_finetuning/run_casual_instruction_gpt2.sh) |
+| GPT-2 | large (0.8B) | English | V100 8*32G | Finish | [run_causal_instruction_gpt2.sh](../../applications/instruction_prompting/HugChat/supervised_finetuning/run_casual_instruction_gpt2.sh) | [wjn1996/hugnlp-hugchat-gpt2-large](https://huggingface.co/wjn1996/hugnlp-hugchat-gpt2-large)
 | GPT-2 | xlarge (1.3B) | English | V100 8*32G | Finish | [run_causal_instruction_gpt2_xl.sh]((../../applications/instruction_prompting/HugChat/supervised_finetuning/run_casual_instruction_gpt2_xl.sh)) | [wjn1996/hugnlp-hugchat-gpt2-xl](https://huggingface.co/wjn1996/hugnlp-hugchat-gpt2-xl)
-| OPT | 1.3B | English | V100 8*32G LoRA (dim=8) | Finish | [run_causal_instruction_opt.sh]((../../applications/instruction_prompting/HugChat/supervised_finetuning/run_casual_instruction_opt.sh)) |
+| OPT | 1.3B | English | V100 8*32G LoRA (dim=8) | Finish | [run_causal_instruction_opt.sh]((../../applications/instruction_prompting/HugChat/supervised_finetuning/run_casual_instruction_opt.sh)) | [wjn1996/hugnlp-hugchat-opt-1.3b](https://huggingface.co/wjn1996/hugnlp-hugchat-opt-1.3b)
 | OPT | 6.7B | English | V100 8*32G ZeRO-3 FP16 LoRA (dim=8) | Finish | [run_causal_instruction_opt_lora.sh]((../../applications/instruction_prompting/HugChat/supervised_finetuning/run_causal_instruction_opt_lora.sh)) |
+| OPT | 13B | English | V100 8*32G ZeRO-3 FP16 LoRA (dim=8) | Developing | [run_causal_instruction_opt_lora.sh]((../../applications/instruction_prompting/HugChat/supervised_finetuning/run_causal_instruction_opt_lora.sh)) |
 | GLM-2B | 2.0B | English | V100 8*32G | Pending | |
-| GPT-Neo | 1.3B | English | V100 8*32G ZeRO-1 FP16 | Finish | [run_causal_instruction_gpt_neo.sh](../../applications/instruction_prompting/HugChat/supervised_finetuning/run_causal_instruction_gpt_neo.sh) | [wjn1996/hugnlp-hugchat-gpt-neo-1.3B](https://huggingface.co/wjn1996/hugnlp-hugchat-gpt-neo-1.3B) |
-| GPT-Neo | 2.7B | English | V100 8*32G ZeRO-3 FP16 | Developing | [run_causal_instruction_gpt_neo.sh](../../applications/instruction_prompting/HugChat/supervised_finetuning/run_causal_instruction_gpt_neo.sh) |
-| LLaMA | 7B | English | V100 8*32G | Pending | |
+| GPT-Neo | 1.3B | English | V100 8*32G ZeRO-1 FP16 | Finish | [run_causal_instruction_gpt_neo.sh](../../applications/instruction_prompting/HugChat/supervised_finetuning/run_causal_instruction_gpt_neo.sh) | [wjn1996/hugnlp-hugchat-gpt-neo-1.3B](https://huggingface.co/wjn1996/hugnlp-hugchat-gpt-neo-1.3B) | [wjn1996/hugnlp-hugchat-gpt-neo-1.3B](https://huggingface.co/wjn1996/hugnlp-hugchat-gpt-neo-1.3B)
+| GPT-Neo | 2.7B | English | V100 8*32G ZeRO-3 FP16 | Finish | [run_causal_instruction_gpt_neo.sh](../../applications/instruction_prompting/HugChat/supervised_finetuning/run_causal_instruction_gpt_neo.sh) | [wjn1996/hugnlp-hugchat-gpt-neo-2.7B](https://huggingface.co/wjn1996/hugnlp-hugchat-gpt-neo-2.7B)
+| LLaMA | 7B | English | V100 8*32G | Pending |  | 
 
 ---