github
/
InternLM
mirror of https://github.com/InternLM/InternLM
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

merge main

pull/325/head
yingtongxiong 2023-09-19 18:04:47 +08:00
parent 6a5915bf0d bfefc4ea3c
commit 30b21075e8
9 changed files with 100 additions and 66 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
.github/workflows/demo_in_readme.yaml vendored
View File

@ -9,11 +9,11 @@ on:
- "**.md"
env:
WORKSPACE_PREFIX: $(echo $GITHUB_WORKSPACE |cut -d '/' -f 1-4)
SLURM_PARTITION: llm
SLURM_PARTITION: llm_s
jobs:
check-requirements:
runs-on: [lmtest]
runs-on: [t_cluster]
steps:
- name: mask env
run: |
@ -37,7 +37,7 @@ jobs:
dataset-preparation:
if: ${{ always() }}
needs: check-requirements
runs-on: [lmtest]
runs-on: [t_cluster]
steps:
- name: mask env
run: |
@ -57,7 +57,7 @@ jobs:
train:
if: ${{ always() }}
needs: check-requirements
runs-on: [lmtest]
runs-on: [t_cluster]
timeout-minutes: 30
steps:
- name: mask env
@ -83,18 +83,19 @@ jobs:
source activate internlm-env-test
export PYTHONPATH=$PWD:$PYTHONPATH
sh ./ci_scripts/train/load_ckpt.sh 7B_load_new_ckpt ${GITHUB_RUN_ID}-${GITHUB_JOB}
rm -rf $GITHUB_WORKSPACE/llm_ckpts
rsync -av --remove-source-files $GITHUB_WORKSPACE/llm_ckpts ${{env.WORKSPACE_PREFIX}}/ci_clean_bak
- name: torchrun-train
run: |
source activate internlm-env-test
sh ./ci_scripts/train/torchrun.sh ${GITHUB_RUN_ID}-${GITHUB_JOB}
rm -rf $GITHUB_WORKSPACE/llm_ckpts
rsync -av --remove-source-files $GITHUB_WORKSPACE/llm_ckpts ${{env.WORKSPACE_PREFIX}}/ci_clean_bak
convert-model-then-load:
if: ${{ always() }}
needs: check-requirements
runs-on: [lmtest]
runs-on: [t_cluster]
timeout-minutes: 15
steps:
- name: mask env
run: |
@ -107,13 +108,14 @@ jobs:
export PYTHONPATH=$PWD:$PYTHONPATH
sh ./ci_scripts/model/convert_to_hf.sh
cd ./hf_ckpt
srun -p ${SLURM_PARTITION} --job-name=${GITHUB_RUN_ID}-${GITHUB_JOB} --gpus-per-task=2 python ../ci_scripts/model/loaded_as_transformer.py
srun -p ${SLURM_PARTITION} --quotatype=spot --job-name=${GITHUB_RUN_ID}-${GITHUB_JOB} --gpus-per-task=2 python ../ci_scripts/model/loaded_as_transformer.py
cd ..
rm -rf $GITHUB_WORKSPACE/hf_ckpt
rsync -av --remove-source-files $GITHUB_WORKSPACE/hf_ckpt ${{env.WORKSPACE_PREFIX}}/ci_clean_bak
load-chat-model-in-hf:
if: ${{ always() }}
needs: check-requirements
runs-on: [lmtest]
runs-on: [t_cluster]
timeout-minutes: 15
steps:
- name: mask env
run: |
@ -123,4 +125,4 @@ jobs:
- name: chat-model-in-hf
run: |
source activate internlm-env-test
srun -p ${SLURM_PARTITION} --job-name=${GITHUB_RUN_ID}-${GITHUB_JOB} --gpus-per-task=2 python ./ci_scripts/model/demo_load_7B_chat_model.py
srun -p ${SLURM_PARTITION} --quotatype=spot --job-name=${GITHUB_RUN_ID}-${GITHUB_JOB} --gpus-per-task=2 python ./ci_scripts/model/demo_load_7B_chat_model.py

1
ci_scripts/common/variables.sh
View File

@ -1,3 +1,4 @@
#!/bin/bash
readonly DATA_VOLUME=$(echo $GITHUB_WORKSPACE | cut -d '/' -f 1-4)/data
readonly CLEAN_PATH=$(echo $GITHUB_WORKSPACE | cut -d '/' -f 1-4)/ci_clean_bak

7
ci_scripts/data/tokenizer_alpaca.sh
View File

@ -3,6 +3,7 @@ set -x
source ./ci_scripts/common/variables.sh
[[ -n ${DATA_VOLUME} ]] || { echo "should set DATA_VOLUME first before ci, exit."; exit 1; }
[[ -n ${CLEAN_PATH} ]] || { echo "should set CLEAN_PATH first before ci, exit."; exit 1; }
readonly SRC_DATASET_META=${DATA_VOLUME}/lm_data/alpaca_data/alpaca_data.json
readonly RESULTS=${DATA_VOLUME}/lm_data/alpaca_data/result
@ -19,7 +20,7 @@ source ./ci_scripts/common/basic_func.sh
echo "start to test alpaca_tokenizer.py."
if [[ -d ${RESULTS} ]]; then
if ! rm -rf ${RESULTS}/*; then
if ! rsync -av --remove-source-files ${RESULTS} ${CLEAN_PATH}; then
echo "cleaning test data in ${RESULTS} failed, exit."
exit 1
fi
@ -41,8 +42,8 @@ for file in ${file_list[@]}; do
fi
done
# clean the test files.
if ! rm -rf ${RESULTS}/*; then
# move the test files.
if ! rsync -av --remove-source-files ${RESULTS} ${CLEAN_PATH}; then
echo "cleaning test data in ${RESULTS} failed."
exit_code=$(($exit_code + 1))
fi

11
ci_scripts/data/tokenizer_chinese.sh
View File

@ -2,7 +2,8 @@
set -x
source ./ci_scripts/common/variables.sh
[[ -n ${DATA_VOLUME} ]] || { echo "should set DATA_VOLUME first before ci."; exit 1; }
[[ -n ${DATA_VOLUME} ]] || { echo "should set DATA_VOLUME first before ci, exit."; exit 1; }
[[ -n ${CLEAN_PATH} ]] || { echo "should set CLEAN_PATH first before ci, exit."; exit 1; }
readonly DATA=${DATA_VOLUME}/lm_data/cn_data/raw_data.txt
readonly RESULT=${DATA_VOLUME}/lm_data/cn_data/result.bin
@ -16,13 +17,13 @@ echo "start to test tokenizer.py."
num=$(num_files "${RESULTS}")
if [[ ${num} -gt 0 ]]; then
if ! rm -rf ${RESULTS}; then
if ! rsync -av --remove-source-files ${RESULTS} ${CLEAN_PATH}; then
echo "cleaning test data ${RESULTS} failed, exit."
exit 1
fi
fi
srun -p ${SLURM_PARTITION} --job-name=$1 --gpus-per-task=1 python tools/tokenizer.py --text_input_path ${DATA} --bin_output_path ${RESULT}
srun -p ${SLURM_PARTITION} --quotatype=spot --job-name=$1 --gpus-per-task=1 python tools/tokenizer.py --text_input_path ${DATA} --bin_output_path ${RESULT}
[[ $? -ne 0 ]] && { echo "test tokenizer.py failed."; exit_code=$(($exit_code + 1)); }
file_list=($RESULT $RESULT_META)
@ -33,8 +34,8 @@ for file in ${file_list[@]}; do
fi
done
# clean the test files.
if ! rm -rf ${RESULTS}/*; then
# move the test files.
if ! rsync -av --remove-source-files ${RESULTS} ${CLEAN_PATH}; then
echo "cleaning cached file in ${RESULTS} failed."
exit_code=$(($exit_code + 1))
fi

3
ci_scripts/model/convert_to_hf.sh
View File

@ -4,6 +4,7 @@ set -x
source ./ci_scripts/common/variables.sh
[[ -n ${DATA_VOLUME} ]] || { echo "should set DATA_VOLUME first before ci, exit."; exit 1; }
[[ -n ${GITHUB_WORKSPACE} ]] || { echo "should set GITHUB_WORKSPACE first before ci, exit."; exit 1; }
[[ -n ${CLEAN_PATH} ]] || { echo "should set CLEAN_PATH first before ci, exit."; exit 1; }
readonly CKPTS_INPUT="${DATA_VOLUME}/lm_data/alpaca_data/llm_ckpts/20"
readonly CKPTS_OUTPUT="${GITHUB_WORKSPACE}/hf_ckpt"
@ -18,7 +19,7 @@ source ./ci_scripts/common/basic_func.sh
echo "start to test convert2hf.py."
if [[ -d ${CKPTS_OUTPUT} ]]; then
if ! rm -rf ${CKPTS_OUTPUT}/*; then
if ! rsync -av --remove-source-files ${CKPTS_OUTPUT}/* ${CLEAN_PATH}; then
echo "cleaning cached file in ${CKPTS_OUTPUT} failed, exit."
exit 1
fi

15
ci_scripts/train/load_ckpt.sh
View File

@ -1,7 +1,10 @@
#!/bin/bash
set -x
source ./ci_scripts/common/variables.sh
[[ -n ${GITHUB_WORKSPACE} ]] || { echo "should set GITHUB_WORKSPACE first before ci, exit."; exit 1; }
[[ -n ${CLEAN_PATH} ]] || { echo "should set CLEAN_PATH first before ci, exit."; exit 1; }
readonly CKPTS_PATH="$GITHUB_WORKSPACE/llm_ckpts"
readonly CKPTS40_PATH="$GITHUB_WORKSPACE/llm_ckpts/40"
readonly CKPTS40_OUTPUT="${CKPTS40_PATH}/*.pt"
@ -19,7 +22,7 @@ if [[ ! -f ${file} ]]; then
exit_code=$(($exit_code + 1))
fi
srun -p ${SLURM_PARTITION} --exclusive --job-name=$2 -n 8 --ntasks-per-node=8 --gpus-per-task=1 python train.py --config ${file}
srun -p ${SLURM_PARTITION} --exclusive --quotatype=spot --job-name=$2 -n 8 --ntasks-per-node=8 --gpus-per-task=1 python train.py --config ${file}
[[ $? -ne 0 ]] && { echo "test slurm training failed."; exit_code=$(($exit_code + 1)); }
@ -29,10 +32,12 @@ if [[ ${num} -ne ${expected_num} ]]; then
exit_code=$(($exit_code + 1))
fi
# clean the test files.
if ! rm -rf ${CKPTS_PATH}/*; then
echo "cleaning cached file in ${CKPTS_PATH} failed."
exit_code=$(($exit_code + 1))
# move the test files.
if [[ -d ${CKPTS_PATH} ]]; then
if ! rsync -av --remove-source-files ${CKPTS_PATH} ${CLEAN_PATH}; then
echo "cleaning cached file in ${CKPTS_PATH} failed."
exit_code=$(($exit_code + 1))
fi
fi
exit $exit_code

7
ci_scripts/train/slurm_train.sh
View File

@ -1,7 +1,10 @@
#!/bin/bash
set -x
source ./ci_scripts/common/variables.sh
[[ -n ${GITHUB_WORKSPACE} ]] || { echo "should set GITHUB_WORKSPACE first before ci, exit."; exit 1; }
[[ -n ${CLEAN_PATH} ]] || { echo "should set CLEAN_PATH first before ci, exit."; exit 1; }
readonly CKPTS_PATH="$GITHUB_WORKSPACE/llm_ckpts"
readonly CKPTS20_PATH="$GITHUB_WORKSPACE/llm_ckpts/20"
readonly CKPTS20_OUTPUT="${CKPTS20_PATH}/*.pt"
@ -13,13 +16,13 @@ source ./ci_scripts/common/basic_func.sh
echo "start to test slurm training."
if [[ -d ${CKPTS20_PATH} ]]; then
if ! rm -rf ${CKPTS20_PATH}/*; then
if ! rsync -av --remove-source-files ${CKPTS20_PATH} ${CLEAN_PATH}; then
echo "cleaning cached file in ${CKPTS20_PATH} failed, exit."
exit 1
fi
fi
srun -p ${SLURM_PARTITION} --exclusive --job-name=$1 -n 8 --ntasks-per-node=8 --gpus-per-task=1 python train.py --config ./ci_scripts/train/ci_7B_sft.py
srun -p ${SLURM_PARTITION} --exclusive --quotatype=spot --job-name=$1 -n 8 --ntasks-per-node=8 --gpus-per-task=1 python train.py --config ./ci_scripts/train/ci_7B_sft.py
[[ $? -ne 0 ]] && { echo "test slurm training failed."; exit_code=$(($exit_code + 1)); }
num=$(num_files "${CKPTS20_OUTPUT}")

11
ci_scripts/train/torchrun.sh
View File

@ -1,7 +1,10 @@
#!/bin/bash
set -x
source ./ci_scripts/common/variables.sh
[[ -n ${GITHUB_WORKSPACE} ]] || { echo "should set GITHUB_WORKSPACE first before ci, exit."; exit 1; }
[[ -n ${CLEAN_PATH} ]] || { echo "should set CLEAN_PATH first before ci, exit."; exit 1; }
readonly CKPTS_PATH="$GITHUB_WORKSPACE/llm_ckpts"
readonly CKPTS20_PATH="$GITHUB_WORKSPACE/llm_ckpts/20"
readonly CKPTS_OUTPUT="${CKPTS20_PATH}/*.pt"
@ -13,13 +16,13 @@ source ./ci_scripts/common/basic_func.sh
echo "start to test torch training."
if [[ -d ${CKPTS20_PATH} ]]; then
if ! rm -rf ${CKPTS20_PATH}/*; then
if ! rsync -av --remove-source-files ${CKPTS20_PATH} ${CLEAN_PATH}; then
echo "cleaning cached file in ${CKPTS20_PATH} failed, exit."
exit 1
fi
fi
srun -p ${SLURM_PARTITION} --exclusive --job-name=$1 -N 1 torchrun --nnodes=1 --nproc_per_node=8 --master_port=29501 train.py --config ./ci_scripts/train/ci_7B_sft.py --launcher torch
srun -p ${SLURM_PARTITION} --exclusive --quotatype=spot --job-name=$1 -N 1 torchrun --nnodes=1 --nproc_per_node=8 --master_port=29501 train.py --config ./ci_scripts/train/ci_7B_sft.py --launcher torch
[[ $? -ne 0 ]] && { echo "test torch training failed."; exit_code=$(($exit_code + 1)); }
num=$(num_files "${CKPTS_OUTPUT}")
@ -28,8 +31,8 @@ if [[ ${num} -ne ${expected_num} ]]; then
exit_code=$(($exit_code + 1))
fi
# clean the test files.
if ! rm -rf ${CKPTS_PATH}/*; then
# move the test files.
if ! rsync -av --remove-source-files ${CKPTS_PATH}/* ${CLEAN_PATH}; then
echo "cleaning cached file in ${CKPTS_PATH} failed."
exit_code=$(($exit_code + 1))
fi

87
tools/transformers/modeling_internlm.py
View File

@ -28,10 +28,19 @@ from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
from transformers.activations import ACT2FN
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast
from transformers.modeling_outputs import (
BaseModelOutputWithPast,
CausalLMOutputWithPast,
SequenceClassifierOutputWithPast,
)
from transformers.modeling_utils import PreTrainedModel
from transformers.generation.streamers import BaseStreamer
from transformers.utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from transformers.utils import (
add_start_docstrings,
add_start_docstrings_to_model_forward,
logging,
replace_return_docstrings,
)
from configuration_internlm import InternLMConfig
@ -39,6 +48,7 @@ logger = logging.get_logger(__name__)
_CONFIG_FOR_DOC = "InternLMConfig"
# Copied from transformers.models.bart.modeling_bart._make_causal_mask
def _make_causal_mask(
input_ids_shape: torch.Size, dtype: torch.dtype, device: torch.device, past_key_values_length: int = 0
@ -437,6 +447,7 @@ class InternLMModel(InternLMPreTrainedModel):
Args:
config: InternLMConfig
"""
_auto_class = "AutoModel"
def __init__(self, config: InternLMConfig):
@ -765,7 +776,7 @@ class InternLMForCausalLM(InternLMPreTrainedModel):
for layer_past in past_key_values:
reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
return reordered_past
def build_inputs(self, tokenizer, query: str, history: List[Tuple[str, str]] = []):
prompt = ""
for record in history:
@ -774,43 +785,49 @@ class InternLMForCausalLM(InternLMPreTrainedModel):
prompt += "<s>"
prompt += f"""<|User|>:{query}<eoh>\n<|Bot|>:"""
return tokenizer([prompt], return_tensors="pt")
@torch.no_grad()
def chat(self,
tokenizer,
query: str,
history: List[Tuple[str, str]] = [],
streamer: Optional[BaseStreamer] = None,
max_new_tokens: int = 1024,
do_sample: bool = True,
temperature: float = 0.8,
top_p: float = 0.8,
**kwargs):
def chat(
self,
tokenizer,
query: str,
history: List[Tuple[str, str]] = [],
streamer: Optional[BaseStreamer] = None,
max_new_tokens: int = 1024,
do_sample: bool = True,
temperature: float = 0.8,
top_p: float = 0.8,
**kwargs,
):
inputs = self.build_inputs(tokenizer, query, history)
inputs = {k: v.to(self.device) for k, v in inputs.items() if torch.is_tensor(v)}
outputs = self.generate(**inputs,
streamer=streamer,
max_new_tokens=max_new_tokens,
do_sample=do_sample,
temperature=temperature,
top_p=top_p,
**kwargs)
outputs = outputs[0].cpu().tolist()[len(inputs["input_ids"][0]):]
outputs = self.generate(
**inputs,
streamer=streamer,
max_new_tokens=max_new_tokens,
do_sample=do_sample,
temperature=temperature,
top_p=top_p,
**kwargs,
)
outputs = outputs[0].cpu().tolist()[len(inputs["input_ids"][0]) :]
response = tokenizer.decode(outputs, skip_special_tokens=True)
response = response.split("<eoa>")[0]
history = history + [(query, response)]
return response, history
@torch.no_grad()
def stream_chat(self,
tokenizer,
query: str,
history: List[Tuple[str, str]] = [],
max_new_tokens: int = 1024,
do_sample: bool = True,
temperature: float = 0.8,
top_p: float = 0.8,
**kwargs):
def stream_chat(
self,
tokenizer,
query: str,
history: List[Tuple[str, str]] = [],
max_new_tokens: int = 1024,
do_sample: bool = True,
temperature: float = 0.8,
top_p: float = 0.8,
**kwargs,
):
"""
Return a generator in format: (response, history)
Eg.
@ -856,12 +873,12 @@ class InternLMForCausalLM(InternLMPreTrainedModel):
tokenizer=tokenizer,
query=query,
streamer=ChatStreamer(tokenizer=tokenizer),
history=history,
history=history,
max_new_tokens=max_new_tokens,
do_sample=do_sample,
temperature=temperature,
top_p=top_p,
**kwargs
**kwargs,
)
def consumer():
@ -869,7 +886,7 @@ class InternLMForCausalLM(InternLMPreTrainedModel):
producer.start()
while True:
res = response_queue.get()
if res is None:
if res is not None:
return
yield res