yangjianxin1 Firefly issues

yangjianxin1 / Firefly

Firefly: 大模型训练工具，支持训练Qwen2.5、Qwen2、Yi1.5、Phi-3、Llama3、Gemma、MiniCPM、Yi、Deepseek、Orion、Xverse、Mixtral-8x7B、Zephyr、Mistral、Baichuan2、Llma2、Llama、Qwen、Baichuan、ChatGLM2、InternLM、Ziya2、Vicuna、Bloom等大模型

5.89k stars 526 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

max_grad_norm不生效的问题

#304 yiyepiaoling0715 opened 2 weeks ago
1
整合LIger Kernel多卡提速20%減低顯存60% 能和Deepspeed兼容

#303 ByronHsu opened 2 weeks ago
0
增加了minicpm3的sft、pretrain、dpo的配置参数

#302 LDLINGLINGLING closed 1 month ago
1
qwen2 pretrain loss非常大

#301 ucaslei closed 4 weeks ago
0
目前调用unsloth是只能支持单卡吗

#300 MinusOne2 opened 2 months ago
1
关于参数remove_unused_columns = false

#299 Kenneth0901 opened 2 months ago
0
template中的system_format字段是必须的吗？对于微调来说，这个字段是不是可以不用，从节省输入token数量的角度来说

#298 Chris-Mraz opened 2 months ago
0
设置了双卡，却信息任然显示单卡。怎么回事？

#297 youzihaha opened 3 months ago
0
多卡load模型OOM

#296 TonyUSTC opened 3 months ago
0
知识蒸馏

#295 zjjznw123 opened 3 months ago
0
是否支持linux平台

#294 Scorponok31 closed 3 months ago
0
loss一直为0

#293 chk4991 opened 3 months ago
1
用qwen1.5 infer的时候他就疯狂写代码

#292 ExeCuteRunrunrun opened 3 months ago
0
有支持华为昇腾npu的计划吗

#291 bestRiven opened 3 months ago
0
如何在训练语料中添加system

#290 louyuanyuan053709 opened 4 months ago
1
llama 3.1, DCLM支持

#288 WinterStraw opened 4 months ago
0
请问有支持GLM4-9B-Chat SFT 的计划吗？

#287 Candy555 opened 4 months ago
0
category 这个字段在推理的时候怎么用

#286 louyuanyuan053709 opened 4 months ago
0
modifiy cache path rule for pretrain datasets to fix bug

#285 ba5bo opened 4 months ago
0
有支持internvl 的计划么？

#284 moyans opened 4 months ago
0
请问我运行“python train.py --train_args_file train_args/sft/qlora/baichuan-7b-sft-qlora.json”指令，为什么报一下错误呀？

#283 lh5533223 opened 4 months ago
1
关于蒸馏训练的代码

#282 rattlesnakey opened 4 months ago
4
训练baichuan-13B qlora, 发现loss查看为空，怎么解决

#281 lovegit2021 opened 4 months ago
0
目前unsloth是不是不支持qwen2-moe结构？

#280 xiaoer2498 opened 4 months ago
0
微调Qwen2-1.5B-Instruct，loss始终是0

#279 frederichen01 opened 4 months ago
7
DPO的训练能否像SFT一样mask掉user，只计算assistant的loss

#278 charliedream1 opened 4 months ago
5
现在是否支持微调后大模型的单机多卡推理。

#277 SunRise-Star opened 4 months ago
0
有计划支持deepseek-coder-v2-lite的微调吗

#276 fengyang95 opened 4 months ago
0
可以支持 llava 微调吗？

#275 vincent507cpu opened 4 months ago
0
expected string or bytes-like object

#274 sankexin opened 5 months ago
0
qwen2-7b lora loss为0

#273 fengyang95 closed 5 months ago
1
Qwen2-7B-Instruct 训练loss 0 或推理 probability tensor contains either `inf`, `nan` or element < 0

#272 WinterStraw closed 5 months ago
6
请问支持codestral微调和部署吗？

#271 SunRise-Star opened 5 months ago
0
ValueError: You can't train a model that has been loaded in 8-bit precision on a different device than the one you're training on.

#270 WeixuanXiong opened 5 months ago
1
Qwen2-7b lora

#269 linzm1007 opened 5 months ago
0
add Qwen2

#268 yangjianxin1 closed 5 months ago
0
为什么llama2的template中没有bos_token

#267 qy1026 opened 5 months ago
0
使用Unsloth反而OOM了是为什么呢？

#266 yuyu990116 closed 5 months ago
2
NameError: name 'FastLanguageModel' is not defined

#265 Jackiexiong opened 5 months ago
0
“max_seq_length：训练时的最大长度。按照自己的设备进行设置，越长需要占用越多显存。”这个有换算规则吗？

#264 sunzx8 opened 6 months ago
0
unsloth不支持多机多卡，而且对比的还是没有开启flashattn2的增益，有些许疑问

#263 MonolithFoundation opened 6 months ago
0
DPO支持多卡的QLora训练吗？

#262 q497629642 opened 6 months ago
0
启动chat.py脚本加载merge_lora生成的checkpoint模型或者adapter参数都会报错

#261 2013303386 opened 6 months ago
0
add yi-1.5

#260 yangjianxin1 closed 6 months ago
0
RuntimeError: FlashAttention only support fp16 and bf16 data type

#259 sankexin opened 6 months ago
1
NotImplementedError: No operator found for `memory_efficient_attention_forward`

#258 sankexin opened 6 months ago
0
kto训练实现

#257 vincezengqiang opened 6 months ago
0
训练llama3-8b-it报错

#256 wx971025 opened 6 months ago
2
deepspeed安装报错

#255 TC10127 opened 6 months ago
1
多轮对话训练效果

#254 chanel111 opened 6 months ago
1