From c2d4717be2bd44deb6733d819628c3b1de9cee2f Mon Sep 17 00:00:00 2001 From: Jingming <109033042+jingmingzhuo@users.noreply.github.com> Date: Fri, 15 Mar 2024 15:21:35 +0800 Subject: [PATCH] [Fix] Fix a bug in internlm2 series configs (#977) --- configs/models/hf_internlm/hf_internlm2_chat_1_8b.py | 1 + configs/models/hf_internlm/hf_internlm2_chat_1_8b_sft.py | 1 + configs/models/hf_internlm/hf_internlm2_chat_20b.py | 1 + configs/models/hf_internlm/hf_internlm2_chat_20b_sft.py | 1 + .../models/hf_internlm/hf_internlm2_chat_20b_with_system.py | 3 ++- configs/models/hf_internlm/hf_internlm2_chat_7b.py | 1 + configs/models/hf_internlm/hf_internlm2_chat_7b_sft.py | 1 + configs/models/hf_internlm/hf_internlm2_chat_7b_with_system.py | 3 ++- 8 files changed, 10 insertions(+), 2 deletions(-) diff --git a/configs/models/hf_internlm/hf_internlm2_chat_1_8b.py b/configs/models/hf_internlm/hf_internlm2_chat_1_8b.py index b5c109fe..48daa8e3 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_1_8b.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_1_8b.py @@ -31,5 +31,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=1, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ] diff --git a/configs/models/hf_internlm/hf_internlm2_chat_1_8b_sft.py b/configs/models/hf_internlm/hf_internlm2_chat_1_8b_sft.py index 09a19e58..24823a07 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_1_8b_sft.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_1_8b_sft.py @@ -31,5 +31,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=1, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ] diff --git a/configs/models/hf_internlm/hf_internlm2_chat_20b.py b/configs/models/hf_internlm/hf_internlm2_chat_20b.py index b017fe41..7d6515c3 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_20b.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_20b.py @@ -31,5 +31,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=2, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ] diff --git a/configs/models/hf_internlm/hf_internlm2_chat_20b_sft.py b/configs/models/hf_internlm/hf_internlm2_chat_20b_sft.py index ce31d29b..bde3c35c 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_20b_sft.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_20b_sft.py @@ -31,5 +31,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=2, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ] diff --git a/configs/models/hf_internlm/hf_internlm2_chat_20b_with_system.py b/configs/models/hf_internlm/hf_internlm2_chat_20b_with_system.py index 02d2d683..29373777 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_20b_with_system.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_20b_with_system.py @@ -3,8 +3,8 @@ from opencompass.models import HuggingFaceCausalLM _meta_template = dict( round=[ - dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'), dict(role='SYSTEM', begin='<|im_start|>system\n', end='<|im_end|>\n'), + dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'), dict(role='BOT', begin='<|im_start|>assistant\n', end='<|im_end|>\n', generate=True), ], eos_token_id=92542 @@ -32,5 +32,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=2, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ] diff --git a/configs/models/hf_internlm/hf_internlm2_chat_7b.py b/configs/models/hf_internlm/hf_internlm2_chat_7b.py index 8cfb680f..14cee114 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_7b.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_7b.py @@ -31,5 +31,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=1, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ] diff --git a/configs/models/hf_internlm/hf_internlm2_chat_7b_sft.py b/configs/models/hf_internlm/hf_internlm2_chat_7b_sft.py index f7d2d190..fe3d5e7e 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_7b_sft.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_7b_sft.py @@ -31,5 +31,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=1, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ] diff --git a/configs/models/hf_internlm/hf_internlm2_chat_7b_with_system.py b/configs/models/hf_internlm/hf_internlm2_chat_7b_with_system.py index ba67d6da..cd000115 100644 --- a/configs/models/hf_internlm/hf_internlm2_chat_7b_with_system.py +++ b/configs/models/hf_internlm/hf_internlm2_chat_7b_with_system.py @@ -3,8 +3,8 @@ from opencompass.models import HuggingFaceCausalLM _meta_template = dict( round=[ - dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'), dict(role='SYSTEM', begin='<|im_start|>system\n', end='<|im_end|>\n'), + dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'), dict(role='BOT', begin='<|im_start|>assistant\n', end='<|im_end|>\n', generate=True), ], eos_token_id=92542 @@ -32,5 +32,6 @@ models = [ meta_template=_meta_template, run_cfg=dict(num_gpus=1, num_procs=1), end_str='<|im_end|>', + generation_kwargs = {"eos_token_id": [2, 92542]}, ) ]