vijaye12 commited on
Commit
fac05af
1 Parent(s): 39239a0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -6
README.md CHANGED
@@ -12,20 +12,19 @@ can be easily fine-tuned for your target data. Refer to our [paper](https://arxi
12
 
13
  **Note that zeroshot, fine-tuning and inference tasks using TTM can easily be executed in 1 GPU machine or in laptops too!!**
14
 
 
15
 
16
  ## Benchmark Highlights:
17
 
18
  - TTM outperforms pre-trained GPT4TS (NeurIPS 23) by 7-12% in few-shot forecasting.
19
  - TTM outperforms pre-trained LLMTime (NeurIPS 23) by 24% in zero-shot forecasting.
20
  - TTM outperforms pre-trained SimMTM (NeurIPS 23) by 17% in few-shot forecasting.
21
- - TTM drastically reduces the compute needs as compared to the LLM-TS pre-training methods, with a 14X cut in
22
- learnable parameters, 106X less total parameters, and substantial reductions in fine-tuning (65X),
23
- inference time (54X), and memory usage (27X).
24
- Zero-shot results of TTM often surpass the few-shot results of many SOTA approaches including
25
  PatchTST (ICLR 23), PatchTSMixer (KDD 23), TimesNet (ICLR 23), DLinear (AAAI 23) and FEDFormer (ICML 22).
26
  - TTM (1024-96, released in this model card) also outperforms pre-trained MOIRAI on FL = 96 by ...
27
  - TTM quick fine-tuning also outperforms the hard statistical baselines (Statistical ensemble and S-Naive) in
28
  M4-hourly dataset which pretrained TS models are finding hard to outperform.
 
29
 
30
 
31
  ## Model Description
@@ -61,8 +60,8 @@ TTM-1 currently supports 2 modes:
61
 
62
  - Finetuned forecasting: Finetune the pre-trained model with your target data to further improve the forecast.
63
 
64
- **Since, TTM models are extremely small and fast, it is practically very easy to finetune the model with your available target data to
65
- get more accurate forecasts.**
66
 
67
  The current release supports multivariate forecasting via both channel independence and channel-mixing approaches.
68
  Decoder Channel-Mixing can be enabled during fine-tuning for capturing strong channel-correlation patterns across
 
12
 
13
  **Note that zeroshot, fine-tuning and inference tasks using TTM can easily be executed in 1 GPU machine or in laptops too!!**
14
 
15
+ TTM Few-shot tuning is
16
 
17
  ## Benchmark Highlights:
18
 
19
  - TTM outperforms pre-trained GPT4TS (NeurIPS 23) by 7-12% in few-shot forecasting.
20
  - TTM outperforms pre-trained LLMTime (NeurIPS 23) by 24% in zero-shot forecasting.
21
  - TTM outperforms pre-trained SimMTM (NeurIPS 23) by 17% in few-shot forecasting.
22
+ - Zero-shot results of TTM often surpass the few-shot results of many SOTA approaches including
 
 
 
23
  PatchTST (ICLR 23), PatchTSMixer (KDD 23), TimesNet (ICLR 23), DLinear (AAAI 23) and FEDFormer (ICML 22).
24
  - TTM (1024-96, released in this model card) also outperforms pre-trained MOIRAI on FL = 96 by ...
25
  - TTM quick fine-tuning also outperforms the hard statistical baselines (Statistical ensemble and S-Naive) in
26
  M4-hourly dataset which pretrained TS models are finding hard to outperform.
27
+ - TTM takes only a few seconds for zeroshot/inference and a few minutes for finetuning in 1 GPU machine.
28
 
29
 
30
  ## Model Description
 
60
 
61
  - Finetuned forecasting: Finetune the pre-trained model with your target data to further improve the forecast.
62
 
63
+ **Since, TTM models are extremely small and fast, it is practically very easy to finetune the model with your available target data in few minutes
64
+ to get more accurate forecasts.**
65
 
66
  The current release supports multivariate forecasting via both channel independence and channel-mixing approaches.
67
  Decoder Channel-Mixing can be enabled during fine-tuning for capturing strong channel-correlation patterns across